L  L is a highlevel, opensource, generalpurpose and system programming language which emphasizes readability, simplicity, extensibility, conciseness and performance. The L compiler features native code generation through LLVM, and is fully documented in a literate programming style. The language and compiler are usable, but are under heavy development as new features are being implemented. 
L1Norm Batch Normalization (L1BN) 
Batch Normalization (BN) has been proven to be quite effective at accelerating and improving the training of deep neural networks (DNNs). However, BN brings additional computation, consumes more memory and generally slows down the training process by a large margin, which aggravates the training effort. Furthermore, the nonlinear square and root operations in BN also impede the low bitwidth quantization techniques, which draws much attention in deep learning hardware community. In this work, we propose an L1norm BN (L1BN) with only linear operations in both the forward and the backward propagations during training. L1BN is shown to be approximately equivalent to the original L2norm BN (L2BN) by multiplying a scaling factor. Experiments on various convolutional neural networks (CNNs) and generative adversarial networks (GANs) reveal that L1BN maintains almost the same accuracies and convergence rates compared to L2BN but with higher computational efficiency. On FPGA platform, the proposed signum and absolute operations in L1BN can achieve 1.5$\times$ speedup and save 50\% power consumption, compared with the original costly square and root operations, respectively. This hardwarefriendly normalization method not only surpasses L2BN in speed, but also simplify the hardware design of ASIC accelerators with higher energy efficiency. Last but not the least, L1BN promises a fully quantized training of DNNs, which is crucial to future adaptive terminal devices. 
L1norm HigherOrder Orthogonal Iteration (L1HOOI) 
Tucker decomposition is a common method for the analysis of multiway/tensor data. Standard Tucker has been shown to be sensitive against heavy corruptions, due to its L2normbased formulation which places squared emphasis to peripheral entries. In this work, we explore L1Tucker, an L1norm based reformulation of standard Tucker decomposition. After formulating the problem, we present two algorithms for its solution, namely L1norm HigherOrder Singular Value Decomposition (L1HOSVD) and L1norm HigherOrder Orthogonal Iterations (L1HOOI). The presented algorithms are accompanied by complexity and convergence analysis. Our numerical studies on tensor reconstruction and classification corroborate that L1Tucker, implemented by means of the proposed methods, attains similar performance to standard Tucker when the processed data are corruptionfree, while it exhibits sturdy resistance against heavily corrupted entries. 
L1norm HigherOrder Singular Value Decomposition (L1HOSVD) 
Tucker decomposition is a common method for the analysis of multiway/tensor data. Standard Tucker has been shown to be sensitive against heavy corruptions, due to its L2normbased formulation which places squared emphasis to peripheral entries. In this work, we explore L1Tucker, an L1norm based reformulation of standard Tucker decomposition. After formulating the problem, we present two algorithms for its solution, namely L1norm HigherOrder Singular Value Decomposition (L1HOSVD) and L1norm HigherOrder Orthogonal Iterations (L1HOOI). The presented algorithms are accompanied by complexity and convergence analysis. Our numerical studies on tensor reconstruction and classification corroborate that L1Tucker, implemented by means of the proposed methods, attains similar performance to standard Tucker when the processed data are corruptionfree, while it exhibits sturdy resistance against heavily corrupted entries. 
L1Norm Kernel PCA  We present the first model and algorithm for L1norm kernel PCA. While L2norm kernel PCA has been widely studied, there has been no work on L1norm kernel PCA. For this nonconvex and nonsmooth problem, we offer geometric understandings through reformulations and present an efficient algorithm where the kernel trick is applicable. To attest the efficiency of the algorithm, we provide a convergence analysis including linear rate of convergence. Moreover, we prove that the output of our algorithm is a local optimal solution to the L1norm kernel PCA problem. We also numerically show its robustness when extracting principal components in the presence of influential outliers, as well as its runtime comparability to L2norm kernel PCA. Lastly, we introduce its application to outlier detection and show that the L1norm kernel PCA based model outperforms especially for high dimensional data. 
L1Penalized Censored Gaussian Graphical Model  Graphical lasso is one of the most used estimators for inferring genetic networks. Despite its diffusion, there are several fields in applied research where the limits of detection of modern measurement technologies make the use of this estimator theoretically unfounded, even when the assumption of a multivariate Gaussian distribution is satisfied. Typical examples are data generated by polymerase chain reactions and flow cytometer. The combination of censoring and highdimensionality make inference of the underlying genetic networks from these data very challenging. In this paper we propose an $\ell_1$penalized Gaussian graphical model for censored data and derive two EMlike algorithms for inference. By an extensive simulation study, we evaluate the computational efficiency of the proposed algorithms and show that our proposal overcomes existing competitors when censored data are available. We apply the proposed method to gene expression data coming from microfluidic RTqPCR technology in order to make inference on the regulatory mechanisms of blood development. 
L1Regularized Maximum Likelihood Estimator  We consider the problem of estimating the parameters of a multivariate Bernoulli process with autoregressive feedback in the highdimensional setting where the number of samples available is much less than the number of parameters. This problem arises in learning interconnections of networks of dynamical systems with spiking or binaryvalued data. We allow the process to depend on its past up to a lag $p$, for a general $p \ge 1$, allowing for more realistic modeling in many applications. We propose and analyze an $\ell_1$regularized maximum likelihood estimator (MLE) under the assumption that the parameter tensor is approximately sparse. Rigorous analysis of such estimators is made challenging by the dependent and nonGaussian nature of the process as well as the presence of the nonlinearities and multilevel feedback. We derive precise upper bounds on the meansquared estimation error in terms of the number of samples, dimensions of the process, the lag $p$ and other key statistical properties of the model. The ideas presented can be used in the highdimensional analysis of regularized $M$estimators for other sparse nonlinear and nonGaussian processes with longrange dependence. 
L2AED  Fewshot learning focuses on learning a new visual concept with very limited labelled examples. A successful approach to tackle this problem is to compare the similarity between examples in a learned metric space based on convolutional neural networks. However, existing methods typically suffer from metalevel overfitting due to the limited amount of training tasks and do not normally consider the importance of the convolutional features of different examples within the same channel. To address these limitations, we make the following two contributions: (a) We propose a novel metalearning approach for aggregating useful convolutional features and suppressing noisy ones based on a channelwise attention mechanism to improve class representations. The proposed model does not require finetuning and can be trained in an endtoend manner. The main novelty lies in incorporating a shared weight generation module that learns to assign different weights to the feature maps of different examples within the same channel. (b) We also introduce a simple metalevel dropout technique that reduces metalevel overfitting in several fewshot learning approaches. In our experiments, we find that this simple technique significantly improves the performance of the proposed method as well as various stateoftheart metalearning algorithms. Applying our method to fewshot image recognition using Omniglot and miniImageNet datasets shows that it is capable of delivering a stateoftheart classification performance. 
L2Nonexpansive Neural Network  This paper proposes a class of wellconditioned neural networks in which a unit amount of change in the inputs causes at most a unit amount of change in the outputs or any of the internal layers. We develop the known methodology of controlling Lipschitz constants to realize its full potential in maximizing robustness: our linear and convolution layers subsume those in the previous Parseval networks as a special case and allow greater degrees of freedom; aggregation, pooling, splitting and other operators are adapted in new ways, and a new loss function is proposed, all for the purpose of improving robustness. With MNIST and CIFAR10 classifiers, we demonstrate a number of advantages. Without needing any adversarial training, the proposed classifiers exceed the state of the art in robustness against whitebox L2bounded adversarial attacks. Their outputs are quantitatively more meaningful than ordinary networks and indicate levels of confidence. They are also free of exploding gradients, among other desirable properties. 
Label Augmentation  A major impediment to the application of deep learning to realworld problems is the scarcity of labeled data. Small training sets are in fact of no use to deep networks as, due to the large number of trainable parameters, they will very likely be subject to overfitting phenomena. On the other hand, the increment of the training set size through further manual or semiautomatic labellings can be costly, if not possible at times. Thus, the standard techniques to address this issue are transfer learning and data augmentation, which consists of applying some sort of ‘transformation’ to existing labeled instances to let the training set grow in size. Although this approach works well in applications such as image classification, where it is relatively simple to design suitable transformation operators, it is not obvious how to apply it in more structured scenarios. Motivated by the observation that in virtually all application domains it is easy to obtain unlabeled data, in this paper we take a different perspective and propose a \emph{label augmentation} approach. We start from a small, curated labeled dataset and let the labels propagate through a larger set of unlabeled data using graph transduction techniques. This allows us to naturally use (secondorder) similarity information which resides in the data, a source of information which is typically neglected by standard augmentation techniques. In particular, we show that by using known game theoretic transductive processes we can create larger and accurate enough labeled datasets which use results in better trained neural networks. Preliminary experiments are reported which demonstrate a consistent improvement over standard image classification datasets. 
Label Consistent ksvd (LCKSVD) 
Label Embedded Dictionary Learning for Image Classification 
Label Embedded Dictionary Learning (LEDL) 
Dictionary learning methods can be split into two categories: i) class specific dictionary learning ii) class shared dictionary learning. The difference between the two categories is how to use the discriminative information. With the first category, samples of different classes are mapped to different subspaces which leads to some redundancy in the base vectors. For the second category, the samples in each specific class can not be described well. Moreover, most class shared dictionary learning methods use the L0norm regularization term as the sparse constraint. In this paper, we first propose a novel class shared dictionary learning method named label embedded dictionary learning (LEDL) by introducing the L1norm sparse constraint to replace the conventional L0norm regularization term in LCKSVD method. Then we propose a novel network named hybrid dictionary learning network (HDLN) to combine the class specific dictionary learning with class shared dictionary learning together to fully describe the feature to boost the performance of classification. Extensive experimental results on six benchmark datasets illustrate that our methods are capable of achieving superior performance compared to several conventional classification algorithms. 
Label Embedding Network  We propose a method, called Label Embedding Network, which can learn label representation (label embedding) during the training process of deep networks. With the proposed method, the label embedding is adaptively and automatically learned through back propagation. The original onehot represented loss function is converted into a new loss function with soft distributions, such that the originally unrelated labels have continuous interactions with each other during the training process. As a result, the trained model can achieve substantially higher accuracy and with faster convergence speed. Experimental results based on competitive tasks demonstrate the effectiveness of the proposed method, and the learned label embedding is reasonable and interpretable. The proposed method achieves comparable or even better results than the stateoftheart systems. The source code is available at \url{https://…/LabelEmb}. 
Label Leakage  
Label Network Embedding for MultiLabel Classification (LNEMLC) 
Multilabel classification aims to classify instances with discrete nonexclusive labels. Most approaches on multilabel classification focus on effective adaptation or transformation of existing binary and multiclass learning approaches but fail in modelling the joint probability of labels or do not preserve generalization abilities for unseen label combinations. To address these issues we propose a new multilabel classification scheme, LNEMLC – Label Network Embedding for MultiLabel Classification, that embeds the label network and uses it to extend input space in learning and inference of any base multilabel classifier. The approach allows capturing of labels’ joint probability at low computational complexity providing results comparable to the best methods reported in the literature. We demonstrate how the method reveals statistically significant improvements over the simple kNN baseline classifier. We also provide hints for selecting the robust configuration that works satisfactorily across data domains. 
Labeled Latent Dirichlet Allocation (LLDA) 
Labeled Latent Dirichlet Allocation (LLDA) is an extension of the standard unsupervised Latent Dirichlet Allocation (LDA) algorithm, to address multilabel learning tasks. Previous work has shown it to perform in par with other stateoftheart multilabel methods. Nonetheless, with increasing label sets sizes LLDA encounters scalability issues. In this work, we introduce Subset LLDA, a simple variant of the standard LLDA algorithm, that not only can effectively scale up to problems with hundreds of thousands of labels but also improves over the LLDA stateoftheart. We conduct extensive experiments on eight data sets, with label sets sizes ranging from hundreds to hundreds of thousands, comparing our proposed algorithm with the previously proposed LLDA algorithms (Prior–LDA, Dep–LDA), as well as the state of the art in extreme multilabel classification. The results show a steady advantage of our method over the other LLDA algorithms and competitive results compared to the extreme multilabel classification algorithms. 
LabelNoise Robust GAN (rGAN) 
Generative adversarial networks (GANs) are a framework that learns a generative distribution through adversarial training. Recently, their class conditional extensions (e.g., conditional GAN (cGAN) and auxiliary classifier GAN (ACGAN)) have attracted much attention owing to their ability to learn the disentangled representations and to improve the training stability. However, their training requires the availability of largescale accurate classlabeled data, which are often laborious or impractical to collect in a realworld scenario. To remedy the drawback, we propose a novel family of GANs called labelnoise robust GANs (rGANs), which, by incorporating a noise transition model, can learn a clean label conditional generative distribution even when training labels are noisy. In particular, we propose two variants: rACGAN, which is a bridging model between ACGAN and the noiserobust classification model, and rcGAN, which is an extension of cGAN and is guaranteed to learn the clean label conditional distribution in an optimal condition. In addition to providing the theoretical background, we demonstrate the effectiveness of our models through extensive experiments using diverse GAN configurations, various noise settings, and multiple evaluation metrics (in which we tested 402 patterns in total). 
Laconic  We motivate a method for transparently identifying ineffectual computations in unmodified Deep Learning models and without affecting accuracy. Specifically, we show that if we decompose multiplications down to the bit level the amount of work performed during inference for image classification models can be consistently reduced by two orders of magnitude. In the best case studied of a sparse variant of AlexNet, this approach can ideally reduce computation work by more than 500x. We present Laconic a hardware accelerator that implements this approach to improve execution time, and energy efficiency for inference with Deep Learning Networks. Laconic judiciously gives up some of the work reduction potential to yield a lowcost, simple, and energy efficient design that outperforms other stateoftheart accelerators. For example, a Laconic configuration that uses a weight memory interface with just 128 wires outperforms a conventional accelerator with a 2Kwire weight memory interface by 2.3x on average while being 2.13x more energy efficient on average. A Laconic configuration that uses a 1Kwire weight memory interface, outperforms the 2Kwire conventional accelerator by 15.4x and is 1.95x more energy efficient. Laconic does not require but rewards advances in model design such as a reduction in precision, the use of alternate numeric representations that reduce the number of bits that are ‘1’, or an increase in weight or activation sparsity. 
LAD Regression  ➘ “Least Absolute Deviations” 
Ladder  The organizer of a machine learning competition faces the problem of maintaining an accurate leaderboard that faithfully represents the quality of the best submission of each competing team. What makes this estimation problem particularly challenging is its sequential and adaptive nature. As participants are allowed to repeatedly evaluate their submissions on the leaderboard, they may begin to overfit to the holdout data that supports the leaderboard. Few theoretical results give actionable advice on how to design a reliable leaderboard. Existing approaches therefore often resort to poorly understood heuristics such as limiting the bit precision of answers and the rate of resubmission. In this work, we introduce a notion of leaderboard accuracy tailored to the format of a competition. We introduce a natural algorithm called the Ladder and demonstrate that it simultaneously supports strong theoretical guarantees in a fully adaptive model of estimation, withstands practical adversarial attacks, and achieves high utility on real submission files from an actual competition hosted by Kaggle. 
LadderNet  UNet has been providing stateoftheart performance in many medical image segmentation problems. Many modifications have been proposed for UNet, such as attention UNet, recurrent residual convolutional UNet (R2UNet), and UNet with residual blocks or blocks with dense connections. However, all these modifications have an encoderdecoder structure with skip connections, and the number of paths for information flow is limited. We propose LadderNet in this paper, which can be viewed as a chain of multiple UNets. Instead of only one pair of encoder branch and decoder branch in UNet, a LadderNet has multiple pairs of encoderdecoder branches, and has skip connections between every pair of adjacent decoder and decoder branches in each level. Inspired by the success of ResNet and R2UNet, we use modified residual blocks where two convolutional layers in one block share the same weights. A LadderNet has more paths for information flow because of skip connections and residual blocks, and can be viewed as an ensemble of Fully Convolutional Networks (FCN). The equivalence to an ensemble of FCNs improves segmentation accuracy, while the shared weights within each residual block reduce parameter number. Semantic segmentation is essential for retinal disease detection. We tested LadderNet on two benchmark datasets for blood vessel segmentation in retinal images, and achieved superior performance over methods in the literature. 
Lagrange Multiplier  In mathematical optimization, the method of Lagrange multipliers is a strategy for finding the local maxima and minima of a function subject to equality constraints. 
Lagrange Policy Gradient  Most algorithms for reinforcement learning work by estimating actionvalue functions. Here we present a method that uses Lagrange multipliers, the costate equation, and multilayer neural networks to compute policy gradients. We show that this method can find solutions to timeoptimal control problems, driving nonlinear mechanical systems quickly to a target configuration. On these tasks its performance is comparable to that of deep deterministic policy gradient, a recent actionvalue method. 
LAHA  Extreme multilabel text classification (XMTC) aims at tagging a document with most relevant labels from an extremely largescale label set. It is a challenging problem especially for the tail labels because there are only few training documents to build classifier. This paper is motivated to better explore the semantic relationship between each document and extreme labels by taking advantage of both document content and label correlation. Our objective is to establish an explicit labelaware representation for each document with a hybrid attention deep neural network model (LAHA). LAHA consists of three parts. The first part adopts a multilabel selfattention mechanism to detect the contribution of each word to labels. The second part exploits the label structure and document content to determine the semantic connection between words and labels in a same latent space. An adaptive fusion strategy is designed in the third part to obtain the final labelaware document representation so that the essence of previous two parts can be sufficiently integrated. Extensive experiments have been conducted on six benchmark datasets by comparing with the stateoftheart methods. The results show the superiority of our proposed LAHA method, especially on the tail labels. 
Lambda Architecture  Lambda Architecture proposes a simpler, elegant paradigm that is designed to tame complexity while being able to store and effectively process large amounts of data. The Lambda Architecture was originally presented by Nathan Marz, who is well known in the big data community for his work on the Storm project. Lambda Architecture Lambda Architecture 
LambdaMART  At a high level, LambdaMART is an algorithm that uses gradient boosting to directly optimize Learning to Rank specific cost functions such as NDCG. 
LAmbdaPACK  Linear algebra operations are widely used in scientific computing and machine learning applications. However, it is challenging for scientists and data analysts to run linear algebra at scales beyond a single machine. Traditional approaches either require access to supercomputing clusters, or impose configuration and cluster management challenges. In this paper we show how the disaggregation of storage and compute resources in socalled ‘serverless’ environments, combined with computeintensive workload characteristics, can be exploited to achieve elastic scalability and ease of management. We present numpywren, a system for linear algebra built on a serverless architecture. We also introduce LAmbdaPACK, a domainspecific language designed to implement highly parallel linear algebra algorithms in a serverless setting. We show that, for certain linear algebra algorithms such as matrix multiply, singular value decomposition, and Cholesky decomposition, numpywren’s performance (completion time) is within 33% of ScaLAPACK, and its compute efficiency (total CPUhours) is up to 240% better due to elasticity, while providing an easier to use interface and better fault tolerance. At the same time, we show that the inability of serverless runtimes to exploit locality across the cores in a machine fundamentally limits their network efficiency, which limits performance on other algorithms such as QR factorization. This highlights how cloud providers could better support these types of computations through small changes in their infrastructure. 
Lambert W Function  In mathematics, the Lambert W function, also called the omega function or product logarithm, is a set of functions, namely the branches of the inverse relation of the function z = f(W) = We^W where e^W is the exponential function and W is any complex number. In other words, the defining equation for W(z) is: z = W(z)e^{W(z)} for any complex number z. http://…/LambertWFunction.html lamW 
LAMVI2  Tuning machine learning models, particularly deep learning architectures, is a complex process. Automated hyperparameter tuning algorithms often depend on specific optimization metrics. However, in many situations, a developer trades one metric against another: accuracy versus overfitting, precision versus recall, smaller models and accuracy, etc. With deep learning, not only are the model’s representations opaque, the model’s behavior when parameters ‘knobs’ are changed may also be unpredictable. Thus, picking the ‘best’ model often requires timeconsuming model comparison. In this work, we introduce LAMVI2, a visual analytics system to support a developer in comparing hyperparameter settings and outcomes. By focusing on wordembedding models (‘deep learning for text’) we integrate views to compare both highlevel statistics as well as internal model behaviors (e.g., comparing word ‘distances’). We demonstrate how developers can work with LAMVI2 to more quickly and accurately narrow down an appropriate and effective applicationspecific model. 
Lanczos Latent Factor Recommender (LLFR) 
The purpose if this master’s thesis is to study and develop a new algorithmic framework for Collaborative Filtering to produce recommendations in the topN recommendation problem. Thus, we propose Lanczos Latent Factor Recommender (LLFR); a novel ‘big data friendly’ collaborative filtering algorithm for topN recommendation. Using a computationally efficient Lanczosbased procedure, LLFR builds a low dimensional item similarity model, that can be readily exploited to produce personalized ranking vectors over the item space. A number of experiments on real datasets indicate that LLFR outperforms other stateoftheart topN recommendation methods from a computational as well as a qualitative perspective. Our experimental results also show that its relative performance gains, compared to competing methods, increase as the data get sparser, as in the Cold Start Problem. More specifically, this is true both when the sparsity is generalized – as in the New Community Problem, a very common problem faced by real recommender systems in their beginning stages, when there is not sufficient number of ratings for the collaborative filtering algorithms to uncover similarities between items or users – and in the very interesting case where the sparsity is localized in a small fraction of the dataset – as in the New Users Problem, where new users are introduced to the system, they have not rated many items and thus, the CF algorithm can not make reliable personalized recommendations yet. 
Lanczos Method  The Lanczos algorithm is a direct algorithm devised by Cornelius Lanczos that is an adaptation of power methods to find the most useful eigenvalues and eigenvectors of an n t h {\displaystyle n^{th}} n^{{th}} order linear system with a limited number of operations, m {\displaystyle m} m, where m {\displaystyle m} m is much smaller than n {\displaystyle n} n. Although computationally efficient in principle, the method as initially formulated was not useful, due to its numerical instability. In 1970, Ojalvo and Newman showed how to make the method numerically stable and applied it to the solution of very large engineering structures subjected to dynamic loading. This was achieved using a method for purifying the vectors to any degree of accuracy, which when not performed, produced a series of vectors that were highly contaminated by those associated with the lowest natural frequencies. In their original work, these authors also suggested how to select a starting vector (i.e. use a random number generator to select each element of the starting vector) and suggested an empirically determined method for determining m {\displaystyle m} m, the reduced number of vectors (i.e. it should be selected to be approximately 1 Ā½ times the number of accurate eigenvalues desired). Soon thereafter their work was followed by Paige who also provided an error analysis. In 1988, Ojalvo produced a more detailed history of this algorithm and an efficient eigenvalue error test. Currently, the method is widely used in a variety of technical fields and has seen a number of variations. 
Lanczos Network (LanczosNet) 
We propose the Lanczos network (LanczosNet), which uses the Lanczos algorithm to construct low rank approximations of the graph Laplacian for graph convolution. Relying on the tridiagonal decomposition of the Lanczos algorithm, we not only efficiently exploit multiscale information via fast approximated computation of matrix power but also design learnable spectral filters. Being fully differentiable, LanczosNet facilitates both graph kernel learning as well as learning node embeddings. We show the connection between our LanczosNet and graph based manifold learning methods, especially the diffusion maps. We benchmark our model against several recent deep graph networks on citation networks and QM8 quantum chemistry dataset. Experimental results show that our model achieves the stateoftheart performance in most tasks. 
LanczOs Variance Estimates (LOVE) 
One of the most compelling features of Gaussian process (GP) regression is its ability to provide well calibrated posterior distributions. Recent advances in inducing point methods have drastically sped up marginal likelihood and posterior mean computations, leaving posterior covariance estimation and sampling as the remaining computational bottlenecks. In this paper we address this shortcoming by using the Lanczos decomposition algorithm to rapidly approximate the predictive covariance matrix. Our approach, which we refer to as LOVE (LanczOs Variance Estimates), substantially reduces the time and space complexity over any previous method. In practice, it can compute predictive covariances up to 2,000 times faster and draw samples 18,000 time faster than existing methods, all without sacrificing accuracy. 
Landmark Retracing Network (LRN) 
Since convolutional neural network (CNN) lacks an inherent mechanism to handle large scale variations, we always need to compute feature maps multiple times for multiscale object detection, which has the bottleneck of computational cost in practice. To address this, we devise a recurrent scale approximation (RSA) to compute feature map once only, and only through this map can we approximate the rest maps on other levels. At the core of RSA is the recursive rolling out mechanism: given an initial map on a particular scale, it generates the prediction on a smaller scale that is half the size of input. To further increase efficiency and accuracy, we (a): design a scaleforecast network to globally predict potential scales in the image since there is no need to compute maps on all levels of the pyramid. (b): propose a landmark retracing network (LRN) to retrace back locations of the regressed landmarks and generate a confidence score for each landmark; LRN can effectively alleviate false positives due to the accumulated error in RSA. The whole system could be trained endtoend in a unified CNN framework. Experiments demonstrate that our proposed algorithm is superior against stateofthearts on face detection benchmarks and achieves comparable results for generic proposal generation. The source code of RSA is available at github.com/sciencefans/RSAforobjectdetection. 
Langevin Monte Carlo  
Language Model  A statistical language model assigns a probability to a sequence of m words by means of a probability distribution. Language modeling is used in many natural language processing applications such as speech recognition, machine translation, partofspeech tagging, parsing and information retrieval. 
Language Model Based Grammatical Error Correction (LMGEC) 
Grammatical error correction (GEC) is one of the areas in natural language processing in which purely neural models have not yet superseded more traditional symbolic models. Hybrid systems combining phrasebased statistical machine translation (SMT) and neural sequence models are currently among the most effective approaches to GEC. However, both SMT and neural sequencetosequence models require large amounts of annotated data. Language model based Grammatical error correction (LMGEC) is a promising alternative which does not rely on annotated training data. We show how to improve LMGEC by applying modelling techniques based on finite state transducers. We report further gains by rescoring with neural language models. We show that our methods developed for LMGEC can also be used with SMT systems if annotated training data is available. Our best system outperforms the best published result on the CoNLL2014 test set, and achieves far better relative improvements over the SMT baselines than previous hybrid systems. 
LanguagEAction Reward Network (LEARN) 
Recent reinforcement learning (RL) approaches have shown strong performance in complex domains such as Atari games, but are often highly sample inefficient. A common approach to reduce interaction time with the environment is to use reward shaping, which involves carefully designing reward functions that provide the agent intermediate rewards for progress towards the goal. However, designing appropriate shaping rewards is known to be difficult as well as timeconsuming. In this work, we address this problem by using natural language instructions to perform reward shaping. We propose the LanguagEAction Reward Network (LEARN), a framework that maps freeform natural language instructions to intermediate rewards based on actions taken by the agent. These intermediate languagebased rewards can seamlessly be integrated into any standard reinforcement learning algorithm. We experiment with Montezuma’s Revenge from the Atari Learning Environment, a popular benchmark in RL. Our experiments on a diverse set of 15 tasks demonstrate that, for the same number of interactions with the environment, languagebased rewards lead to successful completion of the task 60% more often on average, compared to learning without language. 
LanguageAgnostic Keyword Extraction (LAKE) 
Keyword Extraction is an important task in several text analysis endeavors. In this paper, we present a critical discussion of the issues and challenges ingraphbased keyword extraction methods, along with comprehensive empirical analysis. We propose a parameterless method for constructing graph of text that captures the contextual relation between words. A novel word scoring method is also proposed based on the connection between concepts. We demonstrate that both proposals are individually superior to those followed by the stateoftheart graphbased keyword extraction algorithms. Combination of the proposed graph construction and scoring methods leads to a novel, parameterless keyword extraction method (sCAKE) based on semantic connectivity of words in the document. Motivated by limited availability of NLP tools for several languages, we also design and present a languageagnostic keyword extraction (LAKE) method. We eliminate the need of NLP tools by using a statistical filter to identify candidate keywords before constructing the graph. We show that the resulting method is a competent solution for extracting keywords from documents oflanguages lacking sophisticated NLP support. 
LanguageConditioned Graph Network (LCGN) 
Solving grounded language tasks often requires reasoning about relationships between objects in the context of a given task. For example, to answer the question “What color is the mug on the plate?” we must check the color of the specific mug that satisfies the “on” relationship with respect to the plate. Recent work has proposed various methods capable of complex relational reasoning. However, most of their power is in the inference structure, while the scene is represented with simple local appearance features. In this paper, we take an alternate approach and build contextualized representations for objects in a visual scene to support relational reasoning. We propose a general framework of LanguageConditioned Graph Networks (LCGN), where each node represents an object, and is described by a contextaware representation from related objects through iterative message passing conditioned on the textual input. E.g., conditioning on the “on” relationship to the plate, the object “mug” gathers messages from the object “plate” to update its representation to “mug on the plate”, which can be easily consumed by a simple classifier for answer prediction. We experimentally show that our LCGN approach effectively supports relational reasoning and improves performance across several tasks and datasets. 
LanguageConditioned Reward Learning  Reinforcement learning is a promising framework for solving control problems, but its use in practical situations is hampered by the fact that reward functions are often difficult to engineer. Specifying goals and tasks for autonomous machines, such as robots, is a significant challenge: conventionally, reward functions and goal states have been used to communicate objectives. But people can communicate objectives to each other simply by describing or demonstrating them. How can we build learning algorithms that will allow us to tell machines what we want them to do? In this work, we investigate the problem of grounding language commands as reward functions using inverse reinforcement learning, and argue that languageconditioned rewards are more transferable than languageconditioned policies to new environments. We propose languageconditioned reward learning (LCRL), which grounds language commands as a reward function represented by a deep neural network. We demonstrate that our model learns rewards that transfer to novel tasks and environments on realistic, highdimensional visual environments with natural language commands, whereas directly learning a languageconditioned policy leads to poor performance. 
Laplacian Power Network  Deep Neural Networks often suffer from lack of robustness to adversarial noise. To mitigate this drawback, authors have proposed different approaches, such as adding regularizers or training using adversarial examples. In this paper we propose a new regularizer built upon the Laplacian of similarity graphs obtained from the representation of training data at each intermediate representation. This regularizer penalizes large changes (across consecutive layers in the architecture) in the distance between examples of different classes. We provide theoretical justification for this regularizer and demonstrate its effectiveness when facing adversarial noise on classical supervised learning vision datasets. 
Laplacian Pyramid Based Kernel Prediction Network (LPKPN) 
Most of the existing learningbased single image superresolution (SISR) methods are trained and evaluated on simulated datasets, where the lowresolution (LR) images are generated by applying a simple and uniform degradation (i.e., bicubic downsampling) to their highresolution (HR) counterparts. However, the degradations in realworld LR images are far more complicated. As a consequence, the SISR models trained on simulated data become less effective when applied to practical scenarios. In this paper, we build a realworld superresolution (RealSR) dataset where paired LRHR images on the same scene are captured by adjusting the focal length of a digital camera. An image registration algorithm is developed to progressively align the image pairs at different resolutions. Considering that the degradation kernels are naturally nonuniform in our dataset, we present a Laplacian pyramid based kernel prediction network (LPKPN), which efficiently learns perpixel kernels to recover the HR image. Our extensive experiments demonstrate that SISR models trained on our RealSR dataset deliver better visual quality with sharper edges and finer textures on realworld scenes than those trained on simulated datasets. Though our RealSR dataset is built by using only two cameras (Canon 5D3 and Nikon D810), the trained model generalizes well to other camera devices such as Sony a7II and mobile phones. 
Large Deviation Principles  We establish the Large Deviation Principles for a topological Markov shift on a countably infinite number of alphabets which satisfies a strong combinatorial assumption called ‘finite primitiveness’ by Mauldin $\&$ Urba\’nski. More precisely, we assume the existence of a Gibbs measure for a potential $\phi$ in the sense of Bowen, and prove the level2 Large Deviation Principles for the distribution of Birkhoff averages under the Gibbs measure, as well as that of weighted periodic points and iterated preimages. The rate function is in common, written with the pressure and the free energy associated with the potential $\phi$. The Gibbs measure is not assumed to be an equilibrium state for the potential $\phi$, nor is assumed the existence of an equilibrium state. We provide a sufficient condition for minimizers of the rate function to be equilibrium states. We apply our results to the Gauss transformation and obtain a global limit theorem on the frequency of digits in the regular continued fraction expansion. 
Large Margin Deep Network  We present a formulation of deep learning that aims at producing a large margin classifier. The notion of margin, minimum distance to a decision boundary, has served as the foundation of several theoretically profound and empirically successful results for both classification and regression tasks. However, most large margin algorithms are applicable only to shallow models with a preset feature representation; and conventional margin methods for neural networks only enforce margin at the output layer. Such methods are therefore not well suited for deep networks. In this work, we propose a novel loss function to impose a margin on any chosen set of layers of a deep network (including input and hidden layers). Our formulation allows choosing any norm on the metric measuring the margin. We demonstrate that the decision boundary obtained by our loss has nice properties compared to standard classification loss functions. Specifically, we show improved empirical results on the MNIST, CIFAR10 and ImageNet datasets on multiple tasks: generalization from small training sets, corrupted labels, and robustness against adversarial perturbations. The resulting loss is general and complementary to existing data augmentation (such as random/adversarial input transform) and regularization techniques (such as weight decay, dropout, and batch norm). 
Large Vocabulary Continuous Speech Recognition System (LVCSR) 
The search problem in LVCSR can be simply stated: find the most probable sequence of words given a sequence of acoustic observations, an acoustic model and a language model. This is a demanding problem since word boundary information is not available in continuous speech and each word in the dictionary may be hypothesized to start at each frame of acoustic data. The problem is further complicated by the vocabulary size (typically 65,000 words) and the structure imposed on the search space by the language model. Direct evaluation of all the possible word sequences is impossible (given the large vocabulary) and an efficient search algorithm will consider only a very small subset of all possible utterance models. Typically, the effective size of the search space is reduced through pruning of unlikely hypotheses and/or the elimination of repeated computations. 
LargeScale Information Network Embedding (LINE) 
This paper studies the problem of embedding very large information networks into lowdimensional vector spaces, which is useful in many tasks such as visualization, node classification, and link prediction. Most existing graph embedding methods do not scale for real world information networks which usually contain millions of nodes. In this paper, we propose a novel network embedding method called the ‘LINE,’ which is suitable for arbitrary types of information networks: undirected, directed, and/or weighted. The method optimizes a carefully designed objective function that preserves both the local and global network structures. An edgesampling algorithm is proposed that addresses the limitation of the classical stochastic gradient descent and improves both the effectiveness and the efficiency of the inference. Empirical experiments prove the effectiveness of the LINE on a variety of realworld information networks, including language networks, social networks, and citation networks. The algorithm is very efficient, which is able to learn the embedding of a network with millions of vertices and billions of edges in a few hours on a typical single machine. The source code of the LINE is available online. 
Largest Gaps  In this paper, the algorithm $Largest$ $Gaps$ is introduced, for simultaneously clustering both rows and columns of a matrix to form homogeneous blocks. The definition of clustering is modelbased: clusters and data are generated under the Latent Block Model. In comparison with algorithms designed for this model, the major advantage of the $Largest$ $Gaps$ algorithm is to cluster using only some marginals of the matrix, the size of which is much smaller than the whole matrix. The procedure is linear with respect to the number of entries and thus much faster than the classical algorithms. It simultaneously selects the number of classes as well, and the estimation of the parameters is then made very easily once the classification is obtained. Moreover, the paper proves the procedure to be consistent under the LBM, and it illustrates the statistical performance with some numerical experiments. 
Lariat  We propose a new method for supervised learning, especially suited to wide data where the number of features is much greater than the number of observations. The method combines the lasso ($\ell_1$) sparsity penalty with a quadratic penalty that shrinks the coefficient vector toward the leading principal components of the feature matrix. We call the proposed method the ‘Lariat’. The method can be especially powerful if the features are preassigned to groups (such as cellpathways, assays or protein interaction networks). In that case, the Lariat shrinks each groupwise component of the solution toward the leading principal components of that group. In the process, it also carries out selection of the feature groups. We provide some theory for this method and illustrate it on a number of simulated and real data examples. 
Lasagne  In this work we propose Lasagne, a methodology to learn locality and structure aware graph node embeddings in an unsupervised way. In particular, we show that the performance of existing randomwalk based approaches depends strongly on the structural properties of the graph, e.g., the size of the graph, whether the graph has a flat or upwardsloping Network Community Profile (NCP), whether the graph is expanderlike, whether the classes of interest are more kcorelike or more peripheral, etc. For larger graphs with flat NCPs that are strongly expanderlike, existing methods lead to random walks that expand rapidly, touching many dissimilar nodes, thereby leading to lowerquality vector representations that are less useful for downstream tasks. Rather than relying on global random walks or neighbors within fixed hop distances, Lasagne exploits strongly local Approximate Personalized PageRank stationary distributions to more precisely engineer local information into node embeddings. This leads, in particular, to more meaningful and more useful vector representations of nodes in poorlystructured graphs. We show that Lasagne leads to significant improvement in downstream multilabel classification for larger graphs with flat NCPs, that it is comparable for smaller graphs with upwardsloping NCPs, and that is comparable to existing methods for link prediction tasks. 
LaSOT  In this paper, we present LaSOT, a highquality benchmark for Largescale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT the largest, to the best of our knowledge, densely annotated tracking benchmark. The average sequence length of LaSOT is more than 2,500 frames, and each sequence comprises various challenges deriving from the wild where target objects may disappear and reappear again in the view. By releasing LaSOT, we expect to provide the community a largescale dedicated benchmark with highquality for both the training of deep trackers and the veritable evaluation of tracking algorithms. Moreover, considering the close connections of visual appearance and natural language, we enrich LaSOT by providing additional language specification, aiming at encouraging the exploration of natural linguistic feature for tracking. A thorough experimental evaluation of 35 tracking algorithms on LaSOT is presented with detailed analysis, and the results demonstrate that there is still a big room to improvements. The benchmark and evaluation results are made publicly available at https://…/. 
Lasso Penalized Sparse Asymmetric Least Squares (SALES) 
SALES 
Lasso Regression  ➘ “Least Absolute Shrinkage and Selection Operator” 
Lassoing Eigenvalues (elasso) 
The properties of penalized sample covariance matrices depend on the choice of the penalty function. In this paper, we introduce a class of nonsmooth penalty functions for the sample covariance matrix, and demonstrate how this method results in a grouping of the estimated eigenvalues. We refer to this method as ‘lassoing eigenvalues’ or as the ‘elasso’. 
Last Observation Projection  
LaSVM (LaSVM) 
LASVM is an approximate SVM solver that uses online approximation. It reaches accuracies similar to that of a real SVM after performing a single sequential pass through the training examples. Further benefits can be achieved using selective sampling techniques to choose which example should be considered next. As show in the graph, LASVM requires considerably less memory than a regular SVM solver. This becomes a considerable speed advantage for large training sets. In fact LASVM has been used to train a 10 class SVM classifier with 8 million examples on a single processor. lasvmR 
Latent Adversarial Training (LAT) 
Neural networks are vulnerable to adversarial attacks — small visually imperceptible crafted noise which when added to the input drastically changes the output. The most effective method of defending against these adversarial attacks is to use the methodology of adversarial training. We analyze the adversarially trained robust models to study their vulnerability against adversarial attacks at the level of the latent layers. Our analysis reveals that contrary to the input layer which is robust to adversarial attack, the latent layer of these robust models are highly susceptible to adversarial perturbations of small magnitude. Leveraging this information, we introduce a new technique Latent Adversarial Training (LAT) which comprises of finetuning the adversarially trained models to ensure the robustness at the feature layers. We also propose Latent Attack (LA), a novel algorithm for construction of adversarial examples. LAT results in minor improvement in test accuracy and leads to a stateoftheart adversarial accuracy against the universal firstorder adversarial PGD attack which is shown for the MNIST, CIFAR10, CIFAR100 datasets. 
Latent Association Mining in Binary Data (LAMB) 
We consider the problem of identifying groups of mutually associated variables in moderate or high dimensional data. In many cases, ordinary Pearson correlation provides useful information concerning the linear relationship between variables. However, for binary data, ordinary correlation may lose power and may lack interpretability. In this paper, we develop and investigate a new method called Latent Association Mining in Binary Data (LAMB). The LAMB method is built on the assumption that the binary observations represent a random thresholding of a latent continuous variable that may have a complex correlation structure. We consider a new measure of association, latent correlation, that is designed to assess association in the underlying continuous variable, without bias due to the mediating effects of the thresholding procedure. The full LAMB procedure makes use of iterative hypothesis testing to identify groups of latently correlated variables. LAMB is shown to improve power over existing methods in simulated settings, to be computationally efficient for large datasets, and to uncover new meaningful results from common real data types. 
Latent Attention Network  Deep neural networks are able to solve tasks across a variety of domains and modalities of data. Despite many empirical successes, we lack the ability to clearly understand and interpret the learned internal mechanisms that contribute to such effective behaviors or, more critically, failure modes. In this work, we present a general method for visualizing an arbitrary neural network’s inner mechanisms and their power and limitations. Our datasetcentric method produces visualizations of how a trained network attends to components of its inputs. The computed ‘attention masks’ support improved interpretability by highlighting which input attributes are critical in determining output. We demonstrate the effectiveness of our framework on a variety of deep neural network architectures in domains from computer vision, natural language processing, and reinforcement learning. The primary contribution of our approach is an interpretable visualization of attention that provides unique insights into the network’s underlying decisionmaking process irrespective of the data modality. 
Latent Autoregressive Count Models  See Pedeli and Varin (2018) <arXiv:1805.10865> for details. lacm 
Latent Class Analysis (LCA) 
Latent class analysis (LCA) identifies unobservable subgroups within a population. What Is Latent Class Analysis? 
Latent Class Model (LCM) 
In statistics, a latent class model (LCM) relates a set of observed (usually discrete) multivariate variables to a set of latent variables. It is a type of latent variable model. It is called a latent class model because the latent variable is discrete. A class is characterized by a pattern of conditional probabilities that indicate the chance that variables take on certain values. Latent Class Analysis (LCA) is a subset of structural equation modeling, used to find groups or subtypes of cases in multivariate categorical data. These subtypes are called “latent classes”. Confronted with a situation as follows, a researcher might choose to use LCA to understand the data: Imagine that symptoms ad have been measured in a range of patients with diseases X Y and Z, and that disease X is associated with the presence of symptoms a, b, and c, disease Y with symptoms b, c, d, and disease Z with symptoms a, c and d. The LCA will attempt to detect the presence of latent classes (the disease entities), creating patterns of association in the symptoms. As in factor analysis, the LCA can also be used to classify case according to their maximum likelihood class membership. Because the criterion for solving the LCA is to achieve latent classes within which there is no longer any association of one symptom with another (because the class is the disease which causes their association, and the set of diseases a patient has (or class a case is a member of) causes the symptom association, the symptoms will be “conditionally independent”, i.e., conditional on class membership, they are no longer related. 
Latent ClassConditional Noise Model (LCCN) 
Learning with noisy labels, which aims to reduce expensive labors on accurate annotations, has become imperative in the Big Data era. Previous noise transition based method has achieved promising results and presented a theoretical guarantee on performance in the case of classconditional noise. However, this type of approaches critically depend on an accurate preestimation of the noise transition, which is usually impractical. Subsequent improvement adapts the preestimation along with the training progress via a Softmax layer. However, the parameters in the Softmax layer are highly tweaked for the fragile performance due to the illposed stochastic approximation. To address these issues, we propose a Latent ClassConditional Noise model (LCCN) that naturally embeds the noise transition under a Bayesian framework. By projecting the noise transition into a Dirichletdistributed space, the learning is constrained on a simplex based on the whole dataset, instead of some adhoc parametric space. We then deduce a dynamic label regression method for LCCN to iteratively infer the latent labels, to stochastically train the classifier and to model the noise. Our approach safeguards the bounded update of the noise transition, which avoids previous arbitrarily tuning via a batch of samples. We further generalize LCCN for openset noisy labels and the semisupervised setting. We perform extensive experiments with the controllable noise data sets, CIFAR10 and CIFAR100, and the agnostic noise data sets, Clothing1M and WebVision17. The experimental results have demonstrated that the proposed model outperforms several stateoftheart methods. 
Latent Code and TextBased GAN (LATEXTGAN) 
Text generation with generative adversarial networks (GANs) can be divided into the textbased and codebased categories according to the type of signals used for discrimination. In this work, we introduce a novel textbased approach called SoftGAN to effectively exploit GAN setup for text generation. We demonstrate how autoencoders (AEs) can be used for providing a continuous representation of sentences, which we will refer to as softtext. This soft representation will be used in GAN discrimination to synthesize similar softtexts. We also propose hybrid latent code and textbased GAN (LATEXTGAN) approaches with one or more discriminators, in which a combination of the latent code and the softtext is used for GAN discriminations. We perform a number of subjective and objective experiments on two wellknown datasets (SNLI and Image COCO) to validate our techniques. We discuss the results using several evaluation metrics and show that the proposed techniques outperform the traditional GANbased textgeneration methods. 
Latent Constrained Correlation Filter (LCCF) 
Correlation filters are special classifiers designed for shiftinvariant object recognition, which are robust to pattern distortions. The recent literature shows that combining a set of subfilters trained based on a single or a small group of images obtains the best performance. The idea is equivalent to estimating variable distribution based on the data sampling (bagging), which can be interpreted as finding solutions (variable distribution approximation) directly from sampled data space. However, this methodology fails to account for the variations existed in the data. In this paper, we introduce an intermediate step — solution sampling — after the data sampling step to form a subspace, in which an optimal solution can be estimated. More specifically, we propose a new method, named latent constrained correlation filters (LCCF), by mapping the correlation filters to a given latent subspace, and develop a new learning framework in the latent subspace that embeds distributionrelated constraints into the original problem. To solve the optimization problem, we introduce a subspace based alternating direction method of multipliers (SADMM), which is proven to converge at the saddle point. Our approach is successfully applied to three different tasks, including eye localization, car detection and object tracking. Extensive experiments demonstrate that LCCF outperforms the stateoftheart methods. The source code will be publicly available. https://…/. 
Latent Constraints  Deep generative neural networks have proven effective at both conditional and unconditional modeling of complex data distributions. Conditional generation enables interactive control, but creating new controls often requires expensive retraining. In this paper, we develop a method to condition generation without retraining the model. By posthoc learning latent constraints, value functions that identify regions in latent space that generate outputs with desired attributes, we can conditionally sample from these regions with gradientbased optimization or amortized actor functions. Combining attribute constraints with a universal ‘realism’ constraint, which enforces similarity to the data distribution, we generate realistic conditional images from an unconditional variational autoencoder. Further, using gradientbased optimization, we demonstrate identitypreserving transformations that make the minimal adjustment in latent space to modify the attributes of an image. Finally, with discrete sequences of musical notes, we demonstrate zeroshot conditional generation, learning latent constraints in the absence of labeled data or a differentiable reward function. Code with dedicated cloud instance has been made publicly available (https://goo.gl/STGMGx ). 
Latent Convex Hull (LCH) 
Providing unexpected recommendations is an important task for recommender systems. To do this, we need to start from the expectations of users and deviate from these expectations when recommending items. Previously proposed approaches model user expectations in the feature space, making them limited to the items that the user has visited or expected by the deduction of associated rules, without including the items that the user could also expect from the latent, complex and heterogeneous interactions between users, items and entities. In this paper, we define unexpectedness in the latent space rather than in the feature space and develop a novel Latent Convex Hull (LCH) method to provide unexpected recommendations. Extensive experiments on two realworld datasets demonstrate the effectiveness of the proposed model that significantly outperforms alternative stateoftheart unexpected recommendation methods in terms of unexpectedness measures while achieving the same level of accuracy. 
Latent Dirichlet Allocation (LDA) 
In natural language processing, latent Dirichlet allocation (LDA) is a generative model that allows sets of observations to be explained by unobserved groups that explain why some parts of the data are similar. For example, if observations are words collected into documents, it posits that each document is a mixture of a small number of topics and that each word’s creation is attributable to one of the document’s topics. LDA is an example of a topic model and was first presented as a graphical model for topic discovery by David Blei, Andrew Ng, and Michael Jordan in 2003. LDAvis 
Latent Dirichlet allocation Gibbs Newton (LDAGN) 
Hyperparameters play a major role in the learning and inference process of latent Dirichlet allocation (LDA). In order to begin the LDA latent variables learning process, these hyperparameters values need to be predetermined. We propose an extension for LDA that we call ‘Latent Dirichlet allocation Gibbs Newton’ (LDAGN), which places noninformative priors over these hyperparameters and uses Gibbs sampling to learn appropriate values for them. At the heart of LDAGN is our proposed ‘GibbsNewton’ algorithm, which is a new technique for learning the parameters of multivariate Polya distributions. We report GibbsNewton performance results compared with two prominent existing approaches to the latter task: Minka’s fixedpoint iteration method and the Moments method. We then evaluate LDAGN in two ways: (i) by comparing it with standard LDA in terms of the ability of the resulting topic models to generalize to unseen documents; (ii) by comparing it with standard LDA in its performance on a binary classification task. 
Latent Embedding Optimization (LEO) 
MetaLearning with Latent Embedding Optimization 
Latent Entity Typing (LET) 
Classifying semantic relations between entity pairs in sentences is an important task in Natural Language Processing (NLP). Most previous models for relation classification rely on the highlevel lexical and syntactic features obtained by NLP tools such as WordNet, dependency parser, partofspeech (POS) tagger, and named entity recognizers (NER). In addition, stateoftheart neural models based on attention mechanisms do not fully utilize information of entity that may be the most crucial features for relation classification. To address these issues, we propose a novel endtoend recurrent neural model which incorporates an entityaware attention mechanism with a latent entity typing (LET) method. Our model not only utilizes entities and their latent types as features effectively but also is more interpretable by visualizing attention mechanisms applied to our model and results of LET. Experimental results on the SemEval2010 Task 8, one of the most popular relation classification task, demonstrate that our model outperforms existing stateoftheart models without any highlevel features. 
Latent Factor Interpretation (LFI) 
Many machine learning systems utilize latent factors as internal representations for making predictions. However, since these latent factors are largely uninterpreted, predictions made using them are opaque. Collaborative filtering via matrix factorization is a prime example of such an algorithm that uses uninterpreted latent features, and yet has seen widespread adoption for many recommendation tasks. We present Latent Factor Interpretation (LFI), a method for interpreting models by leveraging interpretations of latent factors in terms of humanunderstandable features. The interpretation of latent factors can then replace the uninterpreted latent factors, resulting in a new model that expresses predictions in terms of interpretable features. This new model can then be interpreted using recently developed model explanation techniques. In this paper, we develop LFI for collaborative filtering based recommender systems, which are particularly challenging from an interpretation perspective. We illustrate the use of LFI interpretations on the MovieLens dataset demonstrating that latent factors can be predicted with enough accuracy for accurately replicating the predictions of the true model. Further, we demonstrate the accuracy of interpretations by applying the methodology to a collaborative recommender system using DB tropes and IMDB data and synthetic user preferences. 
Latent Feature Relational Model (LFRM) 
We present a discriminative nonparametric latent feature relational model (LFRM) for link prediction to automatically infer the dimensionality of latent features. Under the generic RegBayes (regularized Bayesian inference) framework, we handily incorporate the prediction loss with probabilistic inference of a Bayesian model; set distinct regularization parameters for different types of links to handle the imbalance issue in real networks; and unify the analysis of both the smooth logistic logloss and the piecewise linear hinge loss. For the nonconjugate posterior inference, we present a simple Gibbs sampler via data augmentation, without making restricting assumptions as done in variational methods. We further develop an approximate sampler using stochastic gradient Langevin dynamics to handle large networks with hundreds of thousands of entities and millions of links, orders of magnitude larger than what existing LFRM models can process. Extensive studies on various real networks show promising performance. 
Latent Gaussian Process Regression  We introduce Latent Gaussian Process Regression which is a latent variable extension allowing modelling of nonstationary processes using stationary GP priors. The approach is built on extending the input space of a regression problem with a latent variable that is used to modulate the covariance function over the input space. We show how our approach can be used to model nonstationary processes but also how multimodal or nonfunctional processes can be described where the input signal cannot fully disambiguate the output. We exemplify the approach on a set of synthetic data and provide results on real data from geostatistics. 
Latent Mixture Sampling (LMS) 
In this paper we propose a novel neural language modelling (NLM) method based on \textit{errorcorrecting output codes} (ECOC), abbreviated as ECOCNLM (errorcorrecting output codes – neural language modelling). This latent variable based approach provides a principled way to choose a varying amount of latent output codes and avoids exact softmax normalization. Instead of minimizing measures between the predicted probability distribution and true distribution, we use errorcorrecting codes to represent both predictions and outputs. Secondly, we propose multiple ways to improve accuracy and convergence rates by maximizing the separability between codes that correspond to classes proportional to word embedding similarities. Lastly, we introduce a novel method called \textit{Latent Mixture Sampling}, a technique that is used to mitigate exposure bias and can be integrated into training latentbased neural language models. This involves mixing the latent codes (i.e variables) of past predictions and past targets in one of two ways: (1) according to a predefined sampling schedule or (2) a differentiable sampling procedure whereby the mixing probability is learned throughout training by replacing the greedy argmax operation with a smooth approximation. In evaluating Codeword Mixture Sampling for ECOCNLM, we also baseline it against CWMS in a closely related Hierarhical Softmaxbased NLM. 
Latent Order Logistic (LOLOG) 
Full probability models are critical for the statistical modeling of complex networks, and yet there are few general, flexible and widely applicable generative methods. We propose a new family of probability models motivated by the idea of network growth, which we call the Latent Order Logistic (LOLOG) model. LOLOG is a fully general framework capable of describing any probability distribution over graph configurations, though not all distributions are easily expressible or estimable as a LOLOG. We develop inferential procedures based on Monte Carlo Method of Moments, Generalized Method of Moments and variational inference. To show the flexibility of the model framework, we show how socalled scalefree networks can be modeled as LOLOGs via preferential attachment. The advantages of LOLOG in terms of avoidance of degeneracy, ease of sampling, and model flexibility are illustrated. Connections with the popular Exponentialfamily Random Graph model (ERGM) are also explored, and we find that they are identical in the case of dyadic independence. Finally, we apply the model to a social network of collaboration within a corporate law firm, a friendship network among adolescent students, and the friendship relations in an online social network. lolog 
Latent Order Logistic Graph Model (LOLOG) 
➘ “Latent Order Logistic” lolog 
Latent Profile Analysis (LPA) 
The main aim of LCA is to split seemingly heterogeneous data into subclasses of two or more homogeneous groups or classes. In contrast, LPA is a method that is conducted with continuously scaled data, the focus being on generating profiles of participants instead of testing a theoretical model in terms of a measurement model, path analytic model, or full structural model, as is the case, for example, with structural equation modeling. An example of LCA and LPA,is sustainable and active travel behaviors among commuters, separating the respondents into classes based on the facilitators of, and hindrances to, certain modes of travel. Quick Example of Latent Profile Analysis in R tidyLPA 
Latent RANSAC  We present a method that can evaluate a RANSAC hypothesis in constant time, i.e. independent of the size of the data. A key observation here is that correct hypotheses are tightly clustered together in the latent parameter domain. In a manner similar to the generalized Hough transform we seek to find this cluster, only that we need as few as two votes for a successful detection. Rapidly locating such pairs of similar hypotheses is made possible by adapting the recent ‘Random Grids’ rangesearch technique. We only perform the usual (costly) hypothesis verification stage upon the discovery of a close pair of hypotheses. We show that this event rarely happens for incorrect hypotheses, enabling a significant speedup of the RANSAC pipeline. The suggested approach is applied and tested on three robust estimation problems: camera localization, 3D rigid alignment and 2Dhomography estimation. We perform rigorous testing on both synthetic and real datasets, demonstrating an improvement in efficiency without a compromise in accuracy. Furthermore, we achieve stateoftheart 3D alignment results on the challenging ‘Redwood’ loopclosure challenge. 
Latent Semantic Analysis (LSA) 
Latent semantic analysis (LSA) is a technique in natural language processing, in particular in vectorial semantics, of analyzing relationships between a set of documents and the terms they contain by producing a set of concepts related to the documents and terms. LSA assumes that words that are close in meaning will occur in similar pieces of text. A matrix containing word counts per paragraph (rows represent unique words and columns represent each paragraph) is constructed from a large piece of text and a mathematical technique called singular value decomposition (SVD) is used to reduce the number of columns while preserving the similarity structure among rows. Words are then compared by taking the cosine of the angle between the two vectors formed by any two rows. Values close to 1 represent very similar words while values close to 0 represent very dissimilar words. 
Latent Semantic Imputation (LSI) 
We present a novel method named Latent Semantic Imputation (LSI) to transfer external knowledge into semantic space for enhancing word embedding. The method integrates graph theory to extract the latent manifold structure of the entities in the affinity space and leverages nonnegative least squares with standard simplex constraints and power iteration method to derive spectral embeddings. It provides an effective and efficient approach to combining entity representations defined in different Euclidean spaces. Specifically, our approach generates and imputes reliable embedding vectors for lowfrequency words in the semantic space and benefits downstream language tasks that depend on word embedding. We conduct comprehensive experiments on a carefully designed classification problem and language modeling and demonstrate the superiority of the enhanced embedding via LSI over several wellknown benchmark embeddings. We also confirm the consistency of the results under different parameter settings of our method. 
Latent Sequence Decompositions (LSD) 
We present the Latent Sequence Decompositions (LSD) framework. LSD decomposes sequences with variable lengthed output units as a function of both the input sequence and the output sequence. We present a training algorithm which samples valid extensions and an approximate decoding algorithm. We experiment with the Wall Street Journal speech recognition task. Our LSD model achieves 12.9% WER compared to a character baseline of 14.8% WER. When combined with a convolutional network on the encoder, we achieve 9.2% WER. 
Latent State Tracking Network (LSTN) 
Recently several deep learning based models have been proposed for endtoend learning of dialogs. While these models can be trained from data without the need for any additional annotations, it is hard to interpret them. On the other hand, there exist traditional state based dialog systems, where the states of the dialog are discrete and hence easy to interpret. However these states need to be handcrafted and annotated in the data. To achieve the best of both worlds, we propose Latent State Tracking Network (LSTN) using which we learn an interpretable model in unsupervised manner. The model defines a discrete latent variable at each turn of the conversation which can take a finite set of values. Since these discrete variables are not present in the training data, we use EM algorithm to train our model in unsupervised manner. In the experiments, we show that LSTN can help achieve interpretability in dialog models without much decrease in performance compared to endtoend approaches. 
Latent Structure Analysis (LSA) 
latent structure analysis (LSA). LSA is a broad category that subsumes several individual methods, including latent class analysis (LCA) and latent trait analysis (LTA). The purpose of LSA is to infer, from observed variables (manifest variables), the structure of other, more fundamental variables that are not directly observed (latent variables). Both manifest variables and latent variables can be binary, nominal, orderedcategorical, or interval/continuous – leading to a large different combinations and different methods. For example, classical latent class analysis considers binary, nominal, or orderedcategorical manifest variables and nominal latent variables, and latent trait analysis considers binary or orderedcategorical variables and continuous latent variables. 
Latent Structure Learning (LSL) 
lsl 
Latent Topic Conversational Model (LTCM) 
Latent variable models have been a preferred choice in conversational modeling compared to sequencetosequence (seq2seq) models which tend to generate generic and repetitive responses. Despite so, training latent variable models remains to be difficult. In this paper, we propose Latent Topic Conversational Model (LTCM) which augments seq2seq with a neural latent topic component to better guide response generation and make training easier. The neural topic component encodes information from the source sentence to build a global ‘topic’ distribution over words, which is then consulted by the seq2seq model at each generation step. We study in details how the latent representation is learnt in both the vanilla model and LTCM. Our extensive experiments contribute to better understanding and training of conditional latent models for languages. Our results show that by sampling from the learnt latent representations, LTCM can generate diverse and interesting responses. In a subjective human evaluation, the judges also confirm that LTCM is the overall preferred option. 
Latent Trait Analysis (LTA) 
Latent Trait Analysis (LTA), a form of latent structure analysis (Lazarsfeld & Henry, 1968), is used for the analysis of categorical data. The simplest way to understand it is that LTA is form of factor analysis for binary (dichotomous) or orderedcategory data. In the area of educational testing and psychological measurement, latent trait analysis is termed Item Response Theory (IRT). There is so much overlap between LTA and IRT that these terms are basically interchangeable. ➚ “Item Response Theory” ltm 
Latent Transition Analysis (LTA) 
Latent transition analysis (LTA) and latent class analysis (LCA) are closely related methods. LCA identifies unobservable (latent) subgroups within a population based on individuals’ responses to multiple observed variables. LTA is an extension of LCA that uses longitudinal data to identify movement between the subgroups over time. 
Latent Tree Learning  ➘ “Latent Tree Models” 
Latent Tree Models  Latent tree models are graphical models defined on trees, in which only a subset of variables is observed. They were first discussed by Judea Pearl as treedecomposable distributions to generalise stardecomposable distributions such as the latent class model. Latent tree models, or their submodels, are widely used in: phylogenetic analysis, network tomography, computer vision, causal modeling, and data clustering. They also contain other wellknown classes of models like hidden Markov models, Brownian motion tree model, the Ising model on a tree, and many popular models used in phylogenetics. This article offers a concise introduction to the theory of latent tree models. We emphasise the role of tree metrics in the structural description of this model class, in designing learning algorithms, and in understanding fundamental limits of what and when can be learned. 
Latent Tree Variational Autoencoder (LTVAE) 
Recently, deep learning based clustering methods are shown superior to traditional ones by jointly conducting representation learning and clustering. These methods rely on the assumptions that the number of clusters is known, and that there is one single partition over the data and all attributes define that partition. However, in realworld applications, prior knowledge of the number of clusters is usually unavailable and there are multiple ways to partition the data based on subsets of attributes. To resolve the issues, we propose latent tree variational autoencoder (LTVAE), which simultaneously performs representation learning and multidimensional clustering. LTVAE learns latent embeddings from data, discovers multifacet clustering structures based on subsets of latent features, and automatically determines the number of clusters in each facet. Experiments show that the proposed method achieves stateoftheart clustering performance and reals reasonable multifacet structures of the data. 
Latent Truth Discovery (LTD) 
Latent truth discovery, LTD for short, refers to the problem of aggregating multiple claims from various sources in order to estimate the plausibility of statements about entities. In the absence of a ground truth, this problem is highly challenging, when some sources provide conflicting claims and others no claims at all. 
Latent Variable Discovery Algorithm (LatentSearch) 
We consider the problem of discovering the simplest latent variable that can make two observed discrete variables conditionally independent. This problem has appeared in the literature as probabilistic latent semantic analysis (pLSA), and has connections to nonnegative matrix factorization. When the simplicity of the variable is measured through its cardinality, we show that a solution to this latent variable discovery problem can be used to distinguish direct causal relations from spurious correlations among almost all joint distributions on simple causal graphs with two observed variables. Conjecturing a similar identifiability result holds with Shannon entropy, we study a loss function that tradesoff between entropy of the latent variable and the conditional mutual information of the observed variables. We then propose a latent variable discovery algorithm — LatentSearch — and show that its stationary points are the stationary points of our loss function. We experimentally show that LatentSearch can indeed be used to distinguish direct causal relations from spurious correlations. 
Latent Variable Mixture Model (LVMM) 
Latent variable mixture modeling (LVMM) is a flexible analytic tool that allows researchers to investigate questions about patterns of data and to determine the extent to which identified patterns relate to important variables. For example, Ā· Do patterns of cooccurring developmental and medical diagnoses influence the severity of pediatric feeding problems (Berlin, Lobato, Pinkos, Cerezo, & LeLeiko, 2011)? Ā· Do differential longitudinal trajectories of glycemic control exist among youth with type 1 diabetes (Helgeson et al., 2010) Ā· Do differential trajectories of adherence among youth newly diagnosed with epilepsy exist (Modi, Rausch, & Glauser, 2011), and if so, Ā· Do psychosocial and demographic variables predict these patterns? Ā· Do patterns of perceived stressors among youth with type 1 diabetes differentially affect glycemic control (Berlin, Rabideau, & Hains, 2012)? http://…cgi?article=1093&context=famconfacpub http://…/latentvariablemixturemodelslvmm.html 
Latent Variable Model  A latent variable model is a statistical model that relates a set of variables (socalled manifest variables) to a set of latent variables. It is assumed that the responses on the indicators or manifest variables are the result of an individual’s position on the latent variable(s), and that the manifest variables have nothing in common after controlling for the latent variable (local independence). Different types of the latent variable model can be grouped according to whether the manifest and latent variables are categorical or continuous. 
Latin Hypercube Design  MOLHD 
Latitude  Nonnegative matrix factorization (NMF) is one of the most frequentlyused matrix factorization models in data analysis. A significant reason to the popularity of NMF is its interpretability and the `parts of whole’ interpretation of its components. Recently, maxtimes, or subtropical, matrix factorization (SMF) has been introduced as an alternative model with equally interpretable `winner takes it all’ interpretation. In this paper we propose a new mixed linear–tropical model, and a new algorithm, called Latitude, that combines NMF and SMF, being able to smoothly alternate between the two. In our model, the data is modeled using the latent factors and latent parameters that control whether the factors are interpreted as NMF or SMF features, or their mixtures. We present an algorithm for our novel matrix factorization. Our experiments show that our algorithm improves over both baselines, and can yield interpretable results that reveal more of the latent structure than either NMF or SMF alone. 
Lavaan Project  The lavaan package is developed to provide useRs, researchers and teachers a free opensource, but commercialquality package for latent variable modeling. You can use lavaan to estimate a large variety of multivariate statistical models, including path analysis, confirmatory factor analysis, structural equation modeling and growth curve models. The official reference to the lavaan package is the following paper: Yves Rosseel (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 136. URL http://…/i02 lavaan,lavaan.shiny,lavaanPlot,blavaan 
Law of Likelihood  If hypothesis A implies that the probability that a random variable X takes the value x is pA(x), while hypothesis B implies that the probability is pB(x), then the observation X = x is evidence supporting A over B if and only if pA(x) > pB(x), and the likelihood ratio, pA(x)/ pB(x), measures the strength of that evidence.’ ‘This says simply that if an event is more probable under hypothesis A than hypothesis B, then the occurrence of that event is evidence supporting A over B – the hypothesis that did the better job of predicting the event is better supported by its occurrence.’ Moreover, ‘the likelihood ratio, is the exact factor by which the probability ratio (ratio of priors in A and B) is changed. 
Law of Total Expectation  The proposition in probability theory known as the law of total expectation, the law of iterated expectations, the tower rule, Adam’s law, and the smoothing theorem, among other names, states that if X is a random variable whose expected value E( X ) is defined, and Y is any random variable on the same probability space, then E(X) = E(E(XY)) , i.e., the expected value of the conditional expected value of X given Y is the same as the expected value of X. 
Layer Reuse Network (LruNet) 
A convolutional layer in a Convolutional Neural Network (CNN) consists of many filters which apply convolution operation to the input, capture some special patterns and pass the result to the next layer. If the same patterns also occur at the deeper layers of the network, why wouldn’t the same convolutional filters be used also in those layers? In this paper, we propose a CNN architecture, Layer Reuse Network (LruNet), where the convolutional layers are used repeatedly without the need of introducing new layers to get a better performance. This approach introduces several advantages: (i) Considerable amount of parameters are saved since we are reusing the layers instead of introducing new layers, (ii) the Memory Access Cost (MAC) can be reduced since reused layer parameters can be fetched only once, (iii) the number of nonlinearities increases with layer reuse, and (iv) reused layers get gradient updates from multiple parts of the network. The proposed approach is evaluated on CIFAR10, CIFAR100 and FashionMNIST datasets for image classification task, and layer reuse improves the performance by 5.14%, 5.85% and 2.29%, respectively. The source code and pretrained models are publicly available. 
Layer Trajectory LSTM (ltLSTM) 
It is popular to stack LSTM layers to get better modeling power, especially when large amount of training data is available. However, an LSTMRNN with too many vanilla LSTM layers is very hard to train and there still exists the gradient vanishing issue if the network goes too deep. This issue can be partially solved by adding skip connections between layers, such as residual LSTM. In this paper, we propose a layer trajectory LSTM (ltLSTM) which builds a layerLSTM using all the layer outputs from a standard multilayer timeLSTM. This layerLSTM scans the outputs from timeLSTMs, and uses the summarized layer trajectory information for final senone classification. The forwardpropagation of timeLSTM and layerLSTM can be handled in two separate threads in parallel so that the network computation time is the same as the standard timeLSTM. With a layerLSTM running through layers, a gated path is provided from the output layer to the bottom layer, alleviating the gradient vanishing issue. Trained with 30 thousand hours of ENUS Microsoft internal data, the proposed ltLSTM performed significantly better than the standard multilayer LSTM and residual LSTM, with up to 9.0% relative word error rate reduction across different tasks. 
Layered SelfOrganizing Map (LSOM) 
This paper defines a new learning architecture, Layered SelfOrganizing Maps (LSOMs), that uses the SOM and supervisedSOM learning algorithms. The architecture is validated with the MNIST database of handwritten digit images. LSOMs are similar to convolutional neural nets (covnets) in the way they sample data, but different in the way they represent features and learn. LSOMs analyze (or generate) image patches with maps of exemplars determined by the SOM learning algorithm rather than feature maps from filterbanks learned via backprop. LSOMs provide an alternative to features derived from covnets. Multilayer LSOMs are trained bottomup, without the use of backprop and therefore may be of interest as a model of the visual cortex. The results show organization at multiple levels. The algorithm appears to be resource efficient in learning, classifying and generating images. Although LSOMs can be used for classification, their validation accuracy for these exploratory runs was well below the state of the art. The goal of this article is to define the architecture and display the structures resulting from its application to the MNIST images. 
Layered TreeBased Pipeline Optimization Tool (Layered TPOT) 
With the demand for machine learning increasing, so does the demand for tools which make it easier to use. Automated machine learning (AutoML) tools have been developed to address this need, such as the TreeBased Pipeline Optimization Tool (TPOT) which uses genetic programming to build optimal pipelines. We introduce Layered TPOT, a modification to TPOT which aims to create pipelines equally good as the original, but in significantly less time. This approach evaluates candidate pipelines on increasingly large subsets of the data according to their fitness, using a modified evolutionary algorithm to allow for separate competition between pipelines trained on different sample sizes. Empirical evaluation shows that, on sufficiently large datasets, Layered TPOT indeed finds better models faster. ➘ “TreeBased Pipeline Optimization Tool” 
LayerWise Relevance Propagation (LRP) 
Despite the tremendous achievements of deep convolutional neural networks~(CNNs) in most of computer vision tasks, understanding how they actually work remains a significant challenge. In this paper, we propose a novel twostep visualization method that aims to shed light on how deep CNNs recognize images and the objects therein. We start out with a layerwise relevance propagation (LRP) step which estimates a pixelwise relevance map over the input image. Following, we construct a contextaware saliency map from the LRPgenerated map which predicts regions close to the foci of attention. We show that our algorithm clearly and concisely identifies the key pixels that contribute to the underlying neural network’s comprehension of images. Experimental results using the ILSVRC2012 validation dataset in conjunction with two wellestablished deep CNNs demonstrate that combining the LRP with the visual salience estimation can give great insight into how a CNNs model perceives and understands a presented scene, in relation to what it has learned in the prior training phase. 
LayerwIseSemanticAccumulation (LISA) 
Recurrent neural networks (RNNs) are temporal networks and cumulative in nature that have shown promising results in various natural language processing tasks. Despite their success, it still remains a challenge to understand their hidden behavior. In this work, we analyze and interpret the cumulative nature of RNN via a proposed technique named as LayerwIseSemanticAccumulation (LISA) for explaining decisions and detecting the most likely (i.e., saliency) patterns that the network relies on while decision making. We demonstrate (1) LISA: ‘How an RNN accumulates or builds semantics during its sequential processing for a given text example and expected response’ (2) Example2pattern: ‘How the saliency patterns look like for each category in the data according to the network in decision making’. We analyse the sensitiveness of RNNs about different inputs to check the increase or decrease in prediction scores and further extract the saliency patterns learned by the network. We employ two relation classification datasets: SemEval 10 Task 8 and TAC KBP Slot Filling to explain RNN predictions via the LISA and example2pattern. 
Lazily Aggregated Gradient Coding (LAGC) 
Gradientbased distributed learning in Parameter Server (PS) computing architectures is subject to random delays due to straggling worker nodes, as well as to possible communication bottlenecks between PS and workers. Solutions have been recently proposed to separately address these impairments based on the ideas of gradient coding, worker grouping, and adaptive worker selection. This paper provides a unified analysis of these techniques in terms of wallclock time, communication, and computation complexity measures. Furthermore, in order to combine the benefits of gradient coding and grouping in terms of robustness to stragglers with the communication and computation load gains of adaptive selection, novel strategies, named Lazily Aggregated Gradient Coding (LAGC) and GroupedLAG (GLAG), are introduced. Analysis and results show that GLAG provides the best wallclock time and communication performance, while maintaining a low computational cost, for two representative distributions of the computing times of the worker nodes. 
Lazy Bayesian Rules (LBR) 
The naive Bayesian classifier provides a simple and effective approach to classifier learning, but its attribute independence assumption is often violated in the real world. A number of approaches have sought to alleviate this problem. A Bayesian tree learning algorithm builds a decision tree, and generates a local naive Bayesian classifier at each leaf. The tests leading to a leaf can alleviate attribute interdependencies for the local naive Bayesian classifier. However, Bayesian tree learning still suffers from the small disjunct problem of tree learning. While inferred Bayesian trees demonstrate low average prediction error rates, there is reason to believe that error rates will be higher for those leaves with few training examples. This paper proposes the application of lazy learning techniques to Bayesian tree induction and presents the resulting lazy Bayesian rule learning algorithm, called Lbr. This algorithm can be justified by a variant of Bayes theorem which supports a weaker conditional attribute independence assumption than is required by naive Bayes. For each test example, it builds a most appropriate rule with a local naive Bayesian classifier as its consequent. It is demonstrated that the computational requirements of Lbr are reasonable in a wide crosssection of natural domains. Experiments with these domains show that, on average, this new algorithm obtains lower error rates significantly more often than the reverse in comparison to a naive Bayesian classifier, C4.5, a Bayesian tree learning algorithm, a constructive Bayesian classifier that eliminates attributes and constructs new attributes using Cartesian products of existing nominal attributes, and a lazy decision tree learning algorithm. It also outperforms, although the result is not statistically significant, a selective naive Bayesian classifier. http://…/ZhengWebbTing99.pdf http://…/CRPITV87Xie.pdf 
Lazy Learning  In artificial intelligence, lazy learning is a learning method in which generalization beyond the training data is delayed until a query is made to the system, as opposed to in eager learning, where the system tries to generalize the training data before receiving queries. The main advantage gained in employing a lazy learning method, such as Case based reasoning, is that the target function will be approximated locally, such as in the knearest neighbor algorithm. Because the target function is approximated locally for each query to the system, lazy learning systems can simultaneously solve multiple problems and deal successfully with changes in the problem domain. The disadvantages with lazy learning include the large space requirement to store the entire training dataset. Particularly noisy training data increases the case base unnecessarily, because no abstraction is made during the training phase. Another disadvantage is that lazy learning methods are usually slower to evaluate, though this is coupled with a faster training phase. Lazy classifiers are most useful for large datasets with few attributes. 
Lazy Stochastic Principal Component Analysis (Lazy SPCA) 
Stochastic principal component analysis (SPCA) has become a popular dimensionality reduction strategy for large, highdimensional datasets. We derive a simplified algorithm, called Lazy SPCA, which has reduced computational complexity and is better suited for largescale distributed computation. We prove that SPCA and Lazy SPCA find the same approximations to the principal subspace, and that the pairwise distances between samples in the lowerdimensional space is invariant to whether SPCA is executed lazily or not. Empirical studies find downstream predictive performance to be identical for both methods, and superior to random projections, across a range of predictive models (linear regression, logistic lasso, and random forests). In our largest experiment with 4.6 million samples, Lazy SPCA reduced 43.7 hours of computation to 9.9 hours. Overall, Lazy SPCA relies exclusively on matrix multiplications, besides an operation on a small square matrix whose size depends only on the target dimensionality. 
lazydata  lazydata is a minimalist library for including data dependencies into Python projects. Problem: Keeping all data files in git (e.g. via gitlfs) results in a bloated repository copy that takes ages to pull. Keeping code and data out of sync is a disaster waiting to happen. Solution: lazydata only stores references to data files in git, and syncs data files ondemand when they are needed. Why: The semantics of code and data are different – code needs to be versioned to merge it, and data just needs to be kept in sync. lazydata achieves exactly this in a minimal way. 
LConvex Set  We investigate an enrichedcategorical approach to a field of discrete mathematics. The main result is a duality theorem between a class of enriched categories (called $\overline{\mathbb{Z}}$ or $\overline{\mathbb{R}}$categories) and that of what we call ($\overline{\mathbb{Z}}$ or $\overline{\mathbb{R}}$) extended Lconvex sets. We introduce extended Lconvex sets as variants of certain discrete structures called Lconvex sets and Lconvex polyhedra, studied in the field of discrete convex analysis. We also introduce homomorphisms between extended Lconvex sets. The theorem claims that there is a one to one correspondence (up to isomorphism) between two classes. The thesis also contains an introductory chapter on enriched categories and no categorical knowledge is assumed. 
lda2vec  Standard natural language processing (NLP) is a messy and difficult affair. It requires teaching a computer about Englishspecific word ambiguities as well as the hierarchical, sparse nature of words in sentences. At Stitch Fix, word vectors help computers learn from the raw text in customer notes. Our systems need to identify a medical professional when she writes that she ‘used to wear scrubs to work’, and distill ‘taking a trip’ into a Fix for vacation clothing. Applied appropriately, word vectors are dramatically more meaningful and more flexible than current techniques and let computers peer into text in a fundamentally new way. I’ll try to convince you that word vectors give us a simple and flexible platform for understanding text while speaking about word2vec, LDA, and introduce our hybrid algorithm lda2vec. 
LDMI  Word2Vec’s Skip Gram model is the current stateoftheart approach for estimating the distributed representation of words. However, it assumes a single vector per word, which is not wellsuited for representing words that have multiple senses. This work presents LDMI, a new model for estimating distributional representations of words. LDMI relies on the idea that, if a word carries multiple senses, then having a different representation for each of its senses should lead to a lower loss associated with predicting its cooccurring words, as opposed to the case when a single vector representation is used for all the senses. After identifying the multisense words, LDMI clusters the occurrences of these words to assign a sense to each occurrence. Experiments on the contextual word similarity task show that LDMI leads to better performance than competing approaches. 
Leader Clustering Algorithm  The leader clustering algorithm provides a means for clustering a set of data points. Unlike many other clustering algorithms it does not require the user to specify the number of clusters, but instead requires the approximate radius of a cluster as its primary tuning parameter. The package provides a fast implementation of this algorithm in ndimensions using Lpdistances (with special cases for p=1,2, and infinity) as well as for spatial data using the Haversine formula, which takes latitude/longitude pairs as inputs and clusters based on great circle distances. leaderCluster 
Leader Gradient Descent (LGD) 
We consider distributed optimization under communication constraints for training deep learning models. We propose a new algorithm, whose parameter updates rely on two forces: a regular gradient step, and a corrective direction dictated by the currently bestperforming worker (leader). Our method differs from the parameteraveraging scheme EASGD in a number of ways: (i) our objective formulation does not change the location of stationary points compared to the original optimization problem; (ii) we avoid convergence decelerations caused by pulling local workers descending to different local minima to each other (i.e. to the average of their parameters); (iii) our update by design breaks the curse of symmetry (the phenomenon of being trapped in poorly generalizing suboptimal solutions in symmetric nonconvex landscapes); and (iv) our approach is more communication efficient since it broadcasts only parameters of the leader rather than all workers. We provide theoretical analysis of the batch version of the proposed algorithm, which we call Leader Gradient Descent (LGD), and its stochastic variant (LSGD). Finally, we implement an asynchronous version of our algorithm and extend it to the multileader setting, where we form groups of workers, each represented by its own local leader (the best performer in a group), and update each worker with a corrective direction comprised of two attractive forces: one to the local, and one to the global leader (the best performer among all workers). The multileader setting is wellaligned with current hardware architecture, where local workers forming a group lie within a single computational node and different groups correspond to different nodes. For training convolutional neural networks, we empirically demonstrate that our approach compares favorably to stateoftheart baselines. 
LeaderFollower Elastic Averaging Stochastic Gradient Descent (LEASGD) 
Distributed learning systems have enabled training largescale models over large amount of data in significantly shorter time. In this paper, we focus on decentralized distributed deep learning systems and aim to achieve differential privacy with good convergence rate and low communication cost. To achieve this goal, we propose a new learning algorithm LEASGD (LeaderFollower Elastic Averaging Stochastic Gradient Descent), which is driven by a novel LeaderFollower topology and a differential privacy model.We provide a theoretical analysis of the convergence rate and the tradeoff between the performance and privacy in the private setting.The experimental results show that LEASGD outperforms stateoftheart decentralized learning algorithm DPSGD by achieving steadily lower loss within the same iterations and by reducing the communication cost by 30%. In addition, LEASGD spends less differential privacy budget and has higher final accuracy result than DPSGD under private setting. 
Leaders and Subleaders Algorithm  An efficient hierarchical clustering algorithm, suitable for large data sets is proposed for effective clustering and prototype selection for pattern classification. It is another simple and efficient technique which uses incremental clustering principles to generate a hierarchical structure for finding the subgroups/subclusters within each cluster. As an example, a two level clustering algorithm – LeadersSubleaders, an extension of the leader algorithm is presented. Classification accuracy (CA) obtained using the representatives generated by the LeadersSubleaders method is found to be better than that of using leaders as representatives. Even if more number of prototypes are generated, classification time is less as only a part of the hierarchical structure is searched. 
Leadlike Recognizer (LeadR) 
A competitive baseline in sentencelevel extractive summarization of news articles is the Lead3 heuristic, where only the first 3 sentences are extracted. The success of this method is due to the tendency for writers to implement progressive elaboration in their work by writing the most important content at the beginning. In this paper, we introduce the Leadlike Recognizer (LeadR) to show how the Lead heuristic can be extended to summarize multisection documents where it would not usually work well. This is done by introducing a neural model which produces a probability distribution over positions for sentences, so that we can locate sentences with introductionlike qualities. To evaluate the performance of our model, we use the task of summarizing multisection documents. LeadR outperforms several baselines on this task, including a simple extension of the Lead heuristic designed for the task. Our work suggests that predicted position is a strong feature to use when extracting summaries. 
LEAF  Deep neural networks (DNNs) have produced stateoftheart results in many benchmarks and problem domains. However, the success of DNNs depends on the proper configuration of its architecture and hyperparameters. Such a configuration is difficult and as a result, DNNs are often not used to their full potential. In addition, DNNs in commercial applications often need to satisfy realworld design constraints such as size or number of parameters. To make configuration easier, automatic machine learning (AutoML) systems for deep learning have been developed, focusing mostly on optimization of hyperparameters. This paper takes AutoML a step further. It introduces an evolutionary AutoML framework called LEAF that not only optimizes hyperparameters but also network architectures and the size of the network. LEAF makes use of both stateoftheart evolutionary algorithms (EAs) and distributed computing frameworks. Experimental results on medical image classification and natural language analysis show that the framework can be used to achieve stateoftheart performance. In particular, LEAF demonstrates that architecture optimization provides a significant boost over hyperparameter optimization, and that networks can be minimized at the same time with little drop in performance. LEAF therefore forms a foundation for democratizing and improving AI, as well as making AI practical in future applications. 
LEAFAGE  As machine learning models become more accurate, they typically become more complex and uninterpretable by humans. The blackbox character of these models holds back its acceptance in practice, especially in highrisk domains where the consequences of failure could be catastrophic such as healthcare or defense. Providing understandable and useful explanations behind ML models or predictions can increase the trust of the user. Examplebased reasoning, which entails leveraging previous experience with analogous tasks to make a decision, is a well known strategy for problem solving and justification. This work presents a new explanation extraction method called LEAFAGE, for a prediction made by any blackbox ML model. The explanation consists of the visualization of similar examples from the training set and the importance of each feature. Moreover, these explanations are contrastive which aims to take the expectations of the user into account. LEAFAGE is evaluated in terms of fidelity to the underlying blackbox model and usefulness to the user. The results showed that LEAFAGE performs overall better than the current stateoftheart method LIME in terms of fidelity, on ML models with nonlinear decision boundary. A userstudy was conducted which focused on revealing the differences between examplebased and feature importancebased explanations. It showed that examplebased explanations performed significantly better than feature importancebased explanation, in terms of perceived transparency, information sufficiency, competence and confidence. Counterintuitively, when the gained knowledge of the participants was tested, it showed that they learned less about the blackbox model after seeing a feature importancebased explanation than seeing no explanation at all. The participants found feature importancebased explanation vague and hard to generalize it to other instances. 
leaflet  Leaflet is a modern opensource JavaScript library for mobilefriendly interactive maps. It is developed by Vladimir Agafonkin with a team of dedicated contributors. Weighing just about 33 KB of JS, it has all the features most developers ever need for online maps. Leaflet is designed with simplicity, performance and usability in mind. It works efficiently across all major desktop and mobile platforms out of the box, taking advantage of HTML5 and CSS3 on modern browsers while still being accessible on older ones. It can be extended with a huge amount of plugins, has a beautiful, easy to use and welldocumented API and a simple, readable source code that is a joy to contribute to. http://…neo4jspatialandleafletjswithmapbox Leaflet: Interactive web maps with R leaflet 
Lean Analytics  Lean Analytics is about measuring the right thing, in the right way, to produce the change the business needs the most at that point in time. With that in mind, here’s some background on metrics that matter. 
LeanResNet  Convolutional Neural Networks (CNNs) filter the input data using a series of spatial convolution operators with compact stencils and pointwise nonlinearities. Commonly, the convolution operators couple features from all channels, which leads to immense computational cost in the training of and prediction with CNNs. To improve the efficiency of CNNs, we introduce lean convolution operators that reduce the number of parameters and computational complexity. Our new operators can be used in a wide range of existing CNNs. Here, we exemplify their use in residual networks (ResNets), which have been very reliable for a few years now and analyzed intensively. In our experiments on three image classification problems, the proposed LeanResNet yields results that are comparable to other recently proposed reduced architectures using similar number of parameters. 
Leap  In complex transfer learning scenarios new tasks might not be tightly linked to previous tasks. Approaches that transfer information contained only in the final parameters of a source model will therefore struggle. Instead, transfer learning at a higher level of abstraction is needed. We propose Leap, a framework that achieves this by transferring knowledge across learning processes. We associate each task with a manifold on which the training process travels from initialization to final parameters and construct a meta learning objective that minimizes the expected length of this path. Our framework leverages only information obtained during training and can be computed on the fly at negligible cost. We demonstrate that our framework outperforms competing methods, both in meta learning and transfer learning, on a set of computer vision tasks. Finally, we demonstrate that Leap can transfer knowledge across learning processes in demanding Reinforcement Learning environments (Atari) that involve millions of gradient steps. 
learn2search  We consider the problem of searching in a set of items by using pairwise comparisons. We aim to locate a target item $t$ by asking an oracle questions of the form ‘Which item from the pair $(i,j)$ is more similar to t?’. We assume a blind setting, where no item features are available to guide the search process; only the oracle sees the features in order to generate an answer. Previous approaches for this problem either assume noiseless answers, or they scale poorly in the number of items, both of which preclude practical applications. In this paper, we present a new scalable learning framework called learn2search that performs efficient comparisonbased search on a set of items despite the presence of noise in the answers. Items live in a space of latent features, and we posit a probabilistic model for the oracle comparing two items $i$ and $j$ with respect to a target $t$. Our algorithm maintains its own representation of the space of items, which it learns incrementally based on past searches. We evaluate the performance of learn2search on both synthetic and realworld data, and show that it learns to search more and more efficiently, over time matching the performance of a scheme with access to the item features. 
Learnable Graph Convolutional Layer (LGCL) 
Convolutional neural networks (CNNs) have achieved great success on gridlike data such as images, but face tremendous challenges in learning from more generic data such as graphs. In CNNs, the trainable local filters enable the automatic extraction of highlevel features. The computation with filters requires a fixed number of ordered units in the receptive fields. However, the number of neighboring units is neither fixed nor are they ordered in generic graphs, thereby hindering the applications of convolutional operations. Here, we address these challenges by proposing the learnable graph convolutional layer (LGCL). LGCL automatically selects a fixed number of neighboring nodes for each feature based on value ranking in order to transform graph data into gridlike structures in 1D format, thereby enabling the use of regular convolutional operations on generic graphs. To enable model training on largescale graphs, we propose a subgraph training method to reduce the excessive memory and computational resource requirements suffered by prior methods on graph convolutions. Our experimental results on node classification tasks in both transductive and inductive learning settings demonstrate that our methods can achieve consistently better performance on the Cora, Citeseer, Pubmed citation network, and proteinprotein interaction network datasets. Our results also indicate that the proposed methods using subgraph training strategy are more efficient as compared to prior approaches. 
Learnable Histogram  Statistical features, such as histogram, BagofWords (BoW) and Fisher Vector, were commonly used with handcrafted features in conventional classification methods, but attract less attention since the popularity of deep learning methods. In this paper, we propose a learnable histogram layer, which learns histogram features within deep neural networks in endtoend training. Such a layer is able to backpropagate (BP) errors, learn optimal bin centers and bin widths, and be jointly optimized with other layers in deep networks during training. Two vision problems, semantic segmentation and object detection, are explored by integrating the learnable histogram layer into deep networks, which show that the proposed layer could be well generalized to different applications. Indepth investigations are conducted to provide insights on the newly introduced layer. 
Learnable ScatterNet  In this paper we explore tying together the ideas from Scattering Transforms and Convolutional Neural Networks (CNN) for Image Analysis by proposing a learnable ScatterNet. Previous attempts at tying them together in hybrid networks have tended to keep the two parts separate, with the ScatterNet forming a fixed front end and a CNN forming a learned backend. We instead look at adding learning between scattering orders, as well as adding learned layers before the ScatterNet. We do this by breaking down the scattering orders into single convolutionallike layers we call ‘locally invariant’ layers, and adding a learned mixing term to this layer. Our experiments show that these locally invariant layers can improve accuracy when added to either a CNN or a ScatterNet. We also discover some surprising results in that the ScatterNet may be best positioned after one or more layers of learning rather than at the front of a neural network. 
LearnAsYouGo (LAGO) 
Analysis of ‘LearnAsYouGo’ (LAGO) Studies 
Learned Global Ranking (LeGR) 
Filter pruning has shown to be effective for learning resourceconstrained convolutional neural networks (CNNs). However, prior methods for resourceconstrained filter pruning have some limitations that hinder their effectiveness and efficiency. When searching for constraintsatisfying CNNs, prior methods either alter the optimization objective or adopt local search algorithms with heuristic parameterization, which are suboptimal, especially in lowresource regime. From the efficiency perspective, prior methods are often costly to search for constraintsatisfying CNNs. In this work, we propose learned global ranking, dubbed LeGR, which improves upon prior art in the two aforementioned dimensions. Inspired by theoretical analysis, LeGR is parameterized to learn layerwise affine transformations over the filter norms to construct a learned global ranking. With global ranking, resourceconstrained filter pruning at various constraint levels can be done efficiently. We conduct extensive empirical analyses to demonstrate the effectiveness of the proposed algorithm with ResNet and MobileNetV2 networks on CIFAR10, CIFAR100, Bird200, and ImageNet datasets. Code is publicly available at https://…/LeGR. 
Learned Group Signals  We study a model where a data collector obtains data from users through a payment mechanism, aiming to learn the underlying state from the elicited data. The private signal of each user represents her knowledge about the state; and through social interactions each user can also learn noisy versions of her social friends’ signals, which is called `learned group signals’. Thanks to social learning, users have richer information about the state beyond their private signals. Based on both her private signal and learned group signals, each user makes strategic decisions to report a privacypreserved version of her data to the data collector. We develop a Bayesian game theoretic framework to study the impact of social learning on users’ data reporting strategies and devise the payment mechanism for the data collector accordingly. Our findings reveal that, in general, the desired data reporting strategy at the BayesianNash equilibrium can be in the form of either a symmetric randomized response (SR) strategy or an informative nondisclosive (ND) strategy. Specifically, a generalized majority voting rule is applied by each user to her noisy group signals to determine which strategy to follow. Further, when a user plays the ND strategy, she reports privacypreserving data completely based on her group signals, independent of her private signal, which indicates that her privacy cost is zero. We emphasize that the reported data when a user plays the ND strategy is still informative about the underlying state because it is based on her learned group signals. As a result, both the data collector and the users can benefit from social learning which drives down the privacy costs and helps to improve the state estimation at a given payment budget. We further derive bounds on the minimum total payment required to achieve a given level of state estimation accuracy. 
Learned Step Size Quantization  We present here Learned Step Size Quantization, a method for training deep networks such that they can run at inference time using low precision integer matrix multipliers, which offer power and space advantages over high precision alternatives. The essence of our approach is to learn the step size parameter of a uniform quantizer by backpropagation of the training loss, applying a scaling factor to its learning rate, and computing its associated loss gradient by ignoring the discontinuity present in the quantizer. This quantization approach can be applied to activations or weights, using different levels of precision as needed for a given system, and requiring only a simple modification of existing training code. As demonstrated on the ImageNet dataset, our approach achieves better accuracy than all previous published methods for creating quantized networks on several ResNet network architectures at 2, 3 and 4bits of precision. 
Learning Active Learning (LAL) 
In this paper, we suggest a novel datadriven approach to active learning: Learning Active Learning (LAL). The key idea behind LAL is to train a regressor that predicts the expected error reduction for a potential sample in a particular learning state. By treating the query selection procedure as a regression problem we are not restricted to dealing with existing AL heuristics; instead, we learn strategies based on experience from previous active learning experiments. We show that LAL can be learnt from a simple artificial 2D dataset and yields strategies that work well on real data from a wide range of domains. Moreover, if some domainspecific samples are available to bootstrap active learning, the LAL strategy can be tailored for a particular problem. 
Learning Analytics (LA) 
Learning analytics is the measurement, collection, analysis and reporting of data about learners and their contexts, for purposes of understanding and optimising learning and the environments in which it occurs. A related field is educational data mining. For general audience introductions, see: Ā· The Educause Learning Initiative Briefing Ā· The Educause Review on Learning analytics Ā· And the UNESCO “Learning Analytics Policy Brief” (2012) 
LEArning and Planning with Semantics (LEAP) 
Building deep reinforcement learning agents that can generalize and adapt to unseen environments remains a fundamental challenge for AI. This paper describes progresses on this challenge in the context of manmade environments, which are visually diverse but contain intrinsic semantic regularities. We propose a hybrid modelbased and modelfree approach, LEArning and Planning with Semantics (LEAPS), consisting of a multitarget subpolicy that acts on visual inputs, and a Bayesian model over semantic structures. When placed in an unseen environment, the agent plans with the semantic model to make highlevel decisions, proposes the next subtarget for the subpolicy to execute, and updates the semantic model based on new observations. We perform experiments in visual navigation tasks using House3D, a 3D environment that contains diverse humandesigned indoor scenes with realworld objects. LEAPS outperforms strong baselines that do not explicitly plan using the semantic content. 
Learning Automata Based SVM (LASVM) 
As an indispensable defensive measure of network security, the intrusion detection is a process of monitoring the events occurring in a computer system or network and analyzing them for signs of possible incidents. It is a classifier to judge the event is normal or malicious. The information used for intrusion detection contains some redundant features which would increase the difficulty of training the classifier for intrusion detection and increase the time of making predictions. To simplify the training process and improve the efficiency of the classifier, it is necessary to remove these dispensable features. in this paper, we propose a novel LASVM scheme to automatically remove redundant features focusing on intrusion detection. This is the first application of learning automata for solving dimension reduction problems. The simulation results indicate that the LASVM scheme achieves a higher accuracy and is more efficient in making predictions compared with traditional SVM. 
Learning AutomataBased QLearning (LAQL) 
An optimization problem of content placement in cooperative caching is formulated, with the aim of maximizing sum mean opinion score (MOS) of mobile users. Firstly, a supervised feedforward backpropagation connectionist model based neural network (SFBCNN) is invoked for user mobility and content popularity prediction. More particularly, practical data collected from GPStracker app on smartphones is tackled to test the accuracy of mobility prediction. Then, a learning automatabased Qlearning (LAQL) algorithm for cooperative caching is proposed, in which learning automata (LA) is invoked for Qlearning to obtain an optimal action selection in a random and stationary environment. It is proven that the LAbased action selection scheme is capable of enabling every state to select the optimal action with arbitrarily high probability if Qlearning is able to converge to the optimal Q value eventually. To characterize the performance of the proposed algorithms, the sum MOS of users is applied to define the reward function. Extensive simulations reveal that: 1) The prediction error of SFBCNN lessen with the increase of iterations and nodes; 2) the proposed LAQL achieves significant performance improvement against traditional Qlearning; 3) the cooperative caching scheme is capable of outperforming noncooperative caching and random caching of 3% and 4%. 
Learning by Association  In many realworld scenarios, labeled data for a specific machine learning task is costly to obtain. Semisupervised training methods make use of abundantly available unlabeled data and a smaller number of labeled examples. We propose a new framework for semisupervised training of deep neural networks inspired by learning in humans. ‘Associations’ are made from embeddings of labeled samples to those of unlabeled ones and back. The optimization schedule encourages correct association cycles that end up at the same class from which the association was started and penalizes wrong associations ending at a different class. The implementation is easy to use and can be added to any existing endtoend training setup. We demonstrate the capabilities of learning by association on several data sets and show that it can improve performance on classification tasks tremendously by making use of additionally available unlabeled data. In particular, for cases with few labeled data, our training scheme outperforms the current state of the art on SVHN. 
Learning by Unsupervised Nonlinear Diffusion (LUND) 
This paper proposes and analyzes a novel clustering algorithm that combines graphbased diffusion geometry with density estimation. The proposed method is suitable for data generated from mixtures of distributions with densities that are both multimodal and have nonlinear shapes. A crucial aspect of this algorithm is to introduce time of a dataadapted diffusion process as a scale parameter that is different from the local spatial scale parameter used in many clustering and learning algorithms. We prove estimates for the behavior of diffusion distances with respect to this time parameter under a flexible nonparametric data model, identifying a range of times in which the mesoscopic equilibria of the underlying process are revealed, corresponding to a gap between withincluster and betweencluster diffusion distances. This analysis is leveraged to prove sufficient conditions guaranteeing the accuracy of the proposed learning by unsupervised nonlinear diffusion (LUND) algorithm. We implement the LUND algorithm numerically and confirm its theoretical properties on illustrative datasets, showing that the proposed method enjoys both theoretical and empirical advantages over current spectral clustering and densitybased clustering techniques. 
Learning Classifier System (LCS) 
Learning classifier systems, or LCS, are a paradigm of rulebased machine learning methods that combine a discovery component (e.g. typically a genetic algorithm) with a learning component (performing either supervised learning, reinforcement learning, or unsupervised learning).[2] Learning classifier systems seek to identify a set of contextdependent rules that collectively store and apply knowledge in a piecewise manner in order to make predictions (e.g. behavior modeling,[3] classification,[4][5] data mining,[5][6][7] regression,[8] function approximation,[9] or game strategy). This approach allows complex solution spaces to be broken up into smaller, simpler parts. The founding concepts behind learning classifier systems came from attempts to model complex adaptive systems, using rulebased agents to form an artificial cognitive system (i.e. artificial intelligence). 
Learning Curve  Plots relating performance to experience are widely used in machine learning. Performance is the error rate or accuracy of the learning system, while experience may be the number of training examples used for learning or the number of iterations used in optimizing the system model parameters. The machine learning curve is useful for many purposes including comparing different algorithms, choosing model parameters during design, adjusting optimization to improve convergence, and determining the amount of data used for training. 
Learning Edge Properties in Graphs from Path Aggregations (LEAP) 
Graph edges, along with their labels, can represent information of fundamental importance, such as links between web pages, friendship between users, the rating given by users to other users or items, and much more. We introduce LEAP, a trainable, general framework for predicting the presence and properties of edges on the basis of the local structure, topology, and labels of the graph. The LEAP framework is based on the exploration and machinelearning aggregation of the paths connecting nodes in a graph. We provide several methods for performing the aggregation phase by training path aggregators, and we demonstrate the flexibility and generality of the framework by applying it to the prediction of links and user ratings in social networks. We validate the LEAP framework on two problems: link prediction, and user rating prediction. On eight large datasets, among which the arXiv collaboration network, the Yeast proteinprotein interaction, and the US airlines routes network, we show that the link prediction performance of LEAP is at least as good as the current state of the art methods, such as SEAL and WLNM. Next, we consider the problem of predicting user ratings on other users: this problem is known as the edgeweight prediction problem in weighted signed networks (WSN). On Bitcoin networks, and Wikipedia RfA, we show that LEAP performs consistently better than the Fairness & Goodness based regression models, varying the amount of training edges between 10 to 90%. These examples demonstrate that LEAP, in spite of its generality, can match or best the performance of approaches that have been especially crafted to solve very specific edge prediction problems. 
Learning From Demonstration (LfD) 

learning from Subgraphs, Embeddings, and Attributes for Link prediction (SEAL) 
Traditional methods for link prediction can be categorized into three main types: graph structure featurebased, latent featurebased, and explicit featurebased. Graph structure feature methods leverage some handcrafted node proximity scores, e.g., common neighbors, to estimate the likelihood of links. Latent feature methods rely on factorizing networks’ matrix representations to learn an embedding for each node. Explicit feature methods train a machine learning model on two nodes’ explicit attributes. Each of the three types of methods has its unique merits. In this paper, we propose SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction), a new framework for link prediction which combines the power of all the three types into a single graph neural network (GNN). GNN is a new type of neural network which directly accepts graphs as input and outputs their labels. In SEAL, the input to the GNN is a local subgraph around each target link. We prove theoretically that our local subgraphs also reserve a great deal of highorder graph structure features related to link existence. Another key feature is that our GNN can naturally incorporate latent features and explicit features. It is achieved by concatenating node embeddings (latent features) and node attributes (explicit features) in the node information matrix for each subgraph, thus combining the three types of features to enhance GNN learning. Through extensive experiments, SEAL shows unprecedentedly strong performance against a wide range of baseline methods, including various link prediction heuristics and network embedding methods. 
Learning Intelligent Distribution Agent (LIDA) 
The LIDA (Learning Intelligent Distribution Agent) cognitive architecture is an integrated artificial cognitive system that attempts to model a broad spectrum of cognition in biological systems, from lowlevel perception/action to highlevel reasoning. Developed primarily by Stan Franklin and colleagues at the University of Memphis, the LIDA architecture is empirically grounded in cognitive science and cognitive neuroscience. In addition to providing hypotheses to guide further research, the architecture can support control structures for software agents and robots. Providing plausible explanations for many cognitive processes, the LIDA conceptual model is also intended as a tool with which to think about how minds work. Two hypotheses underlie the LIDA architecture and its corresponding conceptual model: 1) Much of human cognition functions by means of frequently iterated (~10 Hz) interactions, called cognitive cycles, between conscious contents, the various memory systems and action selection. 2) These cognitive cycles, serve as the ‘atoms’ of cognition of which higherlevel cognitive processes are composed. 
Learning Model Predictive Controller (LMPC) 
We present a samplebased Learning Model Predictive Controller (LMPC) for constrained uncertain linear systems subject to bounded additive disturbances. The proposed controller builds on earlier work on LMPC for deterministic systems. First, we introduce the design of the safe set and value function used to guarantee safety and performance improvement. Afterwards, we show how these quantities can be approximated using noisy historical data. The effectiveness of the proposed approach is demonstrated on a numerical example. We show that the proposed LMPC is able to safely explore the state space and to iteratively improve the worstcase closedloop performance, while robustly satisfying state and input constraints. 
Learning MWay Tree (LMWTree) 
LMWtree is a generic template library written in C++ that implements several algorithms that use the mway nearest neighbor tree structure to store their data. See the related PhD thesis for more details on mway nn trees. The algorithms and data structures are generic to support different data representations such as dense real valued and bit vectors, and sparse vectors. Additionally, it can index any object type that can form a prototype representation of a set of objects. The algorithms are primarily focussed on computationally efficient clustering. Clustering is an unsupervised machine learning process that finds interesting patterns in data. It places similar items into clusters and dissimilar items into different clusters. The data structures and algorithms can also be used for nearest neighbor search, supervised learning and other machine learning applications. The package includes EMtree, Ktree, kmeans, TSVQ, repeated kmeans, clustering, random projections, random indexing, hashing, bit signatures. See the related PhD thesis for more details these algorithms and representations. LMWtree is licensed under the BSD license. 
Learning Solving Procedure  It is expected that progress toward true artificial intelligence will be achieved through the emergence of a system that integrates representation learning and complex reasoning (LeCun et al. 2015). In response to this prediction, research has been conducted on implementing the symbolic reasoning of a von Neumann computer in an artificial neural network (Graves et al. 2016; Graves et al. 2014; Reed et al. 2015). However, these studies have many limitations in realizing neuralsymbolic integration (Jaeger. 2016). Here, we present a new learning paradigm: a learning solving procedure (LSP) that learns the procedure for solving complex problems. This is not accomplished merely by learning inputoutput data, but by learning algorithms through a solving procedure that obtains the output as a sequence of tasks for a given input problem. The LSP neural network system not only learns simple problems of addition and multiplication, but also the algorithms of complicated problems, such as complex arithmetic expression, sorting, and Hanoi Tower. To realize this, the LSP neural network structure consists of a deep neural network and long shortterm memory, which are recursively combined. Through experimentation, we demonstrate the efficiency and scalability of LSP and its validity as a mechanism of complex reasoning. 
Learning Through Deterministic Assignment of Hidden Parameters (LtDaHP) 
Supervised learning frequently boils down to determining hidden and bright parameters in a parameterized hypothesis space based on finite inputoutput samples. The hidden parameters determine the attributions of hidden predictors or the nonlinear mechanism of an estimator, while the bright parameters characterize how hidden predictors are linearly combined or the linear mechanism. In traditional learning paradigm, hidden and bright parameters are not distinguished and trained simultaneously in one learning process. Such an onestage learning (OSL) brings a benefit of theoretical analysis but suffers from the high computational burden. To overcome this difficulty, a twostage learning (TSL) scheme, featured by learning through deterministic assignment of hidden parameters (LtDaHP) was proposed, which suggests to deterministically generate the hidden parameters by using minimal Riesz energy points on a sphere and equally spaced points in an interval. We theoretically show that with such deterministic assignment of hidden parameters, LtDaHP with a neural network realization almost shares the same generalization performance with that of OSL. We also present a series of simulations and application examples to support the outperformance of LtDaHP 
Learning through Probing (LTP) 
Multiagent reinforcement learning has received significant interest in recent years notably due to the advancements made in deep reinforcement learning which have allowed for the developments of new architectures and learning algorithms. Using social dilemmas as the training ground, we present a novel learning architecture, Learning through Probing (LTP), where agents utilize a probing mechanism to incorporate how their opponent’s behavior changes when an agent takes an action. We use distinct training phases and adjust rewards according to the overall outcome of the experiences accounting for changes to the opponents behavior. We introduce a parameter eta to determine the significance of these future changes to opponent behavior. When applied to the Iterated Prisoner’s Dilemma (IPD), LTP agents demonstrate that they can learn to cooperate with each other, achieving higher average cumulative rewards than other reinforcement learning methods while also maintaining good performance in playing against static agents that are present in Axelrod tournaments. We compare this method with traditional reinforcement learning algorithms and agenttracking techniques to highlight key differences and potential applications. We also draw attention to the differences between solving games and societallike interactions and analyze the training of Qlearning agents in makeshift societies. This is to emphasize how cooperation may emerge in societies and demonstrate this using environments where interactions with opponents are determined through a random encounter format of the IPD. 
Learning to Coordinate and Teach Reinforcement (LeCTR) 
We present a framework and algorithm for peertopeer teaching in cooperative multiagent reinforcement learning. Our algorithm, Learning to Coordinate and Teach Reinforcement (LeCTR), trains advising policies by using students’ learning progress as a teaching reward. Agents using LeCTR learn to assume the role of a teacher or student at the appropriate moments, exchanging action advice to accelerate the entire learning process. Our algorithm supports teaching heterogeneous teammates, advising under communication constraints, and learns both what and when to advise. LeCTR is demonstrated to outperform the final performance and rate of learning of prior teaching methods on multiple benchmark domains. To our knowledge, this is the first approach for learning to teach in a multiagent setting. 
Learning to Multitask (L2MT) 
Multitask learning has shown promising performance in many applications and many multitask models have been proposed. In order to identify an effective multitask model for a given multitask problem, we propose a learning framework called learning to multitask (L2MT). To achieve the goal, L2MT exploits historical multitask experience which is organized as a training set consists of several tuples, each of which contains a multitask problem with multiple tasks, a multitask model, and the relative test error. Based on such training set, L2MT first uses a proposed layerwise graph neural network to learn task embeddings for all the tasks in a multitask problem and then learns an estimation function to estimate the relative test error based on task embeddings and the representation of the multitask model based on a unified formulation. Given a new multitask problem, the estimation function is used to identify a suitable multitask model. Experiments on benchmark datasets show the effectiveness of the proposed L2MT framework. 
Learning to Recommend with Missing Modalities (LRMM) 
Multimodal learning has shown promising performance in contentbased recommendation due to the auxiliary user and item information of multiple modalities such as text and images. However, the problem of incomplete and missing modality is rarely explored and most existing methods fail in learning a recommendation model with missing or corrupted modalities. In this paper, we propose LRMM, a novel framework that mitigates not only the problem of missing modalities but also more generally the coldstart problem of recommender systems. We propose modality dropout (mdrop) and a multimodal sequential autoencoder (mauto) to learn multimodal representations for complementing and imputing missing modalities. Extensive experiments on realworld Amazon data show that LRMM achieves stateoftheart performance on rating prediction tasks. More importantly, LRMM is more robust to previous methods in alleviating datasparsity and the coldstart problem. 
Learning to Teach  Teaching plays a very important role in our society, by spreading human knowledge and educating our next generations. A good teacher will select appropriate teaching materials, impact suitable methodologies, and set up targeted examinations, according to the learning behaviors of the students. In the field of artificial intelligence, however, one has not fully explored the role of teaching, and pays most attention to machine \emph{learning}. In this paper, we argue that equal attention, if not more, should be paid to teaching, and furthermore, an optimization framework (instead of heuristics) should be used to obtain good teaching strategies. We call this approach `learning to teach’. In the approach, two intelligent agents interact with each other: a student model (which corresponds to the learner in traditional machine learning algorithms), and a teacher model (which determines the appropriate data, loss function, and hypothesis space to facilitate the training of the student model). The teacher model leverages the feedback from the student model to optimize its own teaching strategies by means of reinforcement learning, so as to achieve teacherstudent coevolution. To demonstrate the practical value of our proposed approach, we take the training of deep neural networks (DNN) as an example, and show that by using the learning to teach techniques, we are able to use much less training data and fewer iterations to achieve almost the same accuracy for different kinds of DNN models (e.g., multilayer perceptron, convolutional neural networks and recurrent neural networks) under various machine learning tasks (e.g., image classification and text understanding). 
Learning to Weight (LTW) 
In information retrieval (IR) and related tasks, term weighting approaches typically consider the frequency of the term in the document and in the collection in order to compute a score reflecting the importance of the term for the document. In tasks characterized by the presence of training data (such as text classification) it seems logical that the term weighting function should take into account the distribution (as estimated from training data) of the term across the classes of interest. Although `supervised term weighting’ approaches that use this intuition have been described before, they have failed to show consistent improvements. In this article we analyse the possible reasons for this failure, and call consolidated assumptions into question. Following this criticism we propose a novel supervised term weighting approach that, instead of relying on any predefined formula, learns a term weighting function optimised on the training set of interest; we dub this approach \emph{Learning to Weight} (LTW). The experiments that we run on several wellknown benchmarks, and using different learning methods, show that our method outperforms previous term weighting approaches in text classification. 
Learning Under Privileged Information (LUPI) 
Conformal Prediction in Learning Under Privileged Information Paradigm with Applications in Drug Discovery 
Learning Using Privileged Information (LUPI) 
We adopt a multiview approach for analyzing two knowledge transfer settings—learning using privileged information (LUPI) and distillation—in a common framework. Under reasonable assumptions about the complexities of hypothesis spaces, and being optimistic about the expected loss achievable by the student (in distillation) and a transformed teacher predictor (in LUPI), we show that encouraging agreement between the teacher and the student leads to reduced search space. As a result, improved convergence rate can be obtained with regularized empirical risk minimization. 
Learning Using Statistical Invariants (LUSI) 
This paper introduces a new learning paradigm, called Learning Using Statistical Invariants (LUSI), which is different from the classical one. In a classical paradigm, the learning machine constructs a classification rule that minimizes the probability of expected error; it is datadriven model of learning. In the LUSI paradigm, in order to construct the desired classification function, a learning machine computes statistical invariants that are specific for the problem, and then minimizes the expected error in a way that preserves these invariants; it is thus both data and invariantdriven learning. From a mathematical point of view, methods of the classical paradigm employ mechanisms of strong convergence of approximations to the desired function, whereas methods of the new paradigm employ both strong and weak convergence mechanisms. This can significantly increase the rate of convergence. 
Learning Vector Quantization (LVQ) 
In computer science, learning vector quantization (LVQ), is a prototypebased supervised classification algorithm. LVQ is the supervised counterpart of vector quantization systems. 
Learning with Counts  
Learning with OpponentLearning Awareness (LOLA) 
Multiagent settings are quickly gathering importance in machine learning. Beyond a plethora of recent work on deep multiagent reinforcement learning, hierarchical reinforcement learning, generative adversarial networks and decentralized optimization can all be seen as instances of this setting. However, the presence of multiple learning agents in these settings renders the training problem nonstationary and often leads to unstable training or undesired final results. We present Learning with OpponentLearning Awareness (LOLA), a method that reasons about the anticipated learning of the other agents. The LOLA learning rule includes an additional term that accounts for the impact of the agent’s policy on the anticipated parameter update of the other agents. We show that the LOLA update rule can be efficiently calculated using an extension of the likelihood ratio policy gradient update, making the method suitable for modelfree reinforcement learning. This method thus scales to large parameter and input spaces and nonlinear function approximators. Preliminary results show that the encounter of two LOLA agents leads to the emergence of titfortat and therefore cooperation in the infinitely iterated prisoners’ dilemma, while independent learning does not. In this domain, LOLA also receives higher payouts compared to a naive learner, and is robust against exploitation by higher order gradientbased methods. Applied to infinitely repeated matching pennies, only LOLA agents converge to the Nash equilibrium. We also apply LOLA to a grid world task with an embedded social dilemma using deep recurrent policies. Again, by considering the learning of the other agent, LOLA agents learn to cooperate out of selfish interests. 
Learning without Memorizing (LwM) 
Incremental learning (IL) is an important task aimed to increase the capability of a trained model, in terms of the number of classes recognizable by the model. The key problem in this task is the requirement of storing data (e.g. images) associated with existing classes, while training the classifier to learn new classes. However, this is impractical as it increases the memory requirement at every incremental step, which makes it impossible to implement IL algorithms on the edge devices with limited memory. Hence, we propose a novel approach, called ‘Learning without Memorizing (LwM)’, to preserve the information with respect to existing (base) classes, without storing any of their data, while making the classifier progressively learn the new classes. In LwM, we present an information preserving penalty: Attention Distillation Loss, and demonstrate that penalizing the changes in classifiers’ attention maps helps to retain information of the base classes, as new classes are added. We show that adding Attention Distillation Loss to the distillation loss which is an existing information preserving loss consistently outperforms the stateoftheart performance in the iILSVRCsmall and iCIFAR100 datasets in terms of the overall accuracy of base and incrementally learned classes. 
LearningBased Visual Saliency Fusion Method for HDR Content (LVBSHDR) 
Saliency prediction for Standard Dynamic Range (SDR) videos has been well explored in the last decade. However, limited studies are available on High Dynamic Range (HDR) Visual Attention Models (VAMs). Considering that the characteristic of HDR content in terms of dynamic range and color gamut is quite different than those of SDR content, it is essential to identify the importance of different saliency attributes of HDR videos for designing a VAM and understand how to combine these features. To this end we propose a learningbased visual saliency fusion method for HDR content (LVBSHDR) to combine various visual saliency features. In our approach various conspicuity maps are extracted from HDR data, and then for fusing conspicuity maps, a Random Forests algorithm is used to train a model based on the collected data from an eyetracking experiment. Performance evaluations demonstrate the superiority of the proposed fusion method against other existing fusion methods. 
LearningtoRank (LETOR) 
As one of the most popular techniques for solving the ranking problem in information retrieval, Learningtorank (LETOR) has received a lot of attention both in academia and industry due to its importance in a wide variety of data mining applications. However, most of existing LETOR approaches choose to learn a single global ranking function to handle all queries, and ignore the substantial differences that exist between queries. In this paper, we propose a domain generalization strategy to tackle this problem. We propose QueryInvariant Listwise Context Modeling (QILCM), a novel neural architecture which eliminates the detrimental influence of interquery variability by learning \textit{queryinvariant} latent representations, such that the ranking system could generalize better to unseen queries. We evaluate our techniques on benchmark datasets, demonstrating that QILCM outperforms previous stateoftheart approaches by a substantial margin. 
Least Absolute Deviation (LAD) 
Least absolute deviations (LAD), also known as least absolute errors (LAE), least absolute value (LAV), least absolute residual (LAR), sum of absolute deviations, or the L1 norm condition, is a statistical optimality criterion and the statistical optimization technique that relies on it. Similar to the popular least squares technique, it attempts to find a function which closely approximates a set of data. In the simple case of a set of (x,y) data, the approximation function is a simple ‘trend line’ in twodimensional Cartesian coordinates. The method minimizes the sum of absolute errors (SAE) (the sum of the absolute values of the vertical ‘residuals’ between points generated by the function and corresponding points in the data). The least absolute deviations estimate also arises as the maximum likelihood estimate if the errors have a Laplace distribution. LadR 
Least Absolute Deviations (LAD) 
Least absolute deviations (LAD), also known as Least Absolute Errors (LAE), Least Absolute Value (LAV), or Least Absolute Residual (LAR) or the L1 norm problem, is a statistical optimization technique similar to the popular least squares technique that attempts to find a function which closely approximates a set of data. In the simple case of a set of (x,y) data, the approximation function is a simple ‘trend line’ in twodimensional Cartesian coordinates. The method minimizes the sum of absolute errors (SAE) (the sum of the absolute values of the vertical ‘residuals’ between points generated by the function and corresponding points in the data). The least absolute deviations estimate also arises as the maximum likelihood estimate if the errors have a Laplace distribution. 
Least Absolute Deviations Estimator (LADE) 
This paper provides an entire inference procedure for the autoregressive model under (conditional) heteroscedasticity of unknown form with a finite variance. We first establish the asymptotic normality of the weighted least absolute deviations estimator (LADE) for the model. Second, we develop the random weighting (RW) method to estimate its asymptotic covariance matrix, leading to the implementation of the Wald test. Third, we construct a portmanteau test for model checking, and use the RW method to obtain its critical values. As a special weighted LADE, the feasible adaptive LADE (ALADE) is proposed and proved to have the same efficiency as its infeasible counterpart. The importance of our entire methodology based on the feasible ALADE is illustrated by simulation results and the real data analysis on three U.S. economic data sets. 
Least Absolute Shrinkage and Screening Operator (LASSO) 
Slide 31: ‘Tibshirani (1996): LASSO = Least Absolute Shrinkage and Selection Operator new translation: LASSO = Least Absolute Shrinkage and Screening Operator’ 
Least Absolute Shrinkage and Selection Operator (LASSO) 
The Lasso is a shrinkage and selection method for linear regression. It minimizes the usual sum of squared errors, with a bound on the sum of the absolute values of the coefficients. It has connections to softthresholding of wavelet coefficients, forward stagewise regression, and boosting methods. 
Least Square Projection (LSP) 
The problem of projecting multidimensional data into lower dimensions has been pursued by many researchers due to its potential application to data analysis of various kinds. This paper presents a novel multidimensional projection technique based on least square approximations. The approximations compute the coordinates of a set of projected points based on the coordinates of a reduced number of control points with defined geometry. We name the technique Least Square Projections (LSP). 
Least Squares AutoTuning  Least squares is by far the simplest and most commonly applied computational method in many fields. In almost all applications, the least squares objective is rarely the true objective. We account for this discrepancy by parametrizing the least squares problem and automatically adjusting these parameters using an optimization algorithm. We apply our method, which we call least squares autotuning, to data fitting. 
Least Squares Deep QNetwork (LSDQN) 
Deep reinforcement learning (DRL) methods such as the Deep QNetwork (DQN) have achieved stateoftheart results in a variety of challenging, highdimensional domains. This success is mainly attributed to the power of deep neural networks to learn rich domain representations for approximating the value function or policy. Batch reinforcement learning methods with linear representations, on the other hand, are more stable and require less hyper parameter tuning. Yet, substantial feature engineering is necessary to achieve good results. In this work we propose a hybrid approach — the Least Squares Deep QNetwork (LSDQN), which combines rich feature representations learned by a DRL algorithm with the stability of a linear least squares method. We do this by periodically retraining the last hidden layer of a DRL network with a batch least squares update. Key to our approach is a Bayesian regularization term for the least squares update, which prevents overfitting to the more recent data. We tested LSDQN on five Atari games and demonstrate significant improvement over vanilla DQN and DoubleDQN. We also investigated the reasons for the superior performance of our method. Interestingly, we found that the performance improvement can be attributed to the large batch size used by the LS method when optimizing the last layer. 
Least Squares Metric Learning (LSML) 
Relational Constraints for Metric Learning on Relational Data 
LeastAngle Regression (LARS) 
In statistics, leastangle regression (LARS) is a regression algorithm for highdimensional data, developed by Bradley Efron, Trevor Hastie, Iain Johnstone and Robert Tibshirani. Suppose we expect a response variable to be determined by a linear combination of a subset of potential covariates. Then the LARS algorithm provides a means of producing an estimate of which variables to include, as well as their coefficients. Instead of giving a vector result, the LARS solution consists of a curve denoting the solution for each value of the L1 norm of the parameter vector. The algorithm is similar to forward stepwise regression, but instead of including variables at each step, the estimated parameters are increased in a direction equiangular to each one’s correlations with the residual. 
LeaveOneOut  In this paper, we introduce a powerful technique, LeaveOneOut, to the analysis of lowrank matrix completion problems. Using this technique, we develop a general approach for obtaining finegrained, entrywise bounds on iterative stochastic procedures. We demonstrate the power of this approach in analyzing two of the most important algorithms for matrix completion: the nonconvex approach based on Singular Value Projection (SVP), and the convex relaxation approach based on nuclear norm minimization (NNM). In particular, we prove for the first time that the original form of SVP, without resampling or sample splitting, converges linearly in the infinity norm. We further apply our leaveoneout approach to an iterative procedure that arises in the analysis of the dual solutions of NNM. Our results show that NNM recovers the true $ d $by$ d $ rank$ r $ matrix with $\mathcal{O}(\mu^2 r^3d \log d )$ observed entries, which has optimal dependence on the dimension and is independent of the condition number of the matrix. To the best of our knowledge, this is the first sample complexity result for a tractable matrix completion algorithm that satisfies these two properties simultaneously. 
LeaveOneOut Cross Validation (LOOCV) 
Leaveoneout crossvalidation (LOOCV) is a particular case of leavepout crossvalidation with p = 1. loo 
LeaveOneOut Least Square Monte Carlo (LOOLSM) 
The least square Monte Carlo (LSM) algorithm proposed by Longstaff and Schwartz [2001] is the most widely used method for pricing options with early exercise features. The LSM estimator contains lookahead bias, and the conventional technique of removing it necessitates an independent set of simulations. This study proposes a new approach for efficiently eliminating lookahead bias by using the leaveoneout method, a wellknown crossvalidation technique for machine learning applications. The leaveoneout LSM (LOOLSM) method is illustrated with examples, including multiasset options whose LSM price is biased high. The asymptotic behavior of lookahead bias is also discussed with the LOOLSM approach. 
LeavepOut Cross Validation (LpOCV) 
As the name suggests, leavepout crossvalidation (LpO CV) involves using p observations as the validation set and the remaining observations as the training set. This is repeated on all ways to cut the original sample on a validation set of p’ observations and a training set. LpO crossvalidation requires to learn and validate times (where n is the number of observation in the original sample). So as soon as n is quite big it becomes impossible to calculate. 
Lecture Hall Tableaux  We introduce lecture hall tableaux, which are fillings of a skew Young diagram satisfying certain conditions. Lecture hall tableaux generalize both lecture hall partitions and antilecture hall compositions, and also contain reverse semistandard Young tableaux as a limit case. We show that the coefficients in the Schur expansion of multivariate little $q$Jacobi polynomials are generating functions for lecture hall tableaux. Using a Selbergtype integral we show that moments of multivariate little $q$Jacobi polynomials, which are equal to generating functions for lecture hall tableaux of a Young diagram, have a product formula. We also explore various combinatorial properties of lecture hall tableaux. 
LectureBank  Recent years have witnessed the rising popularity of Natural Language Processing (NLP) and related fields such as Artificial Intelligence (AI) and Machine Learning (ML). Many online courses and resources are available even for those without a strong background in the field. Often the student is curious about a specific topic but does not quite know where to begin studying. To answer the question of ‘what should one learn first,’ we apply an embeddingbased method to learn prerequisite relations for course concepts in the domain of NLP. We introduce LectureBank, a dataset containing 1,352 English lecture files collected from university courses which are each classified according to an existing taxonomy as well as 208 manuallylabeled prerequisite relation topics, which is publicly available. The dataset will be useful for educational purposes such as lecture preparation and organization as well as applications such as reading list generation. Additionally, we experiment with neural graphbased networks and nonneural classifiers to learn these prerequisite relations from our dataset. 
LeFlow  Recent work has shown that FieldProgrammable Gate Arrays (FPGAs) play an important role in the acceleration of Machine Learning applications. Initial specification of machine learning applications are often done using a highlevel Pythonoriented framework such as Tensorflow, followed by a manual translation to either C or RTL for synthesis using vendor tools. This manual translation step is timeconsuming and requires expertise that limit the applicability of FPGAs in this important domain. In this paper, we present an opensource toolflow that maps numerical computation models written in Tensorflow to synthesizable hardware. Unlike other tools, which are often constrained by a small number of inflexible templates, our flow uses Google’s XLA compiler which emits LLVM code directly from a Tensorflow specification. This LLVM code can then be used with a highlevel synthesis tool to automatically generate hardware. We show that our flow allows users to generate Deep Neural Networks with very few lines of Python code. 
LemmaTag  We present LemmaTag, a featureless recurrent neural network architecture that jointly generates partofspeech tags and lemmatizes sentences of languages with complex morphology, using bidirectional RNNs with characterlevel and wordlevel embeddings. We demonstrate that both tasks benefit from sharing the encoding part of the network and from using the tagger output as an input to the lemmatizer. We evaluate our model across several morphologicallyrich languages, surpassing stateoftheart accuracy in both partofspeech tagging and lemmatization in Czech, German, and Arabic. 
Lemmatization  Lemmatisation (or lemmatization) in linguistics is the process of grouping together the different inflected forms of a word so they can be analysed as a single item. In computational linguistics, lemmatisation is the algorithmic process of determining the lemma for a given word. Since the process may involve complex tasks such as understanding context and determining the part of speech of a word in a sentence (requiring, for example, knowledge of the grammar of a language) it can be a hard task to implement a lemmatiser for a new language. In many languages, words appear in several inflected forms. For example, in English, the verb ‘to walk’ may appear as ‘walk’, ‘walked’, ‘walks’, ‘walking’. The base form, ‘walk’, that one might look up in a dictionary, is called the lemma for the word. The combination of the base form with the part of speech is often called the lexeme of the word. Lemmatisation is closely related to stemming. The difference is that a stemmer operates on a single word without knowledge of the context, and therefore cannot discriminate between words which have different meanings depending on part of speech. However, stemmers are typically easier to implement and run faster, and the reduced accuracy may not matter for some applications. 
Lemotif  We present Lemotif. Lemotif generates a motif for your emotional life. You tell Lemotif a little bit about your day — what were salient events or aspects and how they made you feel. Lemotif will generate a lemotif — a creative abstract visual depiction of your emotions and their sources. Over time, Lemotif can create visual motifs to capture a summary of your emotional states over arbitrary periods of time — making patterns in your emotions and their sources apparent, presenting opportunities to take actions, and measure their effectiveness. The underlying principles in Lemotif are that the lemotif should (1) separate out the sources of the emotions, (2) depict these sources visually, (3) depict the emotions visually, and (4) have a creative aspect to them. We verify via human studies that each of these factors contributes to the proposed lemotifs being favored over corresponding baselines. 
LempelāZivāOberhumer (LZO) 
LempelZivOberhumer (LZO) is a lossless data compression algorithm that is focused on decompression speed. 
Lenia  We report a new model of artificial life called Lenia (from Latin lenis ‘smooth’), a twodimensional cellular automaton with continuous spacetimestate and generalized local rule. Computer simulations show that Lenia supports a great diversity of complex autonomous patterns or ‘lifeforms’ bearing resemblance to realworld microscopic organisms. More than 400 species in 18 families have been identified, many discovered via interactive evolutionary computation. We present basic observations of the model regarding the properties of spacetime and basic settings. We provide a board survey of the lifeforms, categorize them into a hierarchical taxonomy, and map their distribution in the parameter hyperspace. We describe their morphological structures and behavioral dynamics, propose possible mechanisms of their selfpropulsion, selforganization and plasticity. Finally, we discuss how the study of Lenia would be related to biology, artificial life, and artificial intelligence. 
Lenstra Lenstra Lovasz (LLL) 
The LenstraLenstraLovĆ”sz (LLL) lattice basis reduction algorithm is a polynomial time lattice reduction algorithm invented by Arjen Lenstra, Hendrik Lenstra and LĆ”szlĆ³ LovĆ”sz in 1982. 
Level of Measurement  Level of measurement or scale of measure is a classification that describes the nature of information within the values assigned to variables. Psychologist Stanley Smith Stevens developed the bestknown classification with four levels, or scales, of measurement: nominal, ordinal, interval, and ratio. This framework of distinguishing levels of measurement originated in psychology and is widely criticized by scholars in other disciplines. Other classifications include those by Mosteller and Tukey, and by Chrisman. 
Level Weighted Structural Similarity (LWSSIM) 
The Mean Square Error (MSE) has shown its strength when applied in deep generative models such as AutoEncoders to model reconstruction loss. However, in image domain especially, the limitation of MSE is obvious: it assumes pixel independence and ignores spatial relationships of samples. This contradicts most architectures of AutoEncoders which use convolutional layers to extract spatial dependent features. We base on the structural similarity metric (SSIM) and propose a novel level weighted structural similarity (LWSSIM) loss for convolutional AutoEncoders. Experiments on common datasets on various AutoEncoder variants show that our loss is able to outperform the MSE loss and the Vanilla SSIM loss. We also provide reasons why our model is able to succeed in cases where the standard SSIM loss fails. 
LevenbergMarquardt Algorithm (LMA) 
In mathematics and computing, the LevenbergMarquardt algorithm (LMA), also known as the damped leastsquares (DLS) method, is used to solve nonlinear least squares problems. These minimization problems arise especially in least squares curve fitting. The LMA interpolates between the GaussNewton algorithm (GNA) and the method of gradient descent. The LMA is more robust than the GNA, which means that in many cases it finds a solution even if it starts very far off the final minimum. For wellbehaved functions and reasonable starting parameters, the LMA tends to be a bit slower than the GNA. LMA can also be viewed as GaussNewton using a trust region approach. The LMA is a very popular curvefitting algorithm used in many software applications for solving generic curvefitting problems. However, as for many fitting algorithms, the LMA finds only a local minimum, which is not necessarily the global minimum. onls 
Levenshtein Distance  In information theory and computer science, the Levenshtein distance is a string metric for measuring the difference between two sequences. Informally, the Levenshtein distance between two words is the minimum number of singlecharacter edits (i.e. insertions, deletions or substitutions) required to change one word into the other. The phrase edit distance is often used to refer specifically to Levenshtein distance. It is named after Vladimir Levenshtein, who considered this distance in 1965. It is closely related to pairwise string alignments. 
LevyAttack  Developing techniques for adversarial attack and defense is an important research field for establishing reliable machine learning and its applications. Many existing methods employ Gaussian random variables for exploring the data space to find the most adversarial (for attacking) or least adversarial (for defense) point. However, the Gaussian distribution is not necessarily the optimal choice when the exploration is required to follow the complicated structure that most realworld data distributions exhibit. In this paper, we investigate how statistics of random variables affect such random walk exploration. Specifically, we generalize the Boundary Attack, a stateoftheart blackbox decision based attacking strategy, and propose the L\’evyAttack, where the random walk is driven by symmetric $\alpha$stable random variables. Our experiments on MNIST and CIFAR10 datasets show that the L\’evyAttack explores the image data space more efficiently, and significantly improves the performance. Our results also give an insight into the recently found fact in the whitebox attacking scenario that the choice of the norm for measuring the amplitude of the adversarial patterns is essential. 
Lexical Dispersion Plot  A Lexical Dispersion Plot shows the position of words in a given text. On the y axis there is the list of words to be looked at and on the x axis there is the position in the text. Therefore the highest value on the x axis is the lenght of the text. qdap 
Lexical Table  
Lexicase Selection  Lexicase selection is a parent selection method that considers test cases separately, rather than in aggregate, when performing parent selection. It performs well in discrete error spaces but not on the continuousvalued problems that compose most system identification tasks. In this paper, we develop a new form of lexicase selection for symbolic regression, named epsilonlexicase selection, that redefines the pass condition for individuals on each test case in a more effective way. We run a series of experiments on realworld and synthetic problems with several treatments of epsilon and quantify how epsilon affects parent selection and model performance. epsilonlexicase selection is shown to be effective for regression, producing better fit models compared to other techniques such as tournament selection and agefitness Pareto optimization. We demonstrate that epsilon can be adapted automatically for individual test cases based on the population performance distribution. Our experiments show that epsilonlexicase selection with automatic epsilon produces the most accurate models across tested problems with negligible computational overhead. We show that behavioral diversity is exceptionally high in lexicase selection treatments, and that epsilonlexicase selection makes use of more fitness cases when selecting parents than lexicase selection, which helps explain the performance improvement. Batch Tournament Selection for Genetic Programming 
Lexis Surface Map  
LexNLP  LexNLP is an open source Python package focused on natural language processing and machine learning for legal and regulatory text. The package includes functionality to (i) segment documents, (ii) identify key text such as titles and section headings, (iii) extract over eighteen types of structured information like distances and dates, (iv) extract named entities such as companies and geopolitical entities, (v) transform text into features for model training, and (vi) build unsupervised and supervised models such as word embedding or tagging models. LexNLP includes pretrained models based on thousands of unit tests drawn from real documents available from the SEC EDGAR database as well as various judicial and regulatory proceedings. LexNLP is designed for use in both academic research and industrial applications, and is distributed at https://…/lexpredictlexnlp. 
LexVec  In this paper, we propose LexVec, a new method for generating distributed word representations that uses lowrank, weighted factorization of the Positive Pointwise Mutual Information matrix via stochastic gradient descent, employing a weighting scheme that assigns heavier penalties for errors on frequent cooccurrences while still accounting for negative cooccurrence. Evaluation on word similarity and analogy tasks shows that LexVec matches and often outperforms stateoftheart methods on many of these tasks. 
LGMNet  In this work, we propose a novel metalearning approach for fewshot classification, which learns transferable prior knowledge across tasks and directly produces network parameters for similar unseen tasks with training samples. Our approach, called LGMNet, includes two key modules, namely, TargetNet and MetaNet. The TargetNet module is a neural network for solving a specific task and the MetaNet module aims at learning to generate functional weights for TargetNet by observing training samples. We also present an intertask normalization strategy for the training process to leverage common information shared across different tasks. The experimental results on Omniglot and miniImageNet datasets demonstrate that LGMNet can effectively adapt to similar unseen tasks and achieve competitive performance, and the results on synthetic datasets show that transferable prior knowledge is learned by the MetaNet module via mapping training data to functional weights. LGMNet enables fast learning and adaptation since no further tuning steps are required compared to other metalearning approaches. 
libDirectional  In this paper, we present libDirectional, a MATLAB library for directional statistics and directional estimation. It supports a variety of commonly used distributions on the unit circle, such as the von Mises, wrapped normal, and wrapped Cauchy distributions. Furthermore, various distributions on higherdimensional manifolds such as the unit hypersphere and the hypertorus are available. Based on these distributions, several recursive filtering algorithms in libDirectional allow estimation on these manifolds. The functionality is implemented in a clear, welldocumented, and objectoriented structure that is both easy to use and easy to extend. 
libFM  Factorization machines (FM) are a generic approach that allows to mimic most factorization models by feature engineering. This way, factorization machines combine the generality of feature engineering with the superiority of factorization models in estimating interactions between categorical variables of large domain. libFM is a software implementation for factorization machines that features stochastic gradient descent (SGD) and alternating least squares (ALS) optimization as well as Bayesian inference using Markov Chain Monte Carlo (MCMC). 
LibLinear  LibLinear is a linear classifier for data with millions of instances and features. It supports Ā· L2regularized classifiers L2loss linear SVM, Ā· L1loss linear SVM, and logistic regression (LR) Ā· L1regularized classifiers (after version 1.4) Ā· L2loss linear SVM and logistic regression (LR) Ā· L2regularized support vector regression (after version 1.9) Ā· L2loss linear SVR and L1loss linear SVR. 
Library for Online Learning (LIBOL) 
LIBOL is an opensource library for largescale online learning, which consists of a large family of e cient and scalable stateoftheart online learning algorithms for largescale online classification tasks. We have offered easytouse commandline tools and examples for users and developers, and also have made comprehensive documents available for both beginners and advanced users. LIBOL is not only a machine learning toolbox, but also a comprehensive experimental platform for conducting online learning research. http://…/LIBOL_manual.pdf. http://libol.stevenhoi.org 
LIBS2ML  LIBS2ML is a library based on scalable second order learning algorithms for solving largescale problems, i.e., big data problems in machine learning. LIBS2ML has been developed using MEX files, i.e., C++ with MATLAB/Octave interface to take the advantage of both the worlds, i.e., faster learning using C++ and easy I/O using MATLAB. Most of the available libraries are either in MATLAB/Python/R which are very slow and not suitable for largescale learning, or are in C/C++ which does not have easy ways to take input and display results. So LIBS2ML is completely unique due to its focus on the scalable second order methods, the hot research topic, and being based on MEX files. Thus it provides researchers a comprehensive environment to evaluate their ideas and it also provides machine learning practitioners an effective tool to deal with the largescale learning problems. LIBS2ML is an opensource, highly efficient, extensible, scalable, readable, portable and easy to use library. The library can be downloaded from the URL: \url{https://…/LIBS2ML}. 
LIDIOMS  In this paper, we describe the LIDIOMS data set, a multilingual RDF representation of idioms currently containing five languages: English, German, Italian, Portuguese, and Russian. The data set is intended to support natural language processing applications by providing links between idioms across languages. The underlying data was crawled and integrated from various sources. To ensure the quality of the crawled data, all idioms were evaluated by at least two native speakers. Herein, we present the model devised for structuring the data. We also provide the details of linking LIDIOMS to wellknown multilingual data sets such as BabelNet. The resulting data set complies with best practices according to Linguistic Linked Open Data Community. 
Lie Group Machine Learning (LML) 
In this paper a new method for dimensionality reduction in machine learning is pr oposed and called as Lie group Machine Learning (LML). The theory framework of LML is given , including the conception of oneparameter subgroup, Lie algebra and LML; the geometric properties of LML; the generalization hypothesis axiom, the partition independence hypothesis axiom, the duality hypothesis axiom, the learning compatibility hypothesis axiom of LML and the classifiers’ design of LML. The theory framework of Lie group machine learning (LML). Survey on Lie Group Machine Learning 
Lifelong Federated Reinforcement Learning (LFRLA) 
This paper was motivated by the problem of how to make robots fuse and transfer their experience so that they can effectively use prior knowledge and quickly adapt to new environments. To address the problem, we present a learning architecture for navigation in cloud robotic systems: Lifelong Federated Reinforcement Learning (LFRLA). In the work, We propose a knowledge fusion algorithm for upgrading a shared model deployed on the cloud. Then, effective transfer learning methods in LFRLA are introduced. LFRLA is consistent with human cognitive science and fits well in cloud robotic systems. Experiments show that LFRLA greatly improves the efficiency of reinforcement learning for robot navigation. The cloud robotic system deployment also shows that LFRLA is capable of fusing prior knowledge. In addition, we release a cloud robotic navigationlearning website based on LFRLA. 
Lifelong Learning (LL) 
This paper proposes a novel lifelong learning (LL) approach to sentiment classification. LL mimics the human continuous learning process, i.e., retaining the knowledge learned from past tasks and use it to help future learning. In this paper, we first discuss LL in general and then LL for sentiment classification in particular. The proposed LL approach adopts a Bayesian optimization framework based on stochastic gradient descent. Our experimental results show that the proposed method outperforms baseline methods significantly, which demonstrates that lifelong learning is a promising research direction. 
LIFT  Reinforcement learning approaches have long appealed to the data management community due to their ability to learn to control dynamic behavior from raw system performance. Recent successes in combining deep neural networks with reinforcement learning have sparked significant new interest in this domain. However, practical solutions remain elusive due to large training data requirements, algorithmic instability, and lack of standard tools. In this work, we introduce LIFT, an endtoend software stack for applying deep reinforcement learning to data management tasks. While prior work has frequently explored applications in simulations, LIFT centers on utilizing human expertise to learn from demonstrations, thus lowering online training times. We further introduce TensorForce, a TensorFlow library for applied deep reinforcement learning exposing a unified declarative interface to common RL algorithms, thus providing a backend to LIFT. We demonstrate the utility of LIFT in two case studies in database compound indexing and resource management in stream processing. Results show LIFT controllers initialized from demonstrations can outperform human baselines and heuristics across latency metrics and space usage by up to 70%. lift 
Lift Chart  The lift chart provides a visual summary of the usefulness of the information provided by one or more statistical models for predicting a binomial (categorical) outcome variable (dependent variable); for multinomial (multiplecategory) outcome variables, lift charts can be computed for each category. Specifically, the chart summarizes the utility that we may expect by using the respective predictive models, as compared to using baseline information only. The lift chart is applicable to most statistical methods that compute predictions (predicted classifications) for binomial or multinomial responses. Let us start with an example. A marketing agency is planning to send advertisements to selected households with the goal to boost sales of a product. The agency has a list of all households where each household is described by a set of attributes. Each advertisement sent costs a few pennies, but it is well paid off if the customer buys the product. Therefore an agency wants to minimize the number of advertisements sent, while at the same time maximize the number of sold products by reaching only the consumers that will actually buy the product. Therefore it develops a classifier that predicts the probability that a household is a potential customer. To fit this classifier and to express the dependency between the costs and the expected benefit the lift chart can be used. The number off all potential customers P is often unknown, therefore TPrate cannot be computed and the ROC curve cannot used, but the lift chart is useful in such settings. Also the TP is often hard to measure in practice; one might have just a few measurements from a sales analysis. Even in such cases lift chart can help the agency select the amount of most promising households to which an advertisement should be sent. Of course, lift charts are also useful for many other similar problems. http://…/vuk.pdf A lift chart, sometimes called a cumulative gains chart, or a banana chart, is a measure of model performance. It shows how responses, (i.e., to a direct mail solicitation, or a surgical treatment for instance) are changed by applying the model. This change ratio, which is hopefully, the increase in response rate, is called the ‘lift’. A lift chart indicates which subset of the dataset contains the greatest possible proportion of positive responses. The higher the lift curve is from the baseline, the better the performance of the model since the baseline represents the null model, which is no model at all. To explain a lift chart, suppose we had a twoclass prediction where the outcomes were yes (a positive response) or no (a negative response). To create a lift chart, instances in the dataset are sorted in descending probability order according to the predicted probability of a positive response. When the data is plotted, we can see a graphical depiction of the various probabilities. While the example shown in Figure 10 plots the results of different datasets for a single model, a lift chart can also be used to plot the results of a single dataset for different models. Note that the best model is not the one with the highest lift when it is being built. It is the model that performs the best on unseen, future data. http://…/dm_c_ov.pdf http://…/lift_chart.html gains 
Lifted Neural Network  We describe a novel family of models of multi layer feedforward neural networks in which the activation functions are encoded via penalties in the training problem. Our approach is based on representing a nondecreasing activation function as the argmin of an appropriate convex optimization problem. The new framework allows for algorithms such as blockcoordinate descent methods to be applied, in which each step is composed of a simple (no hidden layer) supervised learning problem that is parallelizable across data points and/or layers. Experiments indicate that the proposed models provide excellent initial guesses for weights for standard neural networks. In addition, the model provides avenues for interesting extensions, such as robustness against noisy inputs and optimizing over parameters in activation functions. 
Lifted Proximal Operator Machine (LPOM) 
We propose a new optimization method for training feedforward neural networks. By rewriting the activation function as an equivalent proximal operator, we approximate a feedforward neural network by adding the proximal operators to the objective function as penalties, hence we call the lifted proximal operator machine (LPOM). LPOM is block multiconvex in all layerwise weights and activations. This allows us to use block coordinate descent to update the layerwise weights and activations in parallel. Most notably, we only use the mapping of the activation function itself, rather than its derivatives, thus avoiding the gradient vanishing or blowup issues in gradient based training methods. So our method is applicable to various nondecreasing Lipschitz continuous activation functions, which can be saturating and nondifferentiable. LPOM does not require more auxiliary variables than the layerwise activations, thus using roughly the same amount of memory as stochastic gradient descent (SGD) does. We further prove the convergence of updating the layerwise weights and activations. Experiments on MNIST and CIFAR10 datasets testify to the advantages of LPOM. 
Lifting  The great advances of learningbased approaches in image processing and computer vision are largely based on deeply nested networks that compose linear transfer functions with suitable nonlinearities. Interestingly, the most frequently used nonlinearities in imaging applications (variants of the rectified linear unit) are uncommon in low dimensional approximation problems. In this paper we propose a novel nonlinear transfer function, called lifting, which is motivated from a related technique in convex optimization. A lifting layer increases the dimensionality of the input, naturally yields a linear spline when combined with a fully connected layer, and therefore closes the gap between low and high dimensional approximation problems. Moreover, applying the lifting operation to the loss layer of the network allows us to handle nonconvex and flat (zerogradient) cost functions. We analyze the proposed lifting theoretically, exemplify interesting properties in synthetic experiments and demonstrate its effectiveness in deep learning approaches to image classification and denoising. 
Lifting Autoencoder  In this work we introduce Lifting Autoencoders, a generative 3D surfacebased model of object categories. We bring together ideas from nonrigid structure from motion, image formation, and morphable models to learn a controllable, geometric model of 3D categories in an entirely unsupervised manner from an unstructured set of images. We exploit the 3D geometric nature of our model and use normal information to disentangle appearance into illumination, shading and albedo. We further use weak supervision to disentangle the nonrigid shape variability of human faces into identity and expression. We combine the 3D representation with a differentiable renderer to generate RGB images and append an adversarially trained refinement network to obtain sharp, photorealistic image reconstruction results. The learned generative model can be controlled in terms of interpretable geometry and appearance factors, allowing us to perform photorealistic image manipulation of identity, expression, 3D pose, and illumination properties. 
Light DualTask Neural Network (LDTNet) 
Singleimage dehazing is a challenging problem due to its illposed nature. Existing methods rely on a suboptimal twostep approach, where an intermediate product like a depth map is estimated, based on which the hazefree image is subsequently generated using an artificial prior formula. In this paper, we propose a light dualtask Neural Network called LDTNet that restores the hazefree image in one shot. We use transmission map estimation as an auxiliary task to assist the main task, haze removal, in feature extraction and to enhance the generalization of the network. In LDTNet, the hazefree image and the transmission map are produced simultaneously. As a result, the artificial prior is reduced to the smallest extent. Extensive experiments demonstrate that our algorithm achieves superior performance against the stateoftheart methods on both synthetic and realworld images. 
Light Recurrent Neural Networks (LightRNN) 
Recurrent neural networks (RNNs) have achieved stateoftheart performances in many natural language processing tasks, such as language modeling and machine translation. However, when the vocabulary is large, the RNN model will become very big (e.g., possibly beyond the memory capacity of a GPU device) and its training will become very inefficient. In this work, we propose a novel technique to tackle this challenge. The key idea is to use 2Component (2C) shared embedding for word representations. We allocate every word in the vocabulary into a table, each row of which is associated with a vector, and each column associated with another vector. Depending on its position in the table, a word is jointly represented by two components: a row vector and a column vector. Since the words in the same row share the row vector and the words in the same column share the column vector, we only need $2 \sqrt{V}$ vectors to represent a vocabulary of $V$ unique words, which are far less than the $V$ vectors required by existing approaches. Based on the 2Component shared embedding, we design a new RNN algorithm and evaluate it using the language modeling task on several benchmark datasets. The results show that our algorithm significantly reduces the model size and speeds up the training process, without sacrifice of accuracy (it achieves similar, if not better, perplexity as compared to stateoftheart language models). Remarkably, on the OneBillionWord benchmark Dataset, our algorithm achieves comparable perplexity to previous language models, whilst reducing the model size by a factor of 40100, and speeding up the training process by a factor of 2. We name our proposed algorithm \emph{LightRNN} to reflect its very small model size and very high training speed. 
LightFM Model (LightFM) 
I present a hybrid matrix factorisation model representing users and items as linear combinations of their content features’ latent factors. The model outperforms both collaborative and contentbased models in coldstart or sparse interaction data scenarios (using both user and item metadata), and performs at least as well as a pure collaborative matrix factorisation model where interaction data is abundant. Additionally, feature embeddings produced by the model encode semantic information in a way reminiscent of word embedding approaches, making them useful for a range of related tasks such as tag recommendations. 
Lightweight Convolutional Neural Network (LCNN) 
Edge computing efficiently extends the realm of information technology beyond the boundary defined by cloud computing paradigm. Performing computation near the source and destination, edge computing is promising to address the challenges in many delay sensitive applications, like real time surveillance. Leveraging the ubiquitously connected cameras and smart mobile devices, it enables video analytics at the edge. However, traditional human objects detection and tracking approaches are still computationally too expensive to edge devices. Aiming at intelligent surveillance as an edge network service, this work explored the feasibility of two popular human objects detection schemes, Harr Cascade and SVM, at the edge. Understanding the existing constraints of the algorithms, a lightweight Convolutional Neural Network (LCNN) is proposed using the depthwise separable convolution. The proposed LCNN considerably reduces the number of parameters without affecting the quality of the output, thus it is ideal for an edge device usage. Being trained with Single Shot Multi box Detector (SSD) to pinpoint each human object location, it gives coordination of bounding box around the object. We implemented and tested LCNN on an edge device using Raspberry PI 3. The intensive experimental comparison study has validated that the proposed LCNN is a feasible design for real time human object detection as an edge service. 
Lightweight EncoderDecoder Network (LEDNet) 
The extensive computational burden limits the usage of CNNs in mobile devices for dense estimation tasks. In this paper, we present a lightweight network to address this problem,namely LEDNet, which employs an asymmetric encoderdecoder architecture for the task of realtime semantic segmentation.More specifically, the encoder adopts a ResNet as backbone network, where two new operations, channel split and shuffle, are utilized in each residual block to greatly reduce computation cost while maintaining higher segmentation accuracy. On the other hand, an attention pyramid network (APN) is employed in the decoder to further lighten the entire network complexity. Our model has less than 1M parameters,and is able to run at over 71 FPS in a single GTX 1080Ti GPU. The comprehensive experiments demonstrate that our approach achieves stateoftheart results in terms of speed and accuracy tradeoff on CityScapes dataset. 
Lightweight Feature Fusion Network (LFFN) 
Single image superresolution(SISR) has witnessed great progress as convolutional neural network(CNN) gets deeper and wider. However, enormous parameters hinder its application to real world problems. In this letter, We propose a lightweight feature fusion network (LFFN) that can fully explore multiscale contextual information and greatly reduce network parameters while maximizing SISR results. LFFN is built on spindle blocks and a softmax feature fusion module (SFFM). Specifically, a spindle block is composed of a dimension extension unit, a feature exploration unit and a feature refinement unit. The dimension extension layer expands low dimension to high dimension and implicitly learns the feature maps which is suitable for the next unit. The feature exploration unit performs linear and nonlinear feature exploration aimed at different feature maps. The feature refinement layer is used to fuse and refine features. SFFM fuses the features from different modules in a selfadaptive learning manner with softmax function, making full use of hierarchical information with a small amount of parameter cost. Both qualitative and quantitative experiments on benchmark datasets show that LFFN achieves favorable performance against stateoftheart methods with similar parameters. 
Lightweight Probabilistic Deep Network  Even though probabilistic treatments of neural networks have a long history, they have not found widespread use in practice. Sampling approaches are often too slow already for simple networks. The size of the inputs and the depth of typical CNN architectures in computer vision only compound this problem. Uncertainty in neural networks has thus been largely ignored in practice, despite the fact that it may provide important information about the reliability of predictions and the inner workings of the network. In this paper, we introduce two lightweight approaches to making supervised learning with probabilistic deep networks practical: First, we suggest probabilistic output layers for classification and regression that require only minimal changes to existing networks. Second, we employ assumed density filtering and show that activation uncertainties can be propagated in a practical fashion through the entire network, again with minor changes. Both probabilistic networks retain the predictive power of the deterministic counterpart, but yield uncertainties that correlate well with the empirical error induced by their predictions. Moreover, the robustness to adversarial examples is significantly increased. 
Lightweight Pyramid of Networks (LPNet) 
Existing deep convolutional neural networks have found major success in image deraining, but at the expense of an enormous number of parameters. This limits their potential application, for example in mobile devices. In this paper, we propose a lightweight pyramid of networks (LPNet) for single image deraining. Instead of designing a complex network structures, we use domainspecific knowledge to simplify the learning process. Specifically, we find that by introducing the mature GaussianLaplacian image pyramid decomposition technology to the neural network, the learning problem at each pyramid level is greatly simplified and can be handled by a relatively shallow network with few parameters. We adopt recursive and residual network structures to build the proposed LPNet, which has less than 8K parameters while still achieving stateoftheart performance on rain removal. We also discuss the potential value of LPNet for other low and highlevel vision tasks. 
Likelihood  Likelihood is a funny concept. It’s not a probability, but it is proportional to a probability. The likelihood of a hypothesis (H) given some data (D) is proportional to the probability of obtaining D given that H is true, multiplied by an arbitrary positive constant (K). In other words, L(HD) = K Ā· P(DH). Since a likelihood isn’t actually a probability it doesn’t obey various rules of probability. For example, likelihood need not sum to 1. A critical difference between probability and likelihood is in the interpretation of what is fixed and what can vary. In the case of a conditional probability, P(DH), the hypothesis is fixed and the data are free to vary. Likelihood, however, is the opposite. The likelihood of a hypothesis, L(HD), conditions on the data as if they are fixed while allowing the hypotheses to vary. The distinction is subtle, so I’ll say it again. For conditional probability, the hypothesis is treated as a given and the data are free to vary. For likelihood, the data are a given and the hypotheses vary. ➘ “Likelihood Function” http://…/likelihood 
Likelihood Category Game Model (LCGM) 
To reach consensus among interacting agents is a problem of interest for social, economical, and political systems. A computational and mathematical framework to investigate consensus dynamics on complex networks is naming games. In general, naming is not an independent process but relies on perception and categorization. Existing works focus on consensus process of vocabulary evolution in a population of agents. However, in order to name an object, agents must first be able to distinguish objects according to their features. We articulate a likelihood category game model (LCGM) to integrate feature learning and the naming process. In the LCGM, selforganized agents can define category based on acquired knowledge through learning and use likelihood estimation to distinguish objects. The information communicated among the agents is no longer simply in some form of absolute answer, but involves one’s perception. Extensive simulations with LCGM reveal that a more complex knowledge makes it harder to reach consensus. We also find that agents with larger degree contribute more to the knowledge formation and are more likely to be intelligent. The proposed LCGM and the findings provide new insights into the emergence and evolution of consensus in complex systems in general. 
Likelihood Function  In statistics, a likelihood function (often simply the likelihood) is a function of the parameters of a statistical model. The likelihood of a set of parameter values, theta, given outcomes x, is equal to the probability of those observed outcomes given those parameter values, that is L(thetax) = P(xtheta). Likelihood functions play a key role in statistical inference, especially methods of estimating a parameter from a set of statistics. In informal contexts, “likelihood” is often used as a synonym for “probability.” But in statistical usage, a distinction is made depending on the roles of the outcome or parameter. Probability is used when describing a function of the outcome given a fixed parameter value. For example, if a coin is flipped 10 times and it is a fair coin, what is the probability of it landing headsup every time? Likelihood is used when describing a function of a parameter given an outcome. For example, if a coin is flipped 10 times and it has landed headsup 10 times, what is the likelihood that the coin is fair? 
Likelihood Ratio Similarity (LiRa) 
Recommender system data presents unique challenges to the data mining, machine learning, and algorithms communities. The high missing data rate, in combination with the large scale and high dimensionality that is typical of recommender systems data, requires new tools and methods for efficient data analysis. Here, we address the challenge of evaluating similarity between two users in a recommender system, where for each user only a small set of ratings is available. We present a new similarity score, that we call LiRa, based on a statistical model of user similarity, for largescale, discrete valued data with many missing values. We show that this score, based on a ratio of likelihoods, is more effective at identifying similar users than traditional similarity scores in userbased collaborative filtering, such as the Pearson correlation coefficient. We argue that our approach has significant potential to improve both accuracy and scalability in collaborative filtering. 
LikelihoodRatio Test (LRT) 
In statistics, a likelihood ratio test is a statistical test used to compare the fit of two models, one of which (the null model) is a special case of the other (the alternative model). The test is based on the likelihood ratio, which expresses how many times more likely the data are under one model than the other. This likelihood ratio, or equivalently its logarithm, can then be used to compute a pvalue, or compared to a critical value to decide whether to reject the null model in favour of the alternative model. When the logarithm of the likelihood ratio is used, the statistic is known as a loglikelihood ratio statistic, and the probability distribution of this test statistic, assuming that the null model is true, can be approximated using Wilks’s theorem. In the case of distinguishing between two models, each of which has no unknown parameters, use of the likelihood ratio test can be justified by the NeymanPearson lemma, which demonstrates that such a test has the highest power among all competitors. tsc 
Likert Scale  A Likert scale is a psychometric scale commonly involved in research that employs questionnaires. It is the most widely used approach to scaling responses in survey research, such that the term is often used interchangeably with rating scale, or more accurately the Likerttype scale, even though the two are not synonymous. The scale is named after its inventor, psychologist Rensis Likert. Likert distinguished between a scale proper, which emerges from collective responses to a set of items (usually eight or more), and the format in which responses are scored along a range. Technically speaking, a Likert scale refers only to the former. likert,Scale 
LIMESUP  Supervised Machine Learning (SML) algorithms such as Gradient Boosting, Random Forest, and Neural Networks have become popular in recent years due to their increased predictive performance over traditional statistical methods. This is especially true with large data sets (millions or more observations and hundreds to thousands of predictors). However, the complexity of the SML models makes them opaque and hard to interpret without additional tools. There has been a lot of interest recently in developing global and local diagnostics for interpreting and explaining SML models. In this paper, we propose locally interpretable models and effects based on supervised partitioning (trees) referred to as LIMESUP. This is in contrast with the KLIME approach that is based on clustering the predictor space. We describe LIMESUP based on fitting trees to the fitted response (LIMSUPR) as well as the derivatives of the fitted response (LIMESUPD). We compare the results with KLIME and describe its advantages using simulation and real data. 
Limit Deterministic Buchi Automaton (LDBA) 
LogicallyCorrect Reinforcement Learning 
Limited Gradient Descent  Label noise may handicap the generalization of classifiers, and it is an important issue how to effectively learn main pattern from samples with noisy labels. Recent studies have witnessed that deep neural networks tend to prioritize learning simple patterns and then memorize noise patterns. This suggests a method to search the best generalization, which learns the main pattern until the noise begins to be memorized. A natural idea is to use a supervised approach to find the stop timing of learning, for example resorting clean verification set. In practice, however, a clean verification set is sometimes not easy to obtain. To solve this problem, we propose an unsupervised method called limited gradient descent to estimate the best stop timing. We modified the labels of few samples in noisy dataset to be almost false labels as reverse pattern. By monitoring the learning progresses of the noisy samples and the reverse samples, we can determine the stop timing of learning. In this paper, we also provide some sufficient conditions on learning with noisy labels. Experimental results on CIFAR10 demonstrate that our approach has similar generalization performance to those supervised methods. For uncomplicated datasets, such as MNIST, we add relabeling strategy to further improve generalization and achieve stateoftheart performance. 
Limited Memory Steepest Descent Method (LMSD) 
The possibilities inherent in steepest descent methods have been considerably amplified by the introduction of the BarzilaiBorwein choice of stepsize, and other related ideas. These methods have proved to be competitive with conjugate gradient methods for the minimization of large dimension unconstrained minimization problems. This paper suggests a method which is able to take advantage of the availability of a few additional ālong’ vectors of storage to achieve a significant improvement in performance, both for quadratic and nonquadratic objective functions. It makes use of certain Ritz values related to the Lanczos process (Lanczos in J Res Nat Bur Stand 45:255282, 1950). Some underlying theory is provided, and numerical evidence is set out showing that the new method provides a competitive and more simple alternative to the state of the art lBFGS limited memory method. 
Limitedmemory BFGS (LBFGS) 
Limitedmemory BFGS (LBFGS or LMBFGS) is an optimization algorithm in the family of quasiNewton methods that approximates the BroydenFletcherGoldfarbShanno (BFGS) algorithm using a limited amount of computer memory. It is a popular algorithm for parameter estimation in machine learning. 
Lindy Effect  The Lindy effect is a theory of the life expectancy of nonperishable things that posits for a certain class of nonperishables, like a technology or an idea, every additional day may imply a longer (remaining) life expectancy: the mortality rate decreases with time. This contrasts with living creatures and mechanical things, which instead follow a bathtub curve, where every additional day in its life translates into a shorter additional life expectancy (though longer overall life expectancy, due to surviving this far): after childhood, the mortality rate increases with time. 
Line Map  linemap 
Linear Additive Markov Process (LAMP) 
We introduce LAMP: the Linear Additive Markov Process. Transitions in LAMP may be influenced by states visited in the distant history of the process, but unlike higherorder Markov processes, LAMP retains an efficient parametrization. LAMP also allows the specific dependence on history to be learned efficiently from data. We characterize some theoretical properties of LAMP, including its steadystate and mixing time. We then give an algorithm based on alternating minimization to learn LAMP models from data. Finally, we perform a series of realworld experiments to show that LAMP is more powerful than firstorder Markov processes, and even holds its own against deep sequential models (LSTMs) with a negligible increase in parameter complexity. 
Linear Algebra Package (LAPACK) 
LAPACK (Linear Algebra Package) is a software library for numerical linear algebra. It provides routines for solving systems of linear equations and linear least squares, eigenvalue problems, and singular value decomposition. It also includes routines to implement the associated matrix factorizations such as LU, QR, Cholesky and Schur decomposition. 
Linear Analog SelfAssessment (LASA) 
ordinalCont 
Linear Attention Recurrent Neural Network (LARNN) 
The Linear Attention Recurrent Neural Network (LARNN) is a recurrent attention module derived from the Long ShortTerm Memory (LSTM) cell and ideas from the consciousness Recurrent Neural Network (RNN). Yes, it LARNNs. The LARNN uses attention on its past cell state values for a limited window size $k$. The formulas are also derived from the Batch Normalized LSTM (BNLSTM) cell and the Transformer Network for its MultiHead Attention Mechanism. The MultiHead Attention Mechanism is used inside the cell such that it can query its own $k$ past values with the attention window. This has the effect of augmenting the rank of the tensor with the attention mechanism, such that the cell can perform complex queries to question its previous inner memories, which should augment the long shortterm effect of the memory. With a clever trick, the LARNN cell with attention can be easily used inside a loop on the cell state, just like how any other Recurrent Neural Network (RNN) cell can be looped linearly through time series. This is due to the fact that its state, which is looped upon throughout time steps within time series, stores the inner states in a ‘first in, first out’ queue which contains the $k$ most recent states and on which it is easily possible to add static positional encoding when the queue is represented as a tensor. This neural architecture yields better results than the vanilla LSTM cells. It can obtain results of 91.92% for the test accuracy, compared to the previously attained 91.65% using vanilla LSTM cells. Note that this is not to compare to other research, where up to 93.35% is obtained, but costly using 18 LSTM cells rather than with 2 to 3 cells as analyzed here. Finally, an interesting discovery is made, such that adding activation within the multihead attention mechanism’s linear layers can yield better results in the context researched hereto. 
Linear Centralization Classifier (LCC) 
A classification algorithm, called the Linear Centralization Classifier (LCC), is introduced. The algorithm seeks to find a transformation that best maps instances from the feature space to a space where they concentrate towards the center of their own classes, while maximimizing the distance between class centers. We formulate the classifier as a quadratic program with quadratic constraints. We then simplify this formulation to a linear program that can be solved effectively using a linear programming solver (e.g., simplexdual). We extend the formulation for LCC to enable the use of kernel functions for nonlinear classification applications. We compare our method with two standard classification methods (support vector machine and linear discriminant analysis) and four stateoftheart classification methods when they are applied to eight standard classification datasets. Our experimental results show that LCC is able to classify instances more accurately (based on the area under the receiver operating characteristic) in comparison to other tested methods on the chosen datasets. We also report the results for LCC with a particular kernel to solve for synthetic nonlinear classification problems. 
Linear Congruential Generator (LCG) 
A linear congruential generator (LCG) is an algorithm that yields a sequence of pseudorandomized numbers calculated with a discontinuous piecewise linear equation. The method represents one of the oldest and bestknown pseudorandom number generator algorithms. The theory behind them is relatively easy to understand, and they are easily implemented and fast, especially on computer hardware which can provide modulo arithmetic by storagebit truncation. 
Linear Dimension Reduction  Methods: 1. Principal component analysis (PCA) 2. Canonical correlation analysis (CCA) 3. Linear discriminant analysis (LDA) 4. Nonnegative matrix factorization (NMF) 5. Independent component analysis (ICA) LDRTools 
Linear Discriminant Analysis (LDA) 
Linear discriminant analysis (LDA) and the related Fisher’s linear discriminant are methods used in statistics, pattern recognition and machine learning to find a linear combination of features which characterizes or separates two or more classes of objects or events. The resulting combination may be used as a linear classifier, or, more commonly, for dimensionality reduction before later classification. Fisher’s Linear Discriminant Analysis 
Linear Discriminant Generative Adversarial Networks (LDGAN) 
We develop a novel method for training of GANs for unsupervised and class conditional generation of images, called Linear Discriminant GAN (LDGAN). The discriminator of an LDGAN is trained to maximize the linear separability between distributions of hidden representations of generated and targeted samples, while the generator is updated based on the decision hyperplanes computed by performing LDA over the hidden representations. LDGAN provides a concrete metric of separation capacity for the discriminator, and we experimentally show that it is possible to stabilize the training of LDGAN simply by calibrating the update frequencies between generators and discriminators in the unsupervised case, without employment of normalization methods and constraints on weights. In the class conditional generation tasks, the proposed method shows improved training stability together with better generalization performance compared to WGAN that employs an auxiliary classifier. 
Linear kJunta  We study the problem of testing if a function depends on a small number of linear directions of its input data. We call a function $f$ a \emph{linear $k$junta} if it is completely determined by some $k$dimensional subspace of the input space. In this paper, we study the problem of testing whether a given $n$ variable function $f : \mathbb{R}^n \to $, is a linear $k$junta or $\epsilon$far from all linear $k$juntas, where the closeness is measured with respect to the Gaussian measure on $\mathbb{R}^n$. This problems is a common generalization of (i) The combinatorial problem of junta testing on the hypercube which tests whether a Boolean function is dependent on at most $k$ of its variables and (ii) Geometric testing problems such as testing if a function is an intersection of $k$ halfspaces. We prove the existence of a $\mathsf{poly}(k \cdot s/\epsilon)$query nonadaptive tester for linear $k$juntas with surface area at most $s$. The polynomial dependence on $s$ is necessary as we provide a $\mathsf{poly}(s)$ lower bound on the query complexity of any nonadaptive test for linear juntas. Moreover, we show that if the function is a linear $k$junta with surface area at most $s$, then there is a $(s \cdot k)^{O(k)}$query nonadaptive algorithm to learn the function \emph{up to a rotation of the basis}. {We also use this procedure to obtain a nonadaptive tester (with the same query complexity) for subclasses of linear $k$juntas closed under rotation.} 
Linear Layered Probabilistic Shaping (LLPS) 
Linear layered probabilistic shaping (LLPS) is proposed, an architecture for linear codes to efficiently encode to shaped code words. In the previously proposed probabilistic amplitude shaping (PAS) architecture, a distribution matcher (DM) maps information bits to shaped bits, which are then systematically encoded by appending uniformly distributed parity bits. LLPS extends PAS by probabilistic parity shaping (PPS), which uses a syndrome DM to calculate shaped parity bits. LLPS enables the transmission with any desired distribution using linear codes, furthermore, by LLPS, a given linear code with rate $R_\text{fec}$ can be operated at any rate $R\leq R_\text{fec}$ by changing the distribution. LLPS is used with an LDPC code for dirty paper coding against an interfering BPSK signal, improving the energy efficiency by 0.8 dB. 
Linear Memory Network  Recurrent neural networks can learn complex transduction problems that require maintaining and actively exploiting a memory of their inputs. Such models traditionally consider memory and inputoutput functionalities indissolubly entangled. We introduce a novel recurrent architecture based on the conceptual separation between the functional inputoutput transformation and the memory mechanism, showing how they can be implemented through different neural components. By building on such conceptualization, we introduce the Linear Memory Network, a recurrent model comprising a feedforward neural network, realizing the nonlinear functional transformation, and a linear autoencoder for sequences, implementing the memory component. The resulting architecture can be efficiently trained by building on closedform solutions to linear optimization problems. Further, by exploiting equivalence results between feedforward and recurrent neural networks we devise a pretraining schema for the proposed architecture. Experiments on polyphonic music datasets show competitive results against gated recurrent networks and other state of the art models. 
Linear Mixed Effects Model  CLME,lmenssp 
Linear Mixed Model (LMM) 
A statistical model containing both fixed effects and random effects, that is: mixed effects. LMM is a kind of regression analysis. 
Linear Model UTree (LMUT) 
Deep Reinforcement Learning (DRL) has achieved impressive success in many applications. A key component of many DRL models is a neural network representing a Q function, to estimate the expected cumulative reward following a stateaction pair. The Q function neural network contains a lot of implicit knowledge about the RL problems, but often remains unexamined and uninterpreted. To our knowledge, this work develops the first mimic learning framework for Q functions in DRL. We introduce Linear Model Utrees (LMUTs) to approximate neural network predictions. An LMUT is learned using a novel online algorithm that is wellsuited for an active play setting, where the mimic learner observes an ongoing interaction between the neural net and the environment. Empirical evaluation shows that an LMUT mimics a Q function substantially better than five baseline methods. The transparent tree structure of an LMUT facilitates understanding the network’s learned knowledge by analyzing feature influence, extracting rules, and highlighting the superpixels in image inputs. 
Linear Programming (LP) 
Linear programming (LP; also called linear optimization) is a method to achieve the best outcome (such as maximum profit or lowest cost) in a mathematical model whose requirements are represented by linear relationships. Linear programming is a special case of mathematical programming (mathematical optimization). More formally, linear programming is a technique for the optimization of a linear objective function, subject to linear equality and linear inequality constraints. Its feasible region is a convex polyhedron, which is a set defined as the intersection of finitely many half spaces, each of which is defined by a linear inequality. Its objective function is a realvalued affine function defined on this polyhedron. A linear programming algorithm finds a point in the polyhedron where this function has the smallest (or largest) value if such a point exists. 
Linear Quadratic Estimation (LQE) 
Kalman filtering, also known as linear quadratic estimation (LQE), is an algorithm that uses a series of measurements observed over time, containing noise (random variations) and other inaccuracies, and produces estimates of unknown variables that tend to be more precise than those based on a single measurement alone. More formally, the Kalman filter operates recursively on streams of noisy input data to produce a statistically optimal estimate of the underlying system state. The filter is named after Rudolf (Rudy) E. KĆ”lmĆ”n, one of the primary developers of its theory. The Kalman filter has numerous applications in technology. A common application is for guidance, navigation and control of vehicles, particularly aircraft and spacecraft. Furthermore, the Kalman filter is a widely applied concept in time series analysis used in fields such as signal processing and econometrics. Kalman filters also are one of the main topics in the field of Robotic motion planning and control, and sometimes included in Trajectory optimization. 
Linear Quadratic Regulator  The theory of optimal control is concerned with operating a dynamic system at minimum cost. The case where the system dynamics are described by a set of linear differential equations and the cost is described by a quadratic function is called the LQ problem. One of the main results in the theory is that the solution is provided by the linearquadratic regulator (LQR), a feedback controller whose equations are given below. The LQR is an important part of the solution to the LQG (linearquadraticGaussian) problem. Like the LQR problem itself, the LQG problem is one of the most fundamental problems in control theory. 
Linear Regression  In statistics, linear regression is an approach for modeling the relationship between a scalar dependent variable y and one or more explanatory variables (or independent variable) denoted X. The case of one explanatory variable is called simple linear regression. For more than one explanatory variable, the process is called multiple linear regression. (This term should be distinguished from multivariate linear regression, where multiple correlated dependent variables are predicted, rather than a single scalar variable.) In linear regression, data are modeled using linear predictor functions, and unknown model parameters are estimated from the data. Such models are called linear models. Most commonly, linear regression refers to a model in which the conditional mean of y given the value of X is an affine function of X. Less commonly, linear regression could refer to a model in which the median, or some other quantile of the conditional distribution of y given X is expressed as a linear function of X. Like all forms of regression analysis, linear regression focuses on the conditional probability distribution of y given X, rather than on the joint probability distribution of y and X, which is the domain of multivariate analysis. Linear regression was the first type of regression analysis to be studied rigorously, and to be used extensively in practical applications. This is because models which depend linearly on their unknown parameters are easier to fit than models which are nonlinearly related to their parameters and because the statistical properties of the resulting estimators are easier to determine. Linear regression has many practical uses. Most applications fall into one of the following two broad categories: Ā· If the goal is prediction, or forecasting, or reduction, linear regression can be used to fit a predictive model to an observed data set of y and X values. After developing such a model, if an additional value of X is then given without its accompanying value of y, the fitted model can be used to make a prediction of the value of y. Ā· Given a variable y and a number of variables X1, …, Xp that may be related to y, linear regression analysis can be applied to quantify the strength of the relationship between y and the Xj, to assess which Xj may have no relationship with y at all, and to identify which subsets of the Xj contain redundant information about y. Linear regression models are often fitted using the least squares approach, but they may also be fitted in other ways, such as by minimizing the ‘lack of fit’ in some other norm (as with least absolute deviations regression), or by minimizing a penalized version of the least squares loss function as in ridge regression (L2norm penalty) and lasso (L1norm penalty). Conversely, the least squares approach can be used to fit models that are not linear models. Thus, although the terms ‘least squares’ and ‘linear model’ are closely linked, they are not synonymous. 
Linear Superiorization (LinSup) 
Linear superiorization (abbreviated: LinSup) considers linear programming (LP) problems wherein the constraints as well as the objective function are linear. It allows to steer the iterates of a feasibilityseeking iterative process toward feasible points that have lower (not necessarily minimal) values of the objective function than points that would have been reached by the same feasiblityseeking iterative process without superiorization. Using a feasibilityseeking iterative process that converges even if the linear feasible set is empty, LinSup generates an iterative sequence that converges to a point that minimizes a proximity function which measures the linear constraints violation. In addition, due to LinSup’s repeated objective function reduction steps such a point will most probably have a reduced objective function value. We present an exploratory experimental result that illustrates the behavior of LinSup on an infeasible LP problem. 
Linear Unified LASSO (LLASSO) 
We propose a rescaled LASSO, by premultipying the LASSO with a matrix term, namely linear unified LASSO (LLASSO) for multicollinear situations. Our numerical study has shown that the LLASSO is comparable with other sparse modeling techniques and often outperforms the LASSO and elastic net. Our findings open new visions about using the LASSO still for sparse modeling and variable selection. We conclude our study by pointing that the LLASSO can be solved by the same efficient algorithm for solving the LASSO and suggest to follow the same construction technique for other penalized estimators. 
LinearEpoch GradualWarmup (LEGW) 
Largebatch training approaches have enabled researchers to utilize largescale distributed processing and greatly accelerate deepneural net (DNN) training. For example, by scaling the batch size from 256 to 32K, researchers have been able to reduce the training time of ResNet50 on ImageNet from 29 hours to 2.2 minutes (Ying et al., 2018). In this paper, we propose a new approach called linearepoch gradualwarmup (LEGW) for better largebatch training. With LEGW, we are able to conduct largebatch training for both CNNs and RNNs with the Sqrt Scaling scheme. LEGW enables Sqrt Scaling scheme to be useful in practice and as a result we achieve much better results than the Linear Scaling learning rate scheme. For LSTM applications, we are able to scale the batch size by a factor of 64 without losing accuracy and without tuning the hyperparameters. For CNN applications, LEGW is able to achieve the same accuracy even as we scale the batch size to 32K. LEGW works better than previous largebatch autotuning techniques. LEGW achieves a 5.3X average speedup over the baselines for four LSTMbased applications on the same hardware. We also provide some theoretical explanations for LEGW. 
Linearized Binary Regression  Probit regression was first proposed by Bliss in 1934 to study mortality rates of insects. Since then, an extensive body of work has analyzed and used probit or related binary regression methods (such as logistic regression) in numerous applications and fields. This paper provides a fresh angle to such wellestablished binary regression methods. Concretely, we demonstrate that linearizing the probit model in combination with linear estimators performs on par with stateoftheart nonlinear regression methods, such as posterior mean or maximum aposteriori estimation, for a broad range of realworld regression problems. We derive exact, closedform, and nonasymptotic expressions for the meansquared error of our linearized estimators, which clearly separates them from nonlinear regression methods that are typically difficult to analyze. We showcase the efficacy of our methods and results for a number of synthetic and realworld datasets, which demonstrates that linearized binary regression finds potential use in a variety of inference, estimation, signal processing, and machine learning applications that deal with binaryvalued observations or measurements. 
linearized Gaussian Process  Datadriven Model Predictive Control (MPC), where the system model is learned from data with machine learning, has recently gained increasing interests in the control community. Gaussian Processes (GP), as a type of statistical models, are particularly attractive due to their modeling flexibility and their ability to provide probabilistic estimates of prediction uncertainty. GPbased MPC has been developed and applied, however the optimization problem is typically nonconvex and highly demanding, and scales poorly with model size. This causes unsatisfactory solving performance, even with stateoftheart solvers, and makes the approach less suitable for realtime control. We develop a method based on a new concept, called linearized Gaussian Process, and Sequential Convex Programming, that can significantly improve the solving performance of GPbased MPC. Our method is not only faster but also much more scalable and predictable than other commonly used methods, as it is much less influenced by the model size. The efficiency and advantages of the algorithm are demonstrated clearly in a numerical example. 
LinearTime Clustering Algorithm (Ksets+) 
In this paper, we first propose a new iterative algorithm, called the Ksets+ algorithm for clustering data points in a semimetric space, where the distance measure does not necessarily satisfy the triangular inequality. We show that the Ksets+ algorithm converges in a finite number of iterations and it retains the same performance guarantee as the Ksets algorithm for clustering data points in a metric space. We then extend the applicability of the Ksets+ algorithm from data points in a semimetric space to data points that only have a symmetric similarity measure. Such an extension leads to great reduction of computational complexity. In particular, for an n * n similarity matrix with m nonzero elements in the matrix, the computational complexity of the Ksets+ algorithm is O((Kn + m)I), where I is the number of iterations. The memory complexity to achieve that computational complexity is O(Kn + m). As such, both the computational complexity and the memory complexity are linear in n when the n * n similarity matrix is sparse, i.e., m = O(n). We also conduct various experiments to show the effectiveness of the Ksets+ algorithm by using a synthetic dataset from the stochastic block model and a real network from the WonderNetwork website. 
LinearTime Detection of NonLinear Changes (LIGHT) 
Change detection in multivariate time series has applications in many domains, including health care and network monitoring. A common approach to detect changes is to compare the divergence between the distributions of a reference window and a test window. When the number of dimensions is very large, however, the naĀØÄ±ve approach has both quality and efficiency issues: to ensure robustness the window size needs to be large, which not only leads to missed alarms but also increases runtime. To this end, we propose LIGHT, a lineartime algorithm for robustly detecting nonlinear changes in massively high dimensional time series. Importantly, LIGHT provides high flexibility in choosing the window size, allowing the domain expert to fit the level of details required. To do such, we 1) perform scalable PCA to reduce dimensionality, 2) perform scalable factorization of the joint distribution, and 3) scalably compute divergences between these lower dimensional distributions. Extensive empirical evaluation on both synthetic and realworld data show that LIGHT outperforms state of the art with up to 100% improvement in both quality and efficiency. 
LineNet  High Definition (HD) maps play an important role in modern traffic scenes. However, the development of HD maps coverage grows slowly because of the cost limitation. To efficiently model HD maps, we proposed a convolutional neural network with a novel prediction layer and a zoom module, called LineNet. It is designed for stateoftheart lane detection in an unordered crowdsourced image dataset. And we introduced TTLane, a dataset for efficient lane detection in urban road modeling applications. Combining LineNet and TTLane, we proposed a pipeline to model HD maps with crowdsourced data for the first time. And the maps can be constructed precisely even with inaccurate crowdsourced data. 
Linguistic Descriptions of Complex Phenomena (LDCP) 
Linguistic Descriptions of Complex Phenomena (LDCP) is an architecture and methodology that allows us to model complex phenomena, interpreting input data, and generating automatic text reports customized to the user needs (see <doi:10.1016/j.ins.2016.11.002> and <doi:10.1007/s0050001624305> ). rLDCP 
Lingvo  Lingvo is a Tensorflow framework offering a complete solution for collaborative deep learning research, with a particular focus towards sequencetosequence models. Lingvo models are composed of modular building blocks that are flexible and easily extensible, and experiment configurations are centralized and highly customizable. Distributed training and quantized inference are supported directly within the framework, and it contains existing implementations of a large number of utilities, helper functions, and the newest research ideas. Lingvo has been used in collaboration by dozens of researchers in more than 20 papers over the last two years. This document outlines the underlying design of Lingvo and serves as an introduction to the various pieces of the framework, while also offering examples of advanced features that showcase the capabilities of the framework. 
Link Function  In GLM, the link function provides the relationship between the linear predictor and the mean of the distribution function. There are many commonly used link functions, and their choice can be somewhat arbitrary. It makes sense to try to match the domain of the link function to the range of the distribution function’s mean. 
Link Prediction  Given a snapshot of a social network, can we infer which new interactions among its members are likely to occur in the near future? We formalize this question as the link prediction problem, and develop approaches to link prediction based on measures for analyzing the \proximity” of nodes in a network. Experiments on large coauthorship networks suggest that information about future interactions can be extracted from network topology alone, and that fairly subtle measures for detecting node proximity can outperform more direct measures. 
Link Prediction Adversarial Attack  ➚ “Iterative Gradient Attack” 
Linked Causal Variational Autoencoder (LCVA) 
Modeling spillover effects from observational data is an important problem in economics, business, and other fields of research. % It helps us infer the causality between two seemingly unrelated set of events. For example, if consumer spending in the United States declines, it has spillover effects on economies that depend on the U.S. as their largest export market. In this paper, we aim to infer the causation that results in spillover effects between pairs of entities (or units), we call this effect as \textit{paired spillover}. To achieve this, we leverage the recent developments in variational inference and deep learning techniques to propose a generative model called Linked Causal Variational Autoencoder (LCVA). Similar to variational autoencoders (VAE), LCVA incorporates an encoder neural network to learn the latent attributes and a decoder network to reconstruct the inputs. However, unlike VAE, LCVA treats the \textit{latent attributes as confounders that are assumed to affect both the treatment and the outcome of units}. Specifically, given a pair of units $u$ and $\bar{u}$, their individual treatment and outcomes, the encoder network of LCVA samples the confounders by conditioning on the observed covariates of $u$, the treatments of both $u$ and $\bar{u}$ and the outcome of $u$. Once inferred, the latent attributes (or confounders) of $u$ captures the spillover effect of $\bar{u}$ on $u$. Using a network of users from job training dataset (LaLonde (1986)) and copurchase dataset from Amazon ecommerce domain, we show that LCVA is significantly more robust than existing methods in capturing spillover effects. 
Linked Data Ranking Algorithm (LDRANK) 
The advances of the Linked Open Data (LOD) initiative are giving rise to a more structured Web of data. Indeed, a few datasets act as hubs (e.g., DBpedia) connecting many other datasets. They also made possible new Web services for entity detection inside plain text (e.g., DBpedia Spotlight), thus allowing for new applications that can benefit from a combination of the Web of documents and the Web of data. To ease the emergence of these new applications, we propose a querybiased algorithm (LDRANK) for the ranking of web of data resources with associated textual data. Our algorithm combines link analysis with dimensionality reduction. We use crowdsourcing for building a publicly available and reusable dataset for the evaluation of querybiased ranking of Web of data resources detected in Web pages. We show that, on this dataset, LDRANK outperforms the state of the art. Finally, we use this algorithm for the construction of semantic snippets of which we evaluate the usefulness with a crowdsourcingbased approach. 
Linked Matrix Factorization (LMF) 
In recent years, a number of methods have been developed for the dimension reduction and decomposition of multiple linked highcontent data matrices. Typically these methods assume that just one dimension, rows or columns, is shared among the data sources. This shared dimension may represent common features that are measured for different sample sets (i.e., horizontal integration) or a common set of samples with measurements for different feature sets (i.e., vertical integration). In this article we introduce an approach for simultaneous horizontal and vertical integration, termed Linked Matrix Factorization (LMF), for the more general situation where some matrices share rows (e.g., features) and some share columns (e.g., samples). Our motivating application is a cytotoxicity study with accompanying genomic and molecular chemical attribute data. In this data set, the toxicity matrix (cell lines $\times$ chemicals) shares its sample set with a genotype matrix (cell lines $\times$ SNPs), and shares its feature set with a chemical molecular attribute matrix (chemicals $\times$ attributes). LMF gives a unified lowrank factorization of these three matrices, which allows for the decomposition of systematic variation that is shared among the three matrices and systematic variation that is specific to each matrix. This may be used for efficient dimension reduction, exploratory visualization, and the imputation of missing data even when entire rows or columns are missing from a constituent data matrix. We present theoretical results concerning the uniqueness, identifiability, and minimal parametrization of LMF, and evaluate it with extensive simulation studies. 
Linked Micromaps  Linked Micromaps is a graphing program written in Java. It allows users to view multiple variables interactively and compare statistics across regions (states, counties, registries, hospitals) as well as across time. It supports six types of graph: Ā· bar graphs; Ā· box plots; Ā· raw data tables; Ā· point graphs; Ā· point graphs with arrow; and Ā· point graphs with confidence intervals. 
Linked Recurrent Neural Network (LinkedRNN) 
Recurrent Neural Networks (RNNs) have been proven to be effective in modeling sequential data and they have been applied to boost a variety of tasks such as document classification, speech recognition and machine translation. Most of existing RNN models have been designed for sequences assumed to be identically and independently distributed (i.i.d). However, in many realworld applications, sequences are naturally linked. For example, web documents are connected by hyperlinks; and genes interact with each other. On the one hand, linked sequences are inherently not i.i.d., which poses tremendous challenges to existing RNN models. On the other hand, linked sequences offer link information in addition to the sequential information, which enables unprecedented opportunities to build advanced RNN models. In this paper, we study the problem of RNN for linked sequences. In particular, we introduce a principled approach to capture link information and propose a linked Recurrent Neural Network (LinkedRNN), which models sequential and link information coherently. We conduct experiments on realworld datasets from multiple domains and the experimental results validate the effectiveness of the proposed framework. 
Linking Artificial Intelligence Principles (LAIP) 
Artificial Intelligence principles define social and ethical considerations to develop future AI. They come from research institutes, government organizations and industries. All versions of AI principles are with different considerations covering different perspectives and making different emphasis. None of them can be considered as complete and can cover the rest AI principle proposals. Here we introduce LAIP, an effort and platform for linking and analyzing different Artificial Intelligence Principles. We want to explicitly establish the common topics and links among AI Principles proposed by different organizations and investigate on their uniqueness. Based on these efforts, for the longterm future of AI, instead of directly adopting any of the AI principles, we argue for the necessity of incorporating various AI Principles into a comprehensive framework and focusing on how they can interact and complete each other. 
Links  We present a novel algorithm, called Links, designed to perform online clustering on unit vectors in a highdimensional Euclidean space. The algorithm is appropriate when it is necessary to cluster data efficiently as it streams in, and is to be contrasted with traditional batch clustering algorithms that have access to all data at once. For example, Links has been successfully applied to embedding vectors generated from face images or voice recordings for the purpose of recognizing people, thereby providing realtime identification during video or audio capture. 
Linky  User identity linkage across online social networks is an emerging research topic that has attracted attention in recent years. Many user identity linkage methods have been proposed so far and most of them utilize user profile, content and network information to determine if two social media accounts belong to the same person. In most cases, user identity linkage methods are evaluated by performing some prediction tasks with the results presented using some overall accuracy measures. However, the methods are rarely compared at the individual user level where a predicted matched (or linked) pair of user identities from different online social networks can be visually compared in terms of user profile (e.g. username), content and network information. Such a comparison is critical to determine the relative strengths and weaknesses of each method. In this work, we present Linky, a visual analytical tool which extracts the results from different user identity linkage methods performed on multiple online social networks and visualizes the user profiles, content and ego networks of the linked user identities. Linky is designed to help researchers to (a) inspect the linked user identities at the individual user level, (b) compare results returned by different user linkage methods, and (c) provide a preliminary empirical understanding on which aspects of the user identities, e.g. profile, content or network, contributed to the user identity linkage results. 
LINSPECTOR  Despite an ever growing number of word representation models introduced for a large number of languages, there is a lack of a standardized technique to provide insights into what is captured by these models. Such insights would help the community to get an estimate of the downstream task performance, as well as to design more informed neural architectures, while avoiding extensive experimentation which requires substantial computational resources not all researchers have access to. A recent development in NLP is to use simple classification tasks, also called probing tasks, that test for a single linguistic feature such as partofspeech. Existing studies mostly focus on exploring the information encoded by the sentencelevel representations for English. However, from a typological perspective the morphologically poor English is rather an outlier: the information encoded by the word order and function words in English is often stored on a subword, morphological level in other languages. To address this, we introduce 15 wordlevel probing tasks such as case marking, possession, word length, morphological tag count and pseudoword identification for 24 languages. We present experiments on several state of the art word embedding models, in which we relate the probing task performance for a diverse set of languages to a range of classic NLP tasks such as semantic role labeling and natural language inference. We find that a number of probing tests have significantly high positive correlation to the downstream tasks, especially for morphologically rich languages. We show that our tests can be used to explore word embeddings or blackbox neural models for linguistic cues in a multilingual setting. We release the probing datasets and the evaluation suite with https://…/linspector. 
Lipizzaner  GANs are difficult to train due to convergence pathologies such as mode and discriminator collapse. We introduce Lipizzaner, an open source software system that allows machine learning engineers to train GANs in a distributed and robust way. Lipizzaner distributes a competitive coevolutionary algorithm which, by virtue of dual, adapting, generator and discriminator populations, is robust to collapses. The algorithm is well suited to efficient distribution because it uses a spatial grid abstraction. Training is local to each cell and strong intermediate training results are exchanged among overlapping neighborhoods allowing high performing solutions to propagate and improve with more rounds of training. Experiments on common image datasets overcome critical collapses. Communication overhead scales linearly when increasing the number of compute instances and we observe that increasing scale leads to improved model performance. 
Lipschitz Generative Adversarial Net  In this paper we study the convergence of generative adversarial networks (GANs) from the perspective of the informativeness of the gradient of the optimal discriminative function. We show that GANs without restriction on the discriminative function space commonly suffer from the problem that the gradient produced by the discriminator is uninformative to guide the generator. By contrast, Wasserstein GAN (WGAN), where the discriminative function is restricted to $1$Lipschitz, does not suffer from such a gradient uninformativeness problem. We further show in the paper that the model with a compact dual form of Wasserstein distance, where the Lipschitz condition is relaxed, also suffers from this issue. This implies the importance of Lipschitz condition and motivates us to study the general formulation of GANs with Lipschitz constraint, which leads to a new family of GANs that we call Lipschitz GANs (LGANs). We show that LGANs guarantee the existence and uniqueness of the optimal discriminative function as well as the existence of a unique Nash equilibrium. We prove that LGANs are generally capable of eliminating the gradient uninformativeness problem. According to our empirical analysis, LGANs are more stable and generate consistently higher quality samples compared with WGAN. 
Liquid Analytics  Liquid analytics. That’s the part that automatically updates and refines the training sets, rules, inferences, confidence intervals and predictions, every day, as mutating data keeps pouring nonstop in the databases (be it NoSQL or not). While this (most of the time) still ends up being coded in production mode by software engineers or developers, the framework and logical architecture is designed by data scientists. Because of this, data science is to data floods what statistical science is to frozen data. 
LISAL  Most environmental phenomena, such as wind profiles, ozone concentration and sunlight distribution under a forest canopy, exhibit nonstationary dynamics i.e. phenomenon variation change depending on the location and time of occurrence. Nonstationary dynamics pose both theoretical and practical challenges to statistical machine learning algorithms aiming to accurately capture the complexities governing the evolution of such processes. In this paper, we address the sampling aspects of the problem of learning nonstationary spatiotemporal models, and propose an efficient yet simple algorithm – LISAL. The core idea in LISAL is to learn two models using Gaussian processes (GPs) wherein the first is a nonstationary GP directly modeling the phenomenon. The second model uses a stationary GP representing a latent space corresponding to changes in dynamics, or the nonstationarity characteristics of the first model. LISAL involves adaptively sampling the latent space dynamics using information theory quantities to reduce the computational cost during the learning phase. The relevance of LISAL is extensively validated using multiple real world datasets. Efficiently Learning Nonstationary Gaussian Processes 
ListDecodable Linear Regression  We give the first polynomialtime algorithm for robust regression in the listdecodable setting where an adversary can corrupt a greater than $1/2$ fraction of examples. For any $\alpha < 1$, our algorithm takes as input a sample $_{i \leq n}$ of $n$ linear equations where $\alpha n$ of the equations satisfy $y_i = \langle x_i,\ell^*\rangle +\zeta$ for some small noise $\zeta$ and $(1\alpha)n$ of the equations are \emph{arbitrarily} chosen. It outputs a list $L$ of size $O(1/\alpha)$ – a fixed constant – that contains an $\ell$ that is close to $\ell^*$. Our algorithm succeeds whenever the inliers are chosen from a \emph{certifiably} anticoncentrated distribution $D$. As a special case, this yields a $(d/\alpha)^{O(1/\alpha^8)}$ time algorithm to find a $O(1/\alpha)$ size list when the inlier distribution is a standard Gaussian. The anticoncentration assumption on the inliers is informationtheoretically necessary. Our algorithm works for more general distributions under the additional assumption that $\ell^*$ is Boolean valued. To solve the problem we introduce a new framework for listdecodable learning that strengthens the sumofsquares `identifiability to algorithms’ paradigm. In an independent work, Raghavendra and Yau [RY19] have obtained a similar result for listdecodable regression also using the sumofsquares method. 
ListOps  Latent tree learning models learn to parse a sentence without syntactic supervision, and use that parse to build the sentence representation. Existing work on such models has shown that, while they perform well on tasks like sentence classification, they do not learn grammars that conform to any plausible semantic or syntactic formalism (Williams et al., 2018a). Studying the parsing ability of such models in natural language can be challenging due to the inherent complexities of natural language, like having several valid parses for a single sentence. In this paper we introduce ListOps, a toy dataset created to study the parsing ability of latent tree models. ListOps sequences are in the style of prefix arithmetic. The dataset is designed to have a single correct parsing strategy that a system needs to learn to succeed at the task. We show that the current leading latent tree models are unable to learn to parse and succeed at ListOps. These models achieve accuracies worse than purely sequential RNNs. 
LISTwise ExplaiNer (LISTEN) 
There is an increasing demand for algorithms to explain their outcomes. So far, there is no method that explains the rankings produced by a ranking algorithm. To address this gap we propose LISTEN, a LISTwise ExplaiNer, to explain rankings produced by a ranking algorithm. To efficiently use LISTEN in production, we train a neural network to learn the underlying explanation space created by LISTEN; we call this model QLISTEN. We show that LISTEN produces faithful explanations and that QLISTEN is able to learn these explanations. Moreover, we show that LISTEN is safe to use in a real world environment: users of a news recommendation system do not behave significantly differently when they are exposed to explanations generated by LISTEN instead of manually generated explanations. 
Literate Programming  Literate programming is an approach to programming introduced by Donald Knuth in which a program is given as an explanation of the program logic in a natural language, such as English, interspersed with snippets of macros and traditional source code, from which a compilable source code can be generated. 
littler (“little R”)  littler provides the r program, a simplified commandline interface for GNU R. This allows direct execution of commands, use in piping where the output of one program supplies the input of the next, as well as adding the ability for writing hashbang scripts, i.e. creating executable files starting with, say, #!/usr/bin/r. GNU R, a language and environment for statistical computing and graphics, provides a wonderful system for ‘programming with data’ as well as interactive exploratory analysis, often involving graphs. Sometimes, however, simple scripts are desired. While R can be used in batch mode, and while socalled here documents can be crafted, a longstanding need for a scripting frontend has often been expressed by the R Community. littler (pronounced little R and written r) aims to fill this need. 
LjungBox Test  The LjungBox test (named for Greta M. Ljung and George E. P. Box) is a type of statistical test of whether any of a group of autocorrelations of a time series are different from zero. Instead of testing randomness at each distinct lag, it tests the ‘overall’ randomness based on a number of lags, and is therefore a portmanteau test. This test is sometimes known as the LjungBox Q test, and it is closely connected to the BoxPierce test (which is named after George E. P. Box and David A. Pierce). In fact, the LjungBox test statistic was described explicitly in the paper that led to the use of the BoxPierce statistic, and from which that statistic takes its name. The BoxPierce test statistic is a simplified version of the LjungBox statistic for which subsequent simulation studies have shown poor performance. The LjungBox test is widely applied in econometrics and other applications of time series analysis. http://…/ljungboxtest 
LloydMax  In computer science and electrical engineering, Lloyd’s algorithm, also known as Voronoi iteration or relaxation, is an algorithm named after Stuart P. Lloyd for finding evenly spaced sets of points in subsets of Euclidean spaces, and partitions of these subsets into wellshaped and uniformly sized convex cells. Like the closely related kmeans clustering algorithm, it repeatedly finds the centroid of each set in the partition, and then repartitions the input according to which of these centroids is closest. However, Lloyd’s algorithm differs from kmeans clustering in that its input is a continuous geometric region rather than a discrete set of points. Thus, when repartitioning the input, Lloyd’s algorithm uses Voronoi diagrams rather than simply determining the nearest center to each of a finite set of points as the kmeans algorithm does. Although the algorithm may be applied most directly to the Euclidean plane, similar algorithms may also be applied to higherdimensional spaces or to spaces with other nonEuclidean metrics. Lloyd’s algorithm can be used to construct close approximations to centroidal Voronoi tessellations of the input, which can be used for quantization, dithering, and stippling. Other applications of Lloyd’s algorithm include smoothing of triangle meshes in the finite element method. ➚ “Compressive Kmeans” 
LMKLNet  In this paper we propose solving localized multiple kernel learning (LMKL) using LMKLNet, a feedforward deep neural network. In contrast to previous works, as a learning principle we propose {\em parameterizing} both the gating function for learning kernel combination weights and the multiclass classifier in LMKL using an attentional network (AN) and a multilayer perceptron (MLP), respectively. In this way we can learn the (nonlinear) decision function in LMKL (approximately) by sequential applications of AN and MLP. Empirically on benchmark datasets we demonstrate that overall LMKLNet can not only outperform the stateoftheart MKL solvers in terms of accuracy, but also be trained about {\em two orders of magnitude} faster with much smaller memory footprint for largescale learning. 
LMLDS  ➘ “Multilinear Dynamical System” 
LMPartial Conditional Mutual Information From Mixed Embedding (LMPMIME) 
Causal analysis based on nonuniform embedding schemes is an important way to detect the underlying interactions between dynamic systems. However, there are still some obstacles to estimate highdimensional conditional mutual information and form optimal mixed embedding vector in traditional nonuniform embedding schemes. In this study, we present a new nonuniform embedding method framed in information theory to detect causality for multivariate time series, named LMPMIME, which integrates the lowdimensional approximation of conditional mutual information and the mixed search strategy for the construction of the mixed embedding vector. We apply the proposed method to simulations of linear stochastic, nonlinear stochastic, and chaotic systems, demonstrating its superiority over partial conditional mutual information from mixed embedding (PMIME) method. Moreover, the proposed method works well for multivariate time series with weak coupling strengths, especially for chaotic systems. In the actual application, we show its applicability to epilepsy multichannel electrocorticographic recordings. 
Lmser Network  Proposed in 1991, Least Mean Square Error Reconstruction for selforganizing network, shortly Lmser, was a further development of the traditional autoencoder (AE) by folding the architecture with respect to the central coding layer and thus leading to the features of symmetric weights and neurons, as well as jointly supervised and unsupervised learning. However, its advantages were only demonstrated in a onehiddenlayer implementation due to the lack of computing resources and big data at that time. In this paper, we revisit Lmser from the perspective of deep learning, develop Lmser network based on multiple convolutional layers, which is more suitable for imagerelated tasks, and confirm several Lmser functions with preliminary demonstrations on image recognition, reconstruction, association recall, and so on. Experiments demonstrate that Lmser indeed works as indicated in the original paper, and it has promising performance in various applications. 
LoadSpy  Modern software packages have become increasingly complex with millions of lines of code and references to many external libraries. Redundant operations are a common performance limiter in these code bases. Missed compiler optimization opportunities, inappropriate data structure and algorithm choices, and developers’ inattention to performance are some common reasons for the existence of redundant operations. Developers mainly depend on compilers to eliminate redundant operations. However, compilers’ static analysis often misses optimization opportunities due to ambiguities and limited analysis scope; automatic optimizations to algorithmic and data structural problems are out of scope. We develop LoadSpy, a wholeprogram profiler to pinpoint redundant memory load operations, which are often a symptom of many redundant operations. The strength of LoadSpy exists in identifying and quantifying redundant load operations in programs and associating the redundancies with program execution contexts and scopes to focus developers’ attention on problematic code. LoadSpy works on fully optimized binaries, adopts various optimization techniques to reduce its overhead, and provides a rich graphic user interface, which make it a complete developer tool. Applying LoadSpy showed that a large fraction of redundant loads is common in modern software packages despite highest levels of automatic compiler optimizations. Guided by LoadSpy, we optimize several wellknown benchmarks and realworld applications, yielding significant speedups. 
LOBO  Botnets in online social networks are increasingly often affecting the regular flow of discussion, attacking regular users and their posts, spamming them with irrelevant or offensive content, and even manipulating the popularity of messages and accounts. Researchers and cybercriminals are involved in an arms race, and new and updated botnets designed to defeat current detection systems are constantly developed, rendering such detection systems obsolete. In this paper, we motivate the need for a generalized evaluation in Twitter bot detection and propose a methodology to evaluate bot classifiers by testing them on unseen bot classes. We show that this methodology is empirically robust, using bot classes of varying sizes and characteristics and reaching similar results, and argue that methods trained and tested on single bot classes or datasets might not able to generalize to new bot classes. We train one such classifier on over 200,000 data points and show that it achieves over 97% accuracy. The data used to train and test this classifier includes some of the largest and most varied collections of bots used in literature. We then test this theoretically sound classifier using our methodology, highlighting that it does not generalize well to unseen bot classes. Finally, we discuss the implications of our results, and reasons why some bot classes are easier and faster to detect than others. 
Local Average Treatment Effect (LATE) 

Local Binary Convolution (LBC) 
We propose local binary convolution (LBC), an efficient alternative to convolutional layers in standard convolutional neural networks (CNN). The design principles of LBC are motivated by local binary patterns (LBP). The LBC layer comprises of a set of fixed sparse predefined binary convolutional filters that are not updated during the training process, a nonlinear activation function and a set of learnable linear weights. The linear weights combine the activated filter responses to approximate the corresponding activated filter responses of a standard convolutional layer. The LBC layer affords significant parameter savings, 9x to 169x in the number of learnable parameters compared to a standard convolutional layer. Furthermore, due to lower model complexity and sparse and binary nature of the weights also results in up to 9x to 169x savings in model size compared to a standard convolutional layer. We demonstrate both theoretically and experimentally that our local binary convolution layer is a good approximation of a standard convolutional layer. Empirically, CNNs with LBC layers, called local binary convolutional neural networks (LBCNN), reach stateoftheart performance on a range of visual datasets (MNIST, SVHN, CIFAR10, and a subset of ImageNet) while enjoying significant computational savings. 
Local Binary Pattern Network (LBPNet) 
Memory and computation efficient deep learning architec tures are crucial to continued proliferation of machine learning capabili ties to new platforms and systems. Binarization of operations in convo lutional neural networks has shown promising results in reducing model size and computing efficiency. In this paper, we tackle the problem us ing a strategy different from the existing literature by proposing local binary pattern networks or LBPNet, that is able to learn and perform binary operations in an endtoend fashion. LBPNet1 uses local binary comparisons and random projection in place of conventional convolu tion (or approximation of convolution) operations. These operations can be implemented efficiently on different platforms including direct hard ware implementation. We applied LBPNet and its variants on standard benchmarks. The results are promising across benchmarks while provid ing an important means to improve memory and speed efficiency that is particularly suited for small footprint devices and hardware accelerators. 
Local Coordinate Coding (LCC) 
Generative adversarial networks (GANs) aim to generate realistic data from some prior distribution (e.g., Gaussian noises). However, such prior distribution is often independent of real data and thus may lose semantic information (e.g., geometric structure or content in images) of data. In practice, the semantic information might be represented by some latent distribution learned from data, which, however, is hard to be used for sampling in GANs. In this paper, rather than sampling from the predefined prior distribution, we propose a Local Coordinate Coding (LCC) based sampling method to improve GANs. We derive a generalization bound for LCC based GANs and prove that a small dimensional input is sufficient to achieve good generalization. Extensive experiments on various realworld datasets demonstrate the effectiveness of the proposed method. 
Local DeepFeature Alignment (LDFA) 
This paper presents an unsupervised deeplearning framework named Local DeepFeature Alignment (LDFA) for dimension reduction. We construct neighbourhood for each data sample and learn a local Stacked Contractive Autoencoder (SCAE) from the neighbourhood to extract the local deep features. Next, we exploit an affine transformation to align the local deep features of each neighbourhood with the global features. Moreover, we derive an approach from LDFA to map explicitly a new data sample into the learned lowdimensional subspace. The advantage of the LDFA method is that it learns both local and global characteristics of the data sample set: the local SCAEs capture local characteristics contained in the data set, while the global alignment procedures encode the interdependencies between neighbourhoods into the final lowdimensional feature representations. Experimental results on data visualization, clustering and classification show that the LDFA method is competitive with several wellknown dimension reduction techniques, and exploiting locality in deep learning is a research topic worth further exploring. 
Local Differential Privacy  Realtime datadriven optimization and control problems over networks may require sensitive information of participating users to calculate solutions and decision variables, such as in traffic or energy systems. Adversaries with access to coordination signals may potentially decode information on individual users and put user privacy at risk. We develop \emph{local differential privacy}, which is a strong notion that guarantees user privacy regardless of any auxiliary information an adversary may have, for a larger family of convex distributed optimization problems. The mechanism allows agent to customize their own privacy level based on local needs and parameter sensitivities. We propose a general sampling based approach for determining sensitivity and derive analytical bounds for specific quadratic problems. We analyze inherent tradeoffs between privacy and suboptimality and propose allocation schemes to divide the maximum allowable noise, a \emph{privacy budget}, among all participating agents. Our algorithm is implemented to enable privacy in distributed optimal power flow for electric grids. 
Local Error Driven and Associative Biologically Realistic Algorithm (leabra) 
The algorithm Leabra (local error driven and associative biologically realistic algorithm) allows for the construction of artificial neural networks that are biologically realistic and balance supervised and unsupervised learning within a single framework. leabRa 
Local Expansion via Minimum One Norm (LEMON) 
We propose a novel approach for finding overlapping communities called LEMON (Local Expansion via Minimum One Norm). The algorithm finds the community by seeking a sparse vector in the span of the local spectra such that the seeds are in its support. We show that LEMON can achieve the highest detection accuracy among stateoftheart proposals. The running time depends on the size of the community rather than that of the entire graph. The algorithm is easy to implement, and is highly parallelizable. 
Local False Discovery Rate (LFDR) 
➚ “False Discovery Rate” LFDR.MLE 
Local Fisher Discriminant Analysis (LFDA) 
lfda 
Local Fusion Block  ➚ “Adaptive Weighted SuperResolution Network” 
Local Gradients Smoothing (LGS) 
Deep neural networks (DNNs) have shown vulnerability to adversarial attacks, i.e., carefully perturbed inputs designed to mislead the network at inference time. Recently introduced localized attacks, LaVAN and Adversarial patch, posed a new challenge to deep learning security by adding adversarial noise only within a specific region without affecting the salient objects in an image. Driven by the observation that such attacks introduce concentrated highfrequency changes at a particular image location, we have developed an effective method to estimate noise location in gradient domain and transform those high activation regions caused by adversarial noise in image domain while having minimal effect on the salient object that is important for correct classification. Our proposed Local Gradients Smoothing (LGS) scheme achieves this by regularizing gradients in the estimated noisy region before feeding the image to DNN for inference. We have shown the effectiveness of our method in comparison to other defense methods including JPEG compression, Total Variance Minimization (TVM) and Feature squeezing on ImageNet dataset. In addition, we systematically study the robustness of the proposed defense mechanism against Back Pass Differentiable Approximation (BPDA), a state of the art attack recently developed to break defenses that transform an input sample to minimize the adversarial effect. Compared to other defense mechanisms, LGS is by far the most resistant to BPDA in localized adversarial attack setting. 
LOcal Group Graphical Lasso Estimation (loggle) 
In this paper, we study timevarying graphical models based on data measured over a temporal grid. Such models are motivated by the needs to describe and understand evolving interacting relationships among a set of random variables in many real applications, for instance the study of how stocks interact with each other and how such interactions change over time. We propose a new model, LOcal Group Graphical Lasso Estimation (loggle), under the assumption that the graph topology changes gradually over time. Specifically, loggle uses a novel local grouplasso type penalty to efficiently incorporate information from neighboring time points and to impose structural smoothness of the graphs. We implement an ADMM based algorithm to fit the loggle model. This algorithm utilizes blockwise fast computation and pseudolikelihood approximation to improve computational efficiency. An R package loggle has also been developed. We evaluate the performance of loggle by simulation experiments. We also apply loggle to S&P 500 stock price data and demonstrate that loggle is able to reveal the interacting relationships among stocks and among industrial sectors in a time period that covers the recent global financial crisis. 
Local Interpretable ModelAgnostic Explanation (LIME) 
Machine learning is at the core of many recent advances in science and technology. With computers beating professionals in games like Go, many people have started asking if machines would also make for better drivers or even better doctors. In many applications of machine learning, users are asked to trust a model to help them make decisions. A doctor will certainly not operate on a patient simply because ‘the model said so.’ Even in lowerstakes situations, such as when choosing a movie to watch from Netflix, a certain measure of trust is required before we surrender hours of our time based on a model. Despite the fact that many machine learning models are black boxes, understanding the rationale behind the model’s predictions would certainly help users decide when to trust or not to trust their predictions. An example is shown in Figure 1, in which a model predicts that a certain patient has the flu. The prediction is then explained by an ‘explainer’ that highlights the symptoms that are most important to the model. With this information about the rationale behind the model, the doctor is now empowered to trust the modelor not. 
Local Lift Dependence  We study the RadonNikodym derivative $L \coloneqq d\mathbb{P}(X,Y)/d(\mathbb{P}(X) \times \mathbb{P}(Y))$, between the joint distribution of a random vector $(X,Y)$ and the product measure generated by their marginal distributions, as a scale of local dependence, which we call Local Lift Dependence. This notion of local dependence is extended for when this derivative is not defined, contemplating a large class of distributions which are of interest in statistics. This extension is based on the Hausdorff dimension of the support of the singular part of $\mathbb{P}(X,Y)$. We argue that $L$ is more general and suitable to study variable dependence than other specific local dependence quantifiers and global dependence coefficients, as the Mutual Information, which is the expectation of $\log L$. An outline of how this dependence scale may be useful in statistics and topics for future researches are presented. 
Local Linear Forest  Random forests are a powerful method for nonparametric regression, but are limited in their ability to fit smooth signals, and can show poor predictive performance in the presence of strong, smooth effects. Taking the perspective of random forests as an adaptive kernel method, we pair the forest kernel with a local linear regression adjustment to better capture smoothness. The resulting procedure, local linear forests, enables us to improve on asymptotic rates of convergence for random forests with smooth signals, and provides substantial gains in accuracy on both real and simulated data. 
Local Mahalanobis Distance Learning (LMDL) 
Distance metric learning is a successful way to enhance the performance of the nearest neighbor classifier. In most cases, however, the distribution of data does not obey a regular form and may change in different parts of the feature space. Regarding that, this paper proposes a novel local distance metric learning method, namely Local Mahalanobis Distance Learning (LMDL), in order to enhance the performance of the nearest neighbor classifier. LMDL considers the neighborhood influence and learns multiple distance metrics for a reduced set of input samples. The reduced set is called as prototypes which try to preserve local discriminative information as much as possible. The proposed LMDL can be kernelized very easily, which is significantly desirable in the case of highly nonlinear data. The quality as well as the efficiency of the proposed method assesses through a set of different experiments on various datasets and the obtained results show that LDML as well as the kernelized version is superior to the other related stateoftheart methods. 
Local Orthogonal Decomposition  Inverted file and asymmetric distance computation (IVFADC) have been successfully applied to approximate nearest neighbor search and subsequently maximum inner product search. In such a framework, vector quantization is used for coarse partitioning while product quantization is used for quantizing residuals. In the original IVFADC as well as all of its variants, after residuals are computed, the second production quantization step is completely independent of the first vector quantization step. In this work, we seek to exploit the connection between these two steps when we perform nonexhaustive search. More specifically, we decompose a residual vector locally into two orthogonal components and perform uniform quantization and multiscale quantization to each component respectively. The proposed method, called local orthogonal decomposition, combined with multiscale quantization consistently achieves higher recall than previous methods under the same bitrates. We conduct comprehensive experiments on large scale datasets as well as detailed ablation tests, demonstrating effectiveness of our method. 
Local Outlier Factor (LOF) 
In anomaly detection, the local outlier factor (LOF) is an algorithm proposed by Markus M. Breunig, HansPeter Kriegel, Raymond T. Ng and Jörg Sander in 2000 for finding anomalous data points by measuring the local deviation of a given data point with respect to its neighbours. LOF shares some concepts with DBSCAN and OPTICS such as the concepts of ‘core distance’ and ‘reachability distance’, which are used for local density estimation. Rlof 
Local Projections  In this paper, we propose a novel approach for outlier detection, called local projections, which is based on concepts of Local Outlier Factor (LOF) (Breunig et al., 2000) and RobPCA (Hubert et al., 2005). By using aspects of both methods, our algorithm is robust towards noise variables and is capable of performing outlier detection in multigroup situations. We are further not reliant on a specific underlying data distribution. For each observation of a dataset, we identify a local group of dense nearby observations, which we call a core, based on a modification of the knearest neighbours algorithm. By projecting the dataset onto the space spanned by those observations, two aspects are revealed. First, we can analyze the distance from an observation to the center of the core within the projection space in order to provide a measure of quality of description of the observation by the projection. Second, we consider the distance of the observation to the projection space in order to assess the suitability of the core for describing the outlyingness of the observation. These novel interpretations lead to a univariate measure of outlyingness based on aggregations over all local projections, which outperforms LOF and RobPCA as well as other popular methods like PCOut (Filzmoser et al., 2008) and subspacebased outlier detection (Kriegel et al., 2009) in our simulation setups. Experiments in the context of realword applications employing datasets of various dimensionality demonstrate the advantages of local projections. 
Local Regression (LOESS, LOWESS) 
LOESS and LOWESS (locally weighted scatterplot smoothing) are two strongly related nonparametric regression methods that combine multiple regression models in a knearestneighborbased metamodel. “LOESS” is a later generalization of LOWESS; although it is not a true initialism, it may be understood as standing for “LOcal regrESSion”. 
Local Reparameterization Network (LRNets) 
Recent breakthroughs in computer vision make use of large deep neural networks, utilizing the substantial speedup offered by GPUs. For applications running on limited hardware however, high precision realtime processing can still be a challenge. One approach to solve this problem is learning networks with binary or ternary weights, thus removing the need to calculate multiplications and significantly reduce memory size and access. In this work we introduce LRnets (Local reparameterization networks), a new method for training neural networks with discrete weights using stochastic parameters. We show how a simple modification to the local reparameterization trick, previously used to train Gaussian distributed weights, allows us to train discrete weights. We tested our method on MNIST, CIFAR10 and ImageNet, achieving stateoftheart results compared to previous binary and ternary models. 
Local Segmentation Network (LSN) 
Curve text or arbitrary shape text is very common in realworld scenarios. In this paper, we propose a novel framework with the local segmentation network (LSN) followed by the curve connection to detect text in horizontal, oriented and curved forms. The LSN is composed of two elements, i.e., proposal generation to get the horizontal rectangle proposals with high overlap with text and text segmentation to find the arbitrary shape text region within proposals. The curve connection is then designed to connect the local mask to the detection results. We conduct experiments using the proposed framework on two realworld curve text detection datasets and demonstrate the effectiveness over previous approaches. 
Local Shrunk Discriminant Analysis (LSDA) 
Dimensionality reduction is a crucial step for pattern recognition and data mining tasks to overcome the curse of dimensionality. Principal component analysis (PCA) is a traditional technique for unsupervised dimensionality reduction, which is often employed to seek a projection to best represent the data in a leastsquares sense, but if the original data is nonlinear structure, the performance of PCA will quickly drop. An supervised dimensionality reduction algorithm called Linear discriminant analysis (LDA) seeks for an embedding transformation, which can work well with Gaussian distribution data or singlemodal data, but for nonGaussian distribution data or multimodal data, it gives undesired results. What is worse, the dimension of LDA cannot be more than the number of classes. In order to solve these issues, Local shrunk discriminant analysis (LSDA) is proposed in this work to process the nonGaussian distribution data or multimodal data, which not only incorporate both the linear and nonlinear structures of original data, but also learn the pattern shrinking to make the data more flexible to fit the manifold structure. Further, LSDA has more strong generalization performance, whose objective function will become local LDA and traditional LDA when different extreme parameters are utilized respectively. What is more, a new efficient optimization algorithm is introduced to solve the nonconvex objective function with low computational cost. Compared with other related approaches, such as PCA, LDA and local LDA, the proposed method can derive a subspace which is more suitable for nonGaussian distribution and real data. Promising experimental results on different kinds of data sets demonstrate the effectiveness of the proposed approach. 
Local Spatial Attention (LSA) 
Directly learning features from the point cloud has become an active research direction in 3D understanding. Existing learningbased methods usually construct local regions from the point cloud and extract the corresponding features using shared MultiLayer Perceptron (MLP) and max pooling. However, most of these processes do not adequately take the spatial distribution of the point cloud into account, limiting the ability to perceive finegrained patterns. We design a novel Local Spatial Attention (LSA) module to adaptively generate attention maps according to the spatial distribution of local regions. The feature learning process which integrates with these attention maps can effectively capture the local geometric structure. We further propose the Spatial Feature Extractor (SFE), which constructs a branch architecture, to aggregate the spatial information with associated features in each layer of the network better.The experiments show that our network, named LSANet, can achieve on par or better performance than the stateoftheart methods when evaluating on the challenging benchmark datasets. The source code is available at https://…/LSANet. 
Locality Preserving Projection  With the advantage of low storage cost and high efficiency, hashing learning has received much attention in retrieval field. As multiple modal data representing a common object semantically are complementary, many works focus on learning unified binary codes. However, these works ignore the importance of manifold structre among data. In fact, it is still an interesting problem to directly preserve the local manifold structure among samples in hamming space. Since different modalities are isomerous, we adopt the concatenated feature of multiple modality feature to represent original object. In our framework, Locally Linear Embedding and Locality Preserving Projection are introduced to reconstruct the manifold structure of original space in the Hamming space. Besides, The L21norm regularization are imposed on the projection matrices to further exploit the discriminative features for different modalities simultaneously. Extensive experiments are performed to evaluate the proposed method, dubbed Unsupervised Concatenation Hashing (UCH), on the three publicly available datasets and the experimental results show the superior performance of UCH outperforming most of stateoftheart unsupervised hashing models. 
Locality Sensitive Hashing (LSH) 
Localitysensitive hashing (LSH) is a method of performing probabilistic dimension reduction of highdimensional data. The basic idea is to hash the input items so that similar items are mapped to the same buckets with high probability (the number of buckets being much smaller than the universe of possible input items). This is different from the conventional hash functions, such as those used in cryptography, as in the LSH case the goal is to maximize probability of ‘collision’ of similar items rather than avoid collisions. Note how localitysensitive hashing, in many ways, mirrors data clustering and Nearest neighbor search. http://…/LSH http://…descriptionoflocalitysensitivehashing 
Localization Recall Precision (LRP) 
Average precision (AP), the area under the recallprecision (RP) curve, is the standard performance measure for object detection. Despite its wide acceptance, it has a number of shortcomings, the most important of which are (i) the inability to distinguish very different RP curves, and (ii) the lack of directly measuring bounding box localization accuracy. In this paper, we propose ‘Localization Recall Precision (LRP) Error’, a new metric which we specifically designed for object detection. LRP Error is composed of three components related to localization, false negative (FN) rate and false positive (FP) rate. Based on LRP, we introduce the ‘Optimal LRP’, the minimum achievable LRP error representing the best achievable configuration of the detector in terms of recallprecision and the tightness of the boxes. In contrast to AP, which considers precisions over the entire recall domain, Optimal LRP determines the ‘best’ confidence score threshold for a class, which balances the tradeoff between localization and recallprecision. In our experiments, we show that, for stateoftheart object (SOTA) detectors, Optimal LRP provides richer and more discriminative information than AP. We also demonstrate that the best confidence score thresholds vary significantly among classes and detectors. Moreover, we present LRP results of a simple online video object detector which uses a SOTA still image object detector and show that the classspecific optimized thresholds increase the accuracy against the common approach of using a general threshold for all classes. We provide the source code that can compute LRP for the PASCAL VOC and MSCOCO datasets in https://…/LRP. Our source code can easily be adapted to other datasets as well. 
Localized Information Privacy (LIP) 
In this paper, localized information privacy (LIP) is proposed, as a new privacy definition, which allows statistical aggregation while protecting users’ privacy without relying on a trusted third party. The notion of contextawareness is incorporated in LIP by the introduction of priors, which enables the design of privacypreserving data aggregation with knowledge of priors. We show that LIP relaxes the Localized Differential Privacy (LDP) notion by explicitly modeling the adversary’s knowledge. However, it is stricter than $2\epsilon$LDP and $\epsilon$mutual information privacy. The incorporation of local priors allows LIP to achieve higher utility compared to other approaches. We then present an optimization framework for privacypreserving data aggregation, with the goal of minimizing the expected squared error while satisfying the LIP privacy constraints. Utilityprivacy tradeoffs are obtained under several models in closedform. We then validate our analysis by {numerical analysis} using both synthetic and realworld data. Results show that our LIP mechanism provides better utilityprivacy tradeoffs than LDP and when the prior is not uniformly distributed, the advantage of LIP is even more significant. 
Localized System Solver (locSolver) 
Column generation is often used to solve multicommodity flow problems. A program for column generation always includes a module that solves a linear equation. In this paper, we address three major issues in solving linear problem during column generation procedure which are (1) how to employ the sparse property of the coefficient matrix; (2) how to reduce the size of the coefficient matrix; and (3) how to reuse the solution to a similar equation. To this end, we first analyze the sparse property of coefficient matrix of linear equations and find that the matrices occurring in iteration are very sparse. Then, we present an algorithm locSolver (for localized system solver) for linear equations with sparse coefficient matrices and righthandsides. This algorithm can reduce the number of variables. After that, we present the algorithm incSolver (for incremental system solver) which utilizes similarity in the iterations of the program for a linear equation system. All three techniques can be used in column generation of multicommodity problems. Preliminary numerical experiments show that the incSolver is significantly faster than the existing algorithms. For example, random test cases show that incSolver is at least 37 times and up to 341 times faster than popular solver LAPACK. 
Locally Connected Layer  ➘ “LocallyConnected Spiking Neural Network” 
Locally Estimated Scatterplot Smoothing (LOESS) 
LOESS and LOWESS (locally weighted scatterplot smoothing) are two strongly related nonparametric regression methods that combine multiple regression models in a knearestneighborbased metamodel. “LOESS” is a later generalization of LOWESS; although it is not a true initialism, it may be understood as standing for “LOcal regrESSion”. 
Locally Linear Embedding (LLE) 
With the advantage of low storage cost and high efficiency, hashing learning has received much attention in retrieval field. As multiple modal data representing a common object semantically are complementary, many works focus on learning unified binary codes. However, these works ignore the importance of manifold structre among data. In fact, it is still an interesting problem to directly preserve the local manifold structure among samples in hamming space. Since different modalities are isomerous, we adopt the concatenated feature of multiple modality feature to represent original object. In our framework, Locally Linear Embedding and Locality Preserving Projection are introduced to reconstruct the manifold structure of original space in the Hamming space. Besides, The L21norm regularization are imposed on the projection matrices to further exploit the discriminative features for different modalities simultaneously. Extensive experiments are performed to evaluate the proposed method, dubbed Unsupervised Concatenation Hashing (UCH), on the three publicly available datasets and the experimental results show the superior performance of UCH outperforming most of stateoftheart unsupervised hashing models. 
Locally Selective Combination in Parallel Outlier Ensemble (LSCP) 
In unsupervised outlier ensembles, the absence of ground truth makes the combination of base detectors a challenging task. Specifically, existing parallel outlier ensembles lack a reliable way of selecting competent base detectors, affecting accuracy and stability, during model combination. In this paper, we propose a framework—called Locally Selective Combination in Parallel Outlier Ensembles (LSCP)—which addresses this issue by defining a local region around a test instance using the consensus of its nearest neighbors in randomly generated feature spaces. The topperforming base detectors in this local region are selected and combined as the model’s final output. Four variants of the LSCP framework are compared with six widely used combination algorithms for parallel ensembles. Experimental results demonstrate that one of these LSCP variants consistently outperforms baseline algorithms on the majority of eighteen realworld datasets. 
Locally Smoothed Neural Network (LSNN) 
Convolutional Neural Networks (CNN) and the locally connected layer are limited in capturing the importance and relations of different local receptive fields, which are often crucial for tasks such as face verification, visual question answering, and word sequence prediction. To tackle the issue, we propose a novel locally smoothed neural network (LSNN) in this paper. The main idea is to represent the weight matrix of the locally connected layer as the product of the kernel and the smoother, where the kernel is shared over different local receptive fields, and the smoother is for determining the importance and relations of different local receptive fields. Specifically, a multivariate Gaussian function is utilized to generate the smoother, for modeling the location relations among different local receptive fields. Furthermore, the content information can also be leveraged by setting the mean and precision of the Gaussian function according to the content. Experiments on some variant of MNIST clearly show our advantages over CNN and locally connected layer. 
LocallyConnected Spiking Neural Network (LCSNN) 
In recent years, Spiking Neural Networks (SNNs) have demonstrated great successes in completing various Machine Learning tasks. We introduce a method for learning image features by \textit{locally connected layers} in SNNs using spiketimingdependent plasticity (STDP) rule. In our approach, subnetworks compete via competitive inhibitory interactions to learn features from different locations of the input space. These \textit{LocallyConnected Spiking Neural Networks} (LCSNNs) manifest key topological features of the spatial interaction of biological neurons. We explore biologically inspired ngram classification approach allowing parallel processing over various patches of the the image space. We report the classification accuracy of simple twolayer LCSNNs on two image datasets, which match the stateofart performance and are the first results to date. LCSNNs have the advantage of fast convergence to a dataset representation, and they require fewer learnable parameters than other SNN approaches with unsupervised learning. Robustness tests demonstrate that LCSNNs exhibit graceful degradation of performance despite the random deletion of large amounts of synapses and neurons. 
LocateLinkVisualize (LocLinkVis) 
In this paper we present LocLinkVis (LocateLinkVisualize); a system which supports exploratory information access to a document collection based on georeferencing and visualization. It uses a gazetteer which contains representations of places ranging from countries to buildings, and that is used to recognize toponyms, disambiguate them into places, and to visualize the resulting spatial footprints. 
Location Determination Problem (LDP) 

Log Gaussian Cox Process Network  We generalize the log Gaussian Cox process (LGCP) framework to model multiple correlated point data jointly. The resulting log Gaussian Cox process network (LGCPN) considers the observations as realizations of multiple LGCPs, whose log intensities are given by linear combinations of latent functions drawn from Gaussian process priors. The coefficients of these linear combinations are also drawn from Gaussian processes and can incorporate additional dependencies a priori. We derive closedform expressions for the moments of the intensity functions in our model and use them to develop an efficient variational inference algorithm that is orders of magnitude faster than competing deterministic and stochastic approximations of multivariate LGCP and coregionalization models. Our approach outperforms the state of the art in jointly estimating multiple bovine tuberculosis incidents in Cornwall, UK, and multiple crime type intensities across New York city. 
LOGAN  We present LOGAN, a deep neural network aimed at learning generic shape transforms from unpaired domains. The network is trained on two sets of shapes, e.g., tables and chairs, but there is neither a pairing between shapes in the two domains to supervise the shape translation nor any pointwise correspondence between any shapes. Once trained, LOGAN takes a shape from one domain and transforms it into the other. Our network consists of an autoencoder to encode shapes from the two input domains into a common latent space, where the latent codes encode multiscale shape features in an overcomplete manner. The translator is based on a generative adversarial network (GAN), operating in the latent space, where an adversarial loss enforces crossdomain translation while a feature preservation loss ensures that the right shape features are preserved for a natural shape transform. We conduct various ablation studies to validate each of our key network designs and demonstrate superior capabilities in unpaired shape transforms on a variety of examples over baselines and stateoftheart approaches. We show that our network is able to learn what shape features to preserve during shape transforms, either local or nonlocal, whether content or style, etc., depending solely on the input domain pairs. 
Logarithmic Temporal Coding (LTC) 
A Spiking Neural Network (SNN) can be trained indirectly by first training an Artificial Neural Network (ANN) with the conventional backpropagation algorithm, then converting it into an SNN. The conventional ratecoding method for SNNs uses the number of spikes to encode magnitude of an activation value, and may be computationally inefficient due to the large number of spikes. Temporalcoding is typically more efficient by leveraging the timing of spikes to encode information. In this paper, we present Logarithmic Temporal Coding (LTC), where the number of spikes used to encode an activation value grows logarithmically with the activation value; and the accompanying ExponentiateandFire (EF) spiking neuron model, which only involves efficient bitshift and addition operations. Moreover, we improve the training process of ANN to compensate for approximation errors due to LTC. Experimental results indicate that the resulting SNN achieves competitive performance at significantly lower computational cost than related work. 
LogCanvas  In this demo paper, we introduce LogCanvas, a platform for user search history visualisation. Different from the existing visualisation tools, LogCanvas focuses on helping users reconstruct the semantic relationship among their search activities. LogCanvas segments a user’s search history into different sessions and generates a knowledge graph to represent the information exploration process in each session. A knowledge graph is composed of the most important concepts or entities discovered by each search query as well as their relationships. It thus captures the semantic relationship among the queries. LogCanvas offers a session timeline viewer and a snippets viewer to enable users to refind their previous search results efficiently. LogCanvas also provides a collaborative perspective to support a group of users in sharing search results and experience. 
Logic Learning Machine (LLM) 
Logic Learning Machine (LLM) is a machine learning method based on the generation of intelligible rules. LLM is an efficient implementation of the Switching Neural Network (SNN) paradigm, developed by Marco Muselli, Senior Researcher at the Italian National Research Council CNRIEIIT in Genoa. Logic Learning Machine is implemented in the Rulex suite. LLM has been employed in different fields, including orthopaedic patient classification, DNA microarray analysis and Clinical Decision Support System. 
Logical Positivism  Logical positivism and logical empiricism, which together formed neopositivism, was a movement in Western philosophy whose central thesis was verificationism, a theory of knowledge which asserted that only statements verifiable through empirical observation are cognitively meaningful. 
LogicallyCorrect Reinforcement Learning  We propose a novel Reinforcement Learning (RL) algorithm to synthesize policies for a Markov Decision Process (MDP), such that a linear time property is satisfied. We convert the property into a Limit Deterministic Buchi Automaton (LDBA), then construct a product MDP between the automaton and the original MDP. A reward function is then assigned to the states of the product automaton, according to accepting conditions of the LDBA. With this reward function, RL synthesizes a policy that satisfies the property: as such, the policy synthesis procedure is ‘constrained’ by the given specification. Additionally, we show that the RL procedure sets up an online value iteration method to calculate the maximum probability of satisfying the given property, at any given state of the MDP – a convergence proof for the procedure is provided. Finally, the performance of the algorithm is evaluated via a set of numerical examples. We observe an improvement of one order of magnitude in the number of iterations required for the synthesis compared to existing approaches. 
Logistic Circuits  This paper proposes a new classification model called logistic circuits. On MNIST and Fashion datasets, our learning algorithm outperforms neural networks that have an order of magnitude more parameters. Yet, logistic circuits have a distinct origin in symbolic AI, forming a discriminative counterpart to probabilisticlogical circuits such as ACs, SPNs, and PSDDs. We show that parameter learning for logistic circuits is convex optimization, and that a simple local search algorithm can induce strong model structures from data. 
Logistic Dynamic Weight Based Sine Cosine Search Algorithm (LDWSCSA) 
Particle swarm optimization (PSO) and Sine Cosine algorithm (SCA) have been widely used optimization methods but these methods have some disadvantages such as trapped local optimum point. In order to solve this problem and obtain more successful results than others, a novel logistic dynamic weight based sine cosine search algorithm (LDWSCSA) is presented in this paper. In the LDWSCSA method, logistic map is used as dynamic weight generator. Logistic map is one of the famous and widely used chaotic map in the literature. Search process of SCA is modified in the LDWSCSA. To evaluate performance of the LDWSCSA, the widely used numerical benchmark functions were utilized as test suite and other swarm optimization methods were used to obtain the comparison results. Superior performances of the LDWSCSA are proved success of this method. 
Logistic Regression  In statistics, logistic regression, or logit regression, is a type of probabilistic statistical classification model. It is also used to predict a binary response from a binary predictor, used for predicting the outcome of a categorical dependent variable (i.e., a class label) based on one or more predictor variables (features). That is, it is used in estimating empirical values of the parameters in a qualitative response model. The probabilities describing the possible outcomes of a single trial are modeled, as a function of the explanatory (predictor) variables, using a logistic function. Frequently (and subsequently in this article) “logistic regression” is used to refer specifically to the problem in which the dependent variable is binarythat is, the number of available categories is twowhile problems with more than two categories are referred to as multinomial logistic regression or, if the multiple categories are ordered, as ordered logistic regression. 
logit  The logit function is the inverse of the sigmoidal “logistic” function used in mathematics, especially in statistics. When the function’s parameter represents a probability p, the logit function gives the logodds, or the logarithm of the odds p/(1p). 
LogitBoost Autoregressive Networks  Multivariate binary distributions can be decomposed into products of univariate conditional distributions. Recently popular approaches have modeled these conditionals through neural networks with sophisticated weightsharing structures. It is shown that stateoftheart performance on several standard benchmark datasets can actually be achieved by training separate probability estimators for each dimension. In that case, model training can be trivially parallelized over data dimensions. On the other hand, complexity control has to be performed for each learned conditional distribution. Three possible methods are considered and experimentally compared. The estimator that is employed for each conditional is LogitBoost. Similarities and differences between the proposed approach and autoregressive models based on neural networks are discussed in detail. 
Logitron  Classification is the most important process in data analysis. However, due to the inherent nonconvex and nonsmooth structure of the zeroone loss function of the classification model, various convex surrogate loss functions such as hinge loss, squared hinge loss, logistic loss, and exponential loss are introduced. These loss functions have been used for decades in diverse classification models, such as SVM (support vector machine) with hinge loss, logistic regression with logistic loss, and Adaboost with exponential loss and so on. In this work, we present a Perceptronaugmented convex classification framework, {\it Logitron}. The loss function of it is a smoothly stitched function of the extended logistic loss with the famous Perceptron loss function. The extended logistic loss function is a parameterized function established based on the extended logarithmic function and the extended exponential function. The main advantage of the proposed Logitron classification model is that it shows the connection between SVM and logistic regression via polynomial parameterization of the loss function. In more details, depending on the choice of parameters, we have the HingeLogitron which has the generalized $k$th order hingeloss with an additional $k$th root stabilization function and the LogisticLogitron which has a logisticlike loss function with relatively large $k$. Interestingly, even $k=1$, HingeLogitron satisfies the classificationcalibration condition and shows reasonable classification performance with low computational cost. The numerical experiment in the linear classifier framework demonstrates that HingeLogitron with $k=4$ (the fourthorder SVM with the fourth root stabilization function) outperforms logistic regression, SVM, and other Logitron models in terms of classification accuracy. 
LogLikelihood  For many applications, the natural logarithm of the likelihood function, called the loglikelihood, is more convenient to work with. Because the logarithm is a monotonically increasing function, the logarithm of a function achieves its maximum value at the same points as the function itself, and hence the loglikelihood can be used in place of the likelihood in maximum likelihood estimation and related techniques. Finding the maximum of a function often involves taking the derivative of a function and solving for the parameter being maximized, and this is often easier when the function being maximized is a loglikelihood rather than the original likelihood function. For example, some likelihood functions are for the parameters that explain a collection of statistically independent observations. In such a situation, the likelihood function factors into a product of individual likelihood functions. The logarithm of this product is a sum of individual logarithms, and the derivative of a sum of terms is often easier to compute than the derivative of a product. In addition, several common distributions have likelihood functions that contain products of factors involving exponentiation. The logarithm of such a function is a sum of products, again easier to differentiate than the original function. In phylogenetics the loglikelihood ratio is sometimes termed support and the loglikelihood function support function. However, given the potential for confusion with the mathematical meaning of ‘support’ this terminology is rarely used outside this field. ➚ “Likelihood Function” 
LogLinear Model  A loglinear model is a mathematical model that takes the form of a function whose logarithm is a firstdegree polynomial function of the parameters of the model, which makes it possible to apply (possibly multivariate) linear regression. 
Logrank Test  In statistics, the logrank test is a hypothesis test to compare the survival distributions of two samples. It is a nonparametric test and appropriate to use when the data are right skewed and censored (technically, the censoring must be noninformative). It is widely used in clinical trials to establish the efficacy of a new treatment in comparison with a control treatment when the measurement is the time to event (such as the time from initial treatment to a heart attack). The test is sometimes called the MantelCox test, named after Nathan Mantel and David Cox. The logrank test can also be viewed as a timestratified CochranMantelHaenszel test. glrt 
LoIDE  Logicbased paradigms are nowadays widely used in many different fields, also thank to the availability of robust tools and systems that allow the development of realworld and industrial applications. In this work we present LoIDE, an advanced and modular webeditor for logicbased languages that also integrates with stateoftheart solvers. 
LOKI  Imitation learning (IL) consists of a set of tools that leverage expert demonstrations to quickly learn policies. However, if the expert is suboptimal, IL can yield policies with inferior performance compared to reinforcement learning (RL). In this paper, we aim to provide an algorithm that combines the best aspects of RL and IL. We accomplish this by formulating several popular RL and IL algorithms in a common mirror descent framework, showing that these algorithms can be viewed as a variation on a single approach. We then propose LOKI, a strategy for policy learning that first performs a small but random number of IL iterations before switching to a policy gradient RL method. We show that if the switching time is properly randomized, LOKI can learn to outperform a suboptimal expert and converge faster than running policy gradient from scratch. Finally, we evaluate the performance of LOKI experimentally in several simulated environments. 
Long and ShortTerm TimeSeries Network (LSTNet) 
Multivariate time series forecasting is an important machine learning problem across many domains, including predictions of solar plant energy output, electricity consumption, and traffic jam situation. Temporal data arise in these realworld applications often involves a mixture of longterm and shortterm patterns, for which traditional approaches such as Autoregressive models and Gaussian Process may fail. In this paper, we proposed a novel deep learning framework, namely Long and Shortterm Timeseries network (LSTNet), to address this open challenge. LSTNet uses the Convolution Neural Network (CNN) to extract shortterm local dependency patterns among variables, and the Recurrent Neural Network (RNN) to discover longterm patterns and trends. In our evaluation on realworld data with complex mixtures of repetitive patterns, LSTNet achieved significant performance improvements over that of several stateoftheart baseline methods. 
Long Short Term Memory (LSTM) 
Long short term memory (LSTM) is a recurrent neural network (RNN) architecture (an artificial neural network) published in 1997 by Sepp Hochreiter and Jürgen Schmidhuber. Like most RNNs, an LSTM network is universal in the sense that given enough network units it can compute anything a conventional computer can compute, provided it has the proper weight matrix, which may be viewed as its program. (Of course, finding such a weight matrix is more challenging with some problems than with others.) Unlike traditional RNNs, an LSTM network is wellsuited to learn from experience to classify, process and predict time series when there are very long time lags of unknown size between important events. This is one of the main reasons why LSTM outperforms alternative RNNs and Hidden Markov Models and other sequence learning methods in numerous applications. For example, LSTM achieved the best known results in unsegmented connected handwriting recognition, and in 2009 won the ICDAR handwriting competition. LSTM networks have also been used for automatic speech recognition, and were a major component of a network that recently achieved a record 17.7% phoneme error rate on the classic TIMIT natural speech dataset. 
Long Short Term Memory Convolutional Neural Network (LSTMCNN) 
We propose in this paper a combined model of Long Short Term Memory and Convolutional Neural Networks (LSTMCNN) that exploits word embeddings and positional embeddings for crosssentence nary relation extraction. The proposed model brings together the properties of both LSTMs and CNNs, to simultaneously exploit longrange sequential information and capture most informative features, essential for crosssentence nary relation extraction. The LSTMCNN model is evaluated on standard dataset on crosssentence nary relation extraction, where it significantly outperforms baselines such as CNNs, LSTMs and also a combined CNNLSTM model. The paper also shows that the LSTMCNN model outperforms the current stateoftheart methods on crosssentence nary relation extraction. 
Long ShortTerm Attention (LSTA) 
Egocentric activity recognition is one of the most challenging tasks in video analysis. It requires a finegrained discrimination of small objects and their manipulation. While some methods base on strong supervision and attention mechanisms, they are either annotation consuming or do not take spatiotemporal patterns into account. In this paper we propose LSTA as a mechanism to focus on features from spatial relevant parts while attention is being tracked smoothly across the video sequence. We demonstrate the effectiveness of LSTA on egocentric activity recognition with an endtoend trainable twostream architecture, achieving state of the art performance on four standard benchmarks. 
Long ShortTerm Memory RGCN (LRGCN) 
In this paper we use a timeevolving graph which consists of a sequence of graph snapshots over time to model many realworld networks. We study the path classification problem in a timeevolving graph, which has many applications in realworld scenarios, for example, predicting path failure in a telecommunication network and predicting path congestion in a traffic network in the near future. In order to capture the temporal dependency and graph structure dynamics, we design a novel deep neural network named Long ShortTerm Memory RGCN (LRGCN). LRGCN considers temporal dependency between timeadjacent graph snapshots as a special relation with memory, and uses relational GCN to jointly process both intratime and intertime relations. We also propose a new path representation method named \underline{s}elf\underline{a}ttentive \underline{p}ath \underline{e}mbedding (SAPE), to embed paths of arbitrary length into fixedlength vectors. Through experiments on a realworld telecommunication network and a traffic network in California, we demonstrate the superiority of LRGCN to other competing methods in path failure prediction, and prove the effectiveness of SAPE on path representation. 
Long Term Memory Network (LTM) 
Recurrent Neural Networks (RNN), Long ShortTerm Memory Networks (LSTM), and Memory Networks which contain memory are popularly used to learn patterns in sequential data. Sequential data has long sequences that hold relationships. RNN can handle long sequences but suffers from the vanishing and exploding gradient problems. While LSTM and other memory networks address this problem, they are not capable of handling long sequences (50 or more data points long sequence patterns). Language modelling requiring learning from longer sequences are affected by the need for more information in memory. This paper introduces Long Term Memory network (LTM), which can tackle the exploding and vanishing gradient problems and handles long sequences without forgetting. LTM is designed to scale data in the memory and gives a higher weight to the input in the sequence. LTM avoid overfitting by scaling the cell state after achieving the optimal results. The LTM is tested on Penn treebank dataset, and Text8 dataset and LTM achieves test perplexities of 83 and 82 respectively. 650 LTM cells achieved a test perplexity of 67 for Penn treebank, and 600 cells achieved a test perplexity of 77 for Text8. LTM achieves state of the art results by only using ten hidden LTM cells for both datasets. 
Longitudinal Study  A longitudinal survey is a correlational research study that involves repeated observations of the same variables over long periods of time – often many decades. It is a type of observational study. Longitudinal studies are often used in psychology to study developmental trends across the life span, and in sociology to study life events throughout lifetimes or generations. The reason for this is that, unlike crosssectional studies, in which different individuals with same characteristics are compared, longitudinal studies track the same people, and therefore the differences observed in those people are less likely to be the result of cultural differences across generations. Because of this benefit, longitudinal studies make observing changes more accurate, and they are applied in various other fields. In medicine, the design is used to uncover predictors of certain diseases. In advertising, the design is used to identify the changes that advertising has produced in the attitudes and behaviors of those within the target audience who have seen the advertising campaign. Because most longitudinal studies are observational, in the sense that they observe the state of the world without manipulating it, it has been argued that they may have less power to detect causal relationships than experiments. But because of the repeated observation at the individual level, they have more power than crosssectional observational studies, by virtue of being able to exclude timeinvariant unobserved individual differences, and by virtue of observing the temporal order of events. Some of the disadvantages of longitudinal study include the fact that they take a lot of time and are very expensive. Therefore, they are not very convenient. Longitudinal studies allow social scientists to distinguish short from longterm phenomena, such as poverty. If the poverty rate is 10% at a point in time, this may mean that 10% of the population are always poor, or that the whole population experiences poverty for 10% of the time. It is impossible to conclude which of these possibilities is the case using oneoff crosssectional studies. Types of longitudinal studies include cohort studies and panel studies. Cohort studies sample a cohort, defined as a group experiencing some event (typically birth) in a selected time period, and studying them at intervals through time. Panel studies sample a crosssection, and survey it at (usually regular) intervals. A retrospective study is a longitudinal study that looks back in time. For instance, a researcher may look up the medical records of previous years to look for a trend. 
LongRange Dependency (LRD) 
Longrange dependency (LRD), also called long memory or longrange persistence, is a phenomenon that may arise in the analysis of spatial or time series data. It relates to the rate of decay of statistical dependence, with the implication that this decays more slowly than an exponential decay, typically a powerlike decay. LRD is often related to selfsimilar processes or fields. LRD has been used in various fields such as internet traffic modelling, econometrics, hydrology, linguistics and the earth sciences. Different mathematical definitions of LRD are used for different contexts and purposes. 
Longterm Episodic Memory Network (LEMN) 
Current generation of memoryaugmented neural networks has limited scalability as they cannot efficiently process data that are too large to fit in the external memory storage. One example of this is lifelong learning scenario where the model receives unlimited length of data stream as an input which contains vast majority of uninformative entries. We tackle this problem by proposing a memory network fit for longterm lifelong learning scenario, which we refer to as Longterm Episodic Memory Networks (LEMN), that features a RNNbased retention agent that learns to replace less important memory entries based on the retention probability generated on each entry that is learned to identify data instances of generic importance relative to other memory entries, as well as its historical importance. Such learning of retention agent allows our longterm episodic memory network to retain memory entries of generic importance for a given task. We validate our model on a pathfinding task as well as synthetic and real question answering tasks, on which our model achieves significant improvements over the memory augmented networks with rulebased memory scheduling as well as an RLbased baseline that does not consider relative or historical importance of the memory. 
LOok More than Once (LOMO) 
Previous scene text detection methods have progressed substantially over the past years. However, limited by the receptive field of CNNs and the simple representations like rectangle bounding box or quadrangle adopted to describe text, previous methods may fall short when dealing with more challenging text instances, such as extremely long text and arbitrarily shaped text. To address these two problems, we present a novel text detector namely LOMO, which localizes the text progressively for multiple times (or in other word, LOok More than Once). LOMO consists of a direct regressor (DR), an iterative refinement module (IRM) and a shape expression module (SEM). At first, text proposals in the form of quadrangle are generated by DR branch. Next, IRM progressively perceives the entire long text by iterative refinement based on the extracted feature blocks of preliminary proposals. Finally, a SEM is introduced to reconstruct more precise representation of irregular text by considering the geometry properties of text instance, including text region, text center line and border offsets. The stateoftheart results on several public benchmarks including ICDAR2017RCTW, SCUTCTW1500, TotalText, ICDAR2015 and ICDAR17MLT confirm the striking robustness and effectiveness of LOMO. 
LookupBased Convolutional Neural Network (LCNN) 
Porting state of the art deep learning algorithms to resource constrained compute platforms (e.g. VR, AR, wearables) is extremely challenging. We propose a fast, compact, and accurate model for convolutional neural networks that enables efficient learning and inference. We introduce LCNN, a lookupbased convolutional neural network that encodes convolutions by few lookups to a dictionary that is trained to cover the space of weights in CNNs. Training LCNN involves jointly learning a dictionary and a small set of linear combinations. The size of the dictionary naturally traces a spectrum of tradeoffs between efficiency and accuracy. Our experimental results on ImageNet challenge show that LCNN can offer 3.2x speedup while achieving 55.1% top1 accuracy using AlexNet architecture. Our fastest LCNN offers 37.6x speed up over AlexNet while maintaining 44.3% top1 accuracy. LCNN not only offers dramatic speed ups at inference, but it also enables efficient training. In this paper, we show the benefits of LCNN in fewshot learning and fewiteration learning, two crucial aspects of ondevice training of deep learning models. 
Lord’s Paradox  In statistics, Lord’s paradox raises the issue of when it is appropriate to control for baseline status. In three papers, Frederic Lord noted that different results obtain if researchers adjust for preexisting differences. The paradox was resolved by Paul Holland and Donald Rubin using the Rubin causal model. The most famous formulation of Lord’s paradox was in his 1967 paper and was phrased in terms of weight change over freshman year of college in two different dormitories because Lord did not want readers to assume that measurement error was responsible for the paradox. ‘A large university is interested in investigating the effects on the students of the diet provided in the university dining halls and any sex differences in these effects. Various types of data are gathered. In particular, the weight of each student at the time of his arrival in September and his weight the following June are recorded.’ (Lord 1967, p. 304) Resolving the Lord’s Paradox 
Lorenz Curve  In economics, the Lorenz curve is a graphical representation of the cumulative distribution function of the empirical probability distribution of wealth or income, and was developed by Max O. Lorenz in 1905 for representing inequality of the wealth distribution. The curve is a graph showing the proportion of overall income or wealth assumed by the bottom x% of the people, although this is not rigorously true for a finite population (see below). It is often used to represent income distribution, where it shows for the bottom x% of households, what percentage (y%) of the total income they have. The percentage of households is plotted on the xaxis, the percentage of income on the yaxis. It can also be used to show distribution of assets. In such use, many economists consider it to be a measure of social inequality. The concept is useful in describing inequality among the size of individuals in ecology and in studies of biodiversity, where the cumulative proportion of species is plotted against the cumulative proportion of individuals. It is also useful in business modeling: e.g., in consumer finance, to measure the actual percentage y% of delinquencies attributable to the x% of people with worst risk scores. 
Loss Data Analytics  Loss Data Analytics is an interactive, online, freely available text. The idea behind the name Loss Data Analytics is to integrate classical loss data models from applied probability with modern analytic tools. In particular, we seek to recognize that big data (including social media and usage based insurance) are here and high speed computation is readily available. The online version contains many interactive objects (quizzes, computer demonstrations, interactive graphs, video, and the like) to promote deeper learning. A subset of the book is available for offline reading in pdf and EPUB formats. The online text will be available in multiple languages to promote access to a worldwide audience. 
Loss Distributional Approach (LDA) 
While AMA does not specify the use of any particular modeling technique, one common approach taken in the banking industry is the Loss Distribution Approach (LDA). With LDA, a bank first segments operational losses into homogeneous segments, called unit of measure’s (UoMs). For each unit of measure, the bank then constructs a loss distribution that represents its expectation of total losses that can materialize in a oneyear horizon. Given that data sufficiency is a major challenge for the industry, annual loss distribution cannot be built directly using annual loss figures. Instead, a bank will develop a frequency distribution that describes the number of loss events in a given year, and a severity distribution that describes the loss amount of a single loss event. The frequency and severity distributions are assumed to be independent. The convolution of these two distributions then give rise to the (annual) loss distribution. 
Loss Function  In mathematical optimization, statistics, decision theory and machine learning, a loss function or cost function is a function that maps an event or values of one or more variables onto a real number intuitively representing some ‘cost’ associated with the event. An optimization problem seeks to minimize a loss function. An objective function is either a loss function or its negative (sometimes called a reward function or a utility function), in which case it is to be maximized. In statistics, typically a loss function is used for parameter estimation, and the event in question is some function of the difference between estimated and true values for an instance of data. The concept, as old as Laplace, was reintroduced in statistics by Abraham Wald in the middle of the 20th Century. In the context of economics, for example, this is usually economic cost or regret. In classification, it is the penalty for an incorrect classification of an example. In actuarial science, it is used in an insurance context to model benefits paid over premiums, particularly since the works of Harald CramĆ©r in the 1920s. In optimal control the loss is the penalty for failing to achieve a desired value. In financial risk management the function is precisely mapped to a monetary loss. 
Loss Rank Mining (LRM) 
Modern object detectors usually suffer from low accuracy issues, as foregrounds always drown in tons of backgrounds and become hard examples during training. Compared with those proposalbased ones, realtime detectors are in far more serious trouble since they renounce the use of regionproposing stage which is used to filter a majority of backgrounds for achieving realtime rates. Though foregrounds as hard examples are in urgent need of being mined from tons of backgrounds, a considerable number of stateoftheart realtime detectors, like YOLO series, have yet to profit from existing hard example mining methods, as using these methods need detectors fit series of prerequisites. In this paper, we propose a general hard example mining method named Loss Rank Mining (LRM) to fill the gap. LRM is a general method for realtime detectors, as it utilizes the final feature map which exists in all realtime detectors to mine hard examples. By using LRM, some elements representing easy examples in final feature map are filtered and detectors are forced to concentrate on hard examples during training. Extensive experiments validate the effectiveness of our method. With our method, the improvements of YOLOv2 detector on autodriving related dataset KITTI and more general dataset PASCAL VOC are over 5% and 2% mAP, respectively. In addition, LRM is the first hard example mining strategy which could fit YOLOv2 perfectly and make it better applied in series of real scenarios where both realtime rates and accurate detection are strongly demanded. 
LossBased AdaBoost (LoAdaBoost) 
Medical data are valuable for improvement of health care, policy making and many other purposes. Vast amount of medical data are stored in different locations ,on many different devices and in different data silos. Sharing medical data among different sources is a big challenge due to regulatory , operational and security reasons. One potential solution is federated machine learning ,which a method that sends machine learning algorithms simultaneously to all data sources ,train models in each source and aggregates the learned models. This strategy allows utilization of valuable data without moving them. In this article, we proposed an adaptive boosting method that increases the efficiency of federated machine learning. Using intensive care unit data from hospital, we showed that LoAdaBoost federated learning outperformed baseline method and increased communication efficiency at negligible additional cost. 
Lost Box Recovery  Online signature verification is the process of verifying time series signature data which is generally obtained from the tabletbased device. Unlike offline signature images, the online signature image data consists of points that are arranged in a sequence of time. The aim of this research is to develop an improved approach to map the strokes in both test and reference signatures. Current methods make use of the Dynamic Time Warping (DTW) algorithm and its variant to segment them before comparing each of its data dimension. This paper presents a modified DTW algorithm with the proposed Lost Box Recovery Algorithm aims to improve the mapping performance for online signature verification 
Lotka’s Law  Lotka’s law, named after Alfred J. Lotka, is one of a variety of special applications of Zipf’s law. It describes the frequency of publication by authors in any given field. It states that the number of authors making n contributions is about 1/n^{a} of those making one contribution, where a nearly always equals two. More plainly, the number of authors publishing a certain number of articles is a fixed ratio to the number of authors publishing a single article. As the number of articles published increases, authors producing that many publications become less frequent. There are 1/4 as many authors publishing two articles within a specified time period as there are singlepublication authors, 1/9 as many publishing three articles, 1/16 as many publishing four articles, etc. Though the law itself covers many disciplines, the actual ratios involved (as a function of ‘a’) are very disciplinespecific. LotkasLaw 
Lottery Ticket Hypothesis  Recent work on neural network pruning indicates that, at training time, neural networks need to be significantly larger in size than is necessary to represent the eventual functions that they learn. This paper articulates a new hypothesis to explain this phenomenon. This conjecture, which we term the ‘lottery ticket hypothesis,’ proposes that successful training depends on lucky random initialization of a smaller subcomponent of the network. Larger networks have more of these ‘lottery tickets,’ meaning they are more likely to luck out with a subcomponent initialized in a configuration amenable to successful optimization. This paper conducts a series of experiments with XOR and MNIST that support the lottery ticket hypothesis. In particular, we identify these fortuitouslyinitialized subcomponents by pruning lowmagnitude weights from trained networks. We then demonstrate that these subcomponents can be successfully retrained in isolation so long as the subnetworks are given the same initializations as they had at the beginning of the training process. Initialized as such, these small networks reliably converge successfully, often faster than the original network at the same level of accuracy. However, when these subcomponents are randomly reinitialized or rearranged, they perform worse than the original network. In other words, large networks that train successfully contain small subnetworks with initializations conducive to optimization. The lottery ticket hypothesis and its connection to pruning are a step toward developing architectures, initializations, and training strategies that make it possible to solve the same problems with much smaller networks. 
Louvain Method  Our method, that we call Louvain Method (because, even though the coauthors now hold positions in Paris, London and Louvain, the method was devised when they all were in Louvain), outperforms other methods in terms of computation time, which allows us to analyze networks of unprecedented size (e.g. the analysis of a typical network of 2 million nodes only takes 2 minutes). The Louvain method has also been to shown to be very accurate by focusing on adhoc networks with known community structure. Moreover, due to its hierarchical structure, which is reminiscent of renormalization methods, it allows to look at communities at different resolutions. ➚ “Community Detection” 
Louvain Modularity  The Louvain Method for community detection is a method to extract communities from large networks created by Vincent Blondel. The method is a greedy optimization method that appears to run in time O(n log n). http://…/0803.0476v2.pdf 
Lovasz Convolutional Network (LCN) 
Semisupervised learning on graph structured data has received significant attention with the recent introduction of graph convolution networks (GCN). While traditional methods have focused on optimizing a loss augmented with Laplacian regularization framework, GCNs perform an implicit Laplacian type regularization to capture local graph structure. In this work, we propose Lovasz convolutional network (LCNs) which are capable of incorporating global graph properties. LCNs achieve this by utilizing Lovasz’s orthonormal embeddings of the nodes. We analyse local and global properties of graphs and demonstrate settings where LCNs tend to work better than GCNs. We validate the proposed method on standard random graph models such as stochastic block models (SBM) and certain community structure based graphs where LCNs outperform GCNs and learn more intuitive embeddings. We also perform extensive binary and multiclass classification experiments on real world datasets to demonstrate LCN’s effectiveness. In addition to simple graphs, we also demonstrate the use of LCNs on hypergraphs by identifying settings where they are expected to work better than GCNs. 
Low Algebraic Dimension Matrix Completion (LADMC) 
In the low rank matrix completion (LRMC) problem, the low rank assumption means that the columns (or rows) of the matrix to be completed are points on a lowdimensional linear algebraic variety. This paper extends this thinking to cases where the columns are points on a lowdimensional nonlinear algebraic variety, a problem we call Low Algebraic Dimension Matrix Completion (LADMC). Matrices whose columns belong to a union of subspaces (UoS) are an important special case. We propose a LADMC algorithm that leverages existing LRMC methods on a tensorized representation of the data. For example, a secondorder tensorization representation is formed by taking the outer product of each column with itself, and we consider higher order tensorizations as well. This approach will succeed in many cases where traditional LRMC is guaranteed to fail because the data are lowrank in the tensorized representation but not in the original representation. We also provide a formal mathematical justification for the success of our method. In particular, we show bounds of the rank of these data in the tensorized representation, and we prove sampling requirements to guarantee uniqueness of the solution. Interestingly, the sampling requirements of our LADMC algorithm nearly match the information theoretic lower bounds for matrix completion under a UoS model. We also provide experimental results showing that the new approach significantly outperforms existing stateoftheart methods for matrix completion in many situations. 
Low Complexity Neural Network (LCNN) 
Modern neural network architectures for largescale learning tasks have substantially higher model complexities, which makes understanding, visualizing and training these architectures difficult. Recent contributions to deep learning techniques have focused on architectural modifications to improve parameter efficiency and performance. In this paper, we derive a continuous and differentiable error functional for a neural network that minimizes its empirical error as well as a measure of the model complexity. The latter measure is obtained by deriving a differentiable upper bound on the VapnikChervonenkis (VC) dimension of the classifier layer of a class of deep networks. Using standard backpropagation, we realize a training rule that tries to minimize the error on training samples, while improving generalization by keeping the model complexity low. We demonstrate the effectiveness of our formulation (the Low Complexity Neural Network – LCNN) across several deep learning algorithms, and a variety of large benchmark datasets. We show that hidden layer neurons in the resultant networks learn features that are crisp, and in the case of image datasets, quantitatively sharper. Our proposed approach yields benefits across a wide range of architectures, in comparison to and in conjunction with methods such as Dropout and Batch Normalization, and our results strongly suggest that deep learning techniques can benefit from model complexity control methods such as the LCNN learning rule. 
Low Dimensional Manifold Regularized Neural Network (LDMNet) 
Deep neural networks have proved very successful on archetypal tasks for which large training sets are available, but when the training data are scarce, their performance suffers from overfitting. Many existing methods of reducing overfitting are dataindependent, and their efficacy is often limited when the training set is very small. Datadependent regularizations are mostly motivated by the observation that data of interest lie close to a manifold, which is typically hard to parametrize explicitly and often requires human input of tangent vectors. These methods typically only focus on the geometry of the input data, and do not necessarily encourage the networks to produce geometrically meaningful features. To resolve this, we propose a new framework, the LowDimensionalManifoldregularized neural Network (LDMNet), which incorporates a feature regularization method that focuses on the geometry of both the input data and the output features. In LDMNet, we regularize the network by encouraging the combination of the input data and the output features to sample a collection of low dimensional manifolds, which are searched efficiently without explicit parametrization. To achieve this, we directly use the manifold dimension as a regularization term in a variational functional. The resulting EulerLagrange equation is a LaplaceBeltrami equation over a point cloud, which is solved by the point integral method without increasing the computational complexity. We demonstrate two benefits of LDMNet in the experiments. First, we show that LDMNet significantly outperforms widelyused network regularizers such as weight decay and DropOut. Second, we show that LDMNet can be designed to extract common features of an object imaged via different modalities, which proves to be very useful in realworld applications such as crossspectral face recognition. 
Low Power WideArea Network MAC Protocol (LPWAMAC) 
LowPower WideArea Networks (LPWANs) are being successfully used for the monitoring of largescale systems that are delaytolerant and which have lowbandwidth requirements. The next step would be instrumenting these for the control of CyberPhysical Systems (CPSs) distributed over large areas which require more bandwidth, bounded delays and higher reliability or at least more rigorous guarantees therein. This paper presents LPWAMAC, a novel Low Power WideArea network MAC protocol, that ensures bounded endtoend delays, high channel utility and supports many of the different traffic patterns and datarates typical of CPS. 
Lowest Posterior Loss (LPL) 
This paper defines intrinsic credible regions, a method to produce objective Bayesian credible regions which only depends on the assumed model and the available data. Lowest posterior loss (LPL) regions are defined as Bayesian credible regions which contain values of minimum posterior expected loss: they depend both on the loss function and on the prior specification. An invariant, informationtheory based loss function, the intrinsic discrepancy is argued to be appropriate for scientific communication. Intrinsic credible regions are the lowest posterior loss regions with respect to the intrinsic discrepancy loss and the appropriate reference prior. The proposed procedure is completely general, and it is invariant under both reparametrization and marginalization. The exact derivation of intrinsic credible regions often requires numerical integration, but good analytical approximations are provided. Special attention is given to onedimensional intrinsic credible intervals; their coverage properties show that they are always approximate (and sometimes exact) frequentist confidence intervals. 
Lowest Posterior Loss Interval (LPLI) 
The Lowest Posterior Loss (LPL) interval (Bernardo, 2005), or LPLI, is a probability interval based on intrinsic discrepancy loss between prior and posterior distributions. The expected posterior loss is the loss associated with using a particular value theta[i] in theta of the parameter as the unknown true value of theta (Bernardo, 2005). Parameter values with smaller expected posterior loss should always be preferred. The LPL interval includes a region in which all parameter values have smaller expected posterior loss than those outside the region. Although any loss function could be used, the loss function should be invariant under reparameterization. Any intrinsic loss function is invariant under reparameterization, but not necessarily invariant under onetoone transformations of data x. When a loss function is also invariant under onetoone transformations, it is usually also invariant when reduced to a sufficient statistic. Only an intrinsic loss function that is invariant when reduced to a sufficient statistic should be considered. The intrinsic discrepancy loss is easily a superior loss function to the overused quadratic loss function, and is more appropriate than other popular measures, such as Hellinger distance, KullbackLeibler divergence (KLD), and Jeffreys logarithmic divergence. The intrinsic discrepancy loss is also an informationtheory related divergence measure. Intrinsic discrepancy loss is a symmetric, nonnegative loss function, and is a continuous, convex function. Intrinsic discrepancy loss was introduced by Bernardo and Rueda (2002) in a different context: hypothesis testing. Formally, it is: delta f(p[2],p[1]) = min[kappa(p[2]  p[1]), kappa(p[1]  p[2])] where delta is the discrepancy, kappa is the KLD, and p[1] and p[2] are the probability distributions. The intrinsic discrepancy loss is the loss function, and the expected posterior loss is the mean of the directed divergences. The LPL interval is also called an intrinsic credible interval or intrinsic probability interval, and the area inside the interval is often called an intrinsic credible region or intrinsic probability region. In practice, whether a reference prior or weakly informative prior (WIP) is used, the LPL interval is usually very close to the HPD interval, though the posterior losses may be noticeably different. If LPL used a zeroone loss function, then the HPD interval would be produced. An advantage of the LPL interval over HPD interval (see p.interval) is that the LPL interval is invariant to reparameterization. This is due to the invariant reparameterization property of reference priors. The quantilebased probability interval is also invariant to reparameterization. The LPL interval enjoys the same advantage as the HPD interval does over the quantilebased probability interval: it does not produce equal tails when inappropriate. Compared with probability intervals, the LPL interval is slightly less convenient to calculate. Although the prior distribution is specified within the Model specification function, the user must specify it for the LPL.interval function as well. A comparison of the quantilebased probability interval, HPD interval, and LPL interval is available here: http://…/credible. 
LowLevel, FirstOrder Probabilistic Programming Language (LFPPL) 
We develop a new Lowlevel, Firstorder Probabilistic Programming Language (LFPPL) suited for models containing a mix of continuous, discrete, and/or piecewisecontinuous variables. The key success of this language and its compilation scheme is in its ability to automatically distinguish parameters the density function is discontinuous with respect to, while further providing runtime checks for boundary crossings. This enables the introduction of new inference engines that are able to exploit gradient information, while remaining efficient for models which are not everywhere differentiable. We demonstrate this ability by incorporating a discontinuous Hamiltonian Monte Carlo (DHMC) inference engine that is able to deliver automated and efficient inference for nondifferentiable models. Our system is backed up by a mathematical formalism that ensures that any model expressed in this language has a density with measure zero discontinuities to maintain the validity of the inference engine. 
Lowpass Recurrent Neural Network  Reinforcement learning (RL) agents performing complex tasks must be able to remember observations and actions across sizable time intervals. This is especially true during the initial learning stages, when exploratory behaviour can increase the delay between specific actions and their effects. Many new or popular approaches for learning these distant correlations employ backpropagation through time (BPTT), but this technique requires storing observation traces long enough to span the interval between cause and effect. Besides memory demands, learning dynamics like vanishing gradients and slow convergence due to infrequent weight updates can reduce BPTT’s practicality; meanwhile, although online recurrent network learning is a developing topic, most approaches are not efficient enough to use as replacements. We propose a simple, effective memory strategy that can extend the window over which BPTT can learn without requiring longer traces. We explore this approach empirically on a few tasks and discuss its implications. 
LowRank Discriminative Least Squares Regression Model (LRDLSR) 
Latest least squares regression (LSR) methods mainly try to learn slack regression targets to replace strict zeroone labels. However, the difference of intraclass targets can also be highlighted when enlarging the distance between different classes, and roughly persuing relaxed targets may lead to the problem of overfitting. To solve above problems, we propose a lowrank discriminative least squares regression model (LRDLSR) for multiclass image classification. Specifically, LRDLSR classwisely imposes lowrank constraint on the intraclass regression targets to encourage its compactness and similarity. Moreover, LRDLSR introduces an additional regularization term on the learned targets to avoid the problem of overfitting. These two improvements are helpful to learn a more discriminative projection for regression and thus achieving better classification performance. Experimental results over a range of image databases demonstrate the effectiveness of the proposed LRDLSR method. 
LowRank Kernel Subspace Clustering  Most stateoftheart subspace clustering methods only work with linear (or affine) subspaces. In this paper, we present a kernel subspace clustering method that can handle nonlinear models. While an arbitrary kernel can nonlinearly map data into highdimensional Hilbert feature space, the data in the resulting feature space are very unlikely to have the desired subspace structures. By contrast, we propose to learn a lowrank kernel mapping, with which the mapped data in feature space are not only lowrank but also selfexpressive, such that the lowdimensional subspace structures are present and manifested in the highdimensional feature space. We have evaluated the proposed method extensively on both motion segmentation and image clustering benchmarks, and obtained superior results, outperforming the kernel subspace clustering method that uses standard kernels~\cite{patel2014kernel} and other stateoftheart linear subspace clustering methods. 
LowRank Principal Eigenmatrix Analysis  Sparse PCA is a widely used technique for highdimensional data analysis. In this paper, we propose a new method called lowrank principal eigenmatrix analysis. Different from sparse PCA, the dominant eigenvectors are allowed to be dense but are assumed to have a lowrank structure when matricized appropriately. Such a structure arises naturally in several practical cases: Indeed the top eigenvector of a circulant matrix, when matricized appropriately is a rank1 matrix. We propose a matricized ranktruncated power method that could be efficiently implemented and establish its computational and statistical properties. Extensive experiments on several synthetic data sets demonstrate the competitive empirical performance of our method. 
LowRankFunctional (LRF) 
This preliminary note presents a heuristic for determining rank constrained solutions to linear matrix equations (LME). The method proposed here is based on minimizing a nonconvex quadratic functional, which will henceforth be termed as the \textit{LowRankFunctional} (LRF). Although this method lacks a formal proof/comprehensive analysis, for example in terms of a probabilistic guarantee for converging to a solution, the proposed idea is intuitive and has been seen to perform well in simulations. To that end, many numerical examples are provided to corroborate the idea. 
LowShot Transfer Detector (LSTD) 
Recent advances in object detection are mainly driven by deep learning with largescale detection benchmarks. However, the fullyannotated training set is often limited for a target detection task, which may deteriorate the performance of deep detectors. To address this challenge, we propose a novel lowshot transfer detector (LSTD) in this paper, where we leverage rich sourcedomain knowledge to construct an effective targetdomain detector with very few training examples. The main contributions are described as follows. First, we design a flexible deep architecture of LSTD to alleviate transfer difficulties in lowshot detection. This architecture can integrate the advantages of both SSD and Faster RCNN in a unified deep framework. Second, we introduce a novel regularized transfer learning framework for lowshot detection, where the transfer knowledge (TK) and background depression (BD) regularizations are proposed to leverage object knowledge respectively from source and target domains, in order to further enhance finetuning with a few target images. Finally, we examine our LSTD on a number of challenging lowshot detection experiments, where LSTD outperforms other stateoftheart approaches. The results demonstrate that LSTD is a preferable deep detector for lowshot scenarios. 
Lp Space  In mathematics, the Lp spaces are function spaces defined using a natural generalization of the pnorm for finitedimensional vector spaces. They are sometimes called Lebesgue spaces, named after Henri Lebesgue (Dunford & Schwartz 1958, III.3), although according to the Bourbaki group (Bourbaki 1987) they were first introduced by Frigyes Riesz (Riesz 1910). Lp spaces form an important class of Banach spaces in functional analysis, and of topological vector spaces. Lebesgue spaces have applications in physics, statistics, finance, engineering, and other disciplines. 
LP3DCNN  Traditional 3D Convolutional Neural Networks (CNNs) are computationally expensive, memory intensive, prone to overfit, and most importantly, there is a need to improve their feature learning capabilities. To address these issues, we propose Rectified Local Phase Volume (ReLPV) block, an efficient alternative to the standard 3D convolutional layer. The ReLPV block extracts the phase in a 3D local neighborhood (e.g., 3x3x3) of each position of the input map to obtain the feature maps. The phase is extracted by computing 3D Short Term Fourier Transform (STFT) at multiple fixed low frequency points in the 3D local neighborhood of each position. These feature maps at different frequency points are then linearly combined after passing them through an activation function. The ReLPV block provides significant parameter savings of at least, 3^3 to 13^3 times compared to the standard 3D convolutional layer with the filter sizes 3x3x3 to 13x13x13, respectively. We show that the feature learning capabilities of the ReLPV block are significantly better than the standard 3D convolutional layer. Furthermore, it produces consistently better results across different 3D data representations. We achieve stateoftheart accuracy on the volumetric ModelNet10 and ModelNet40 datasets while utilizing only 11% parameters of the current stateoftheart. We also improve the stateoftheart on the UCF101 split1 action recognition dataset by 5.68% (when trained from scratch) while using only 15% of the parameters of the stateoftheart. The project webpage is available at https://…/home. 
LPDNN  Deep Learning is moving to edge devices, ushering in a new age of distributed Artificial Intelligence (AI). The high demand of computational resources required by deep neural networks may be alleviated by approximate computing techniques, and most notably reducedprecision arithmetic with coarsely quantized numerical representations. In this context, Bonseyes comes in as an initiative to enable stakeholders to bring AI to lowpower and autonomous environments such as: Automotive, Medical Healthcare and Consumer Electronics. To achieve this, we introduce LPDNN, a framework for optimized deployment of Deep Neural Networks on heterogeneous embedded devices. In this work, we detail the quantization engine that is integrated in LPDNN. The engine depends on a finegrained workflow which enables a Neural Network Design Exploration and a sensitivity analysis of each layer for quantization. We demonstrate the engine with a case study on Alexnet and VGG16 for three different techniques for direct quantization: standard fixedpoint, dynamic fixedpoint and kmeans clustering, and demonstrate the potential of the latter. We argue that using a Gaussian quantizer with kmeans clustering can achieve better performance than linear quantizers. Without retraining, we achieve over 55.64\% saving for weights’ storage and 69.17\% for runtime memory accesses with less than 1\% drop in top5 accuracy in Imagenet. 
LRGLM  Due to the ease of modern data collection, applied statisticians often have access to a large set of covariates that they wish to relate to some observed outcome. Generalized linear models (GLMs) offer a particularly interpretable framework for such an analysis. In these highdimensional problems, the number of covariates is often large relative to the number of observations, so we face nontrivial inferential uncertainty; a Bayesian approach allows coherent quantification of this uncertainty. Unfortunately, existing methods for Bayesian inference in GLMs require running times roughly cubic in parameter dimension, and so are limited to settings with at most tens of thousand parameters. We propose to reduce time and memory costs with a lowrank approximation of the data in an approach we call LRGLM. When used with the Laplace approximation or Markov chain Monte Carlo, LRGLM provides a full Bayesian posterior approximation and admits running times reduced by a full factor of the parameter dimension. We rigorously establish the quality of our approximation and show how the choice of rank allows a tunable computationalstatistical tradeoff. Experiments support our theory and demonstrate the efficacy of LRGLM on real largescale datasets. 
LSANet  Directly learning features from the point cloud has become an active research direction in 3D understanding. Existing learningbased methods usually construct local regions from the point cloud and extract the corresponding features using shared MultiLayer Perceptron (MLP) and max pooling. However, most of these processes do not adequately take the spatial distribution of the point cloud into account, limiting the ability to perceive finegrained patterns. We design a novel Local Spatial Attention (LSA) module to adaptively generate attention maps according to the spatial distribution of local regions. The feature learning process which integrates with these attention maps can effectively capture the local geometric structure. We further propose the Spatial Feature Extractor (SFE), which constructs a branch architecture, to aggregate the spatial information with associated features in each layer of the network better.The experiments show that our network, named LSANet, can achieve on par or better performance than the stateoftheart methods when evaluating on the challenging benchmark datasets. The source code is available at https://…/LSANet. 
LSH Count  An important question that arises in the study of high dimensional vector representations learned from data is: given a set $\mathcal{D}$ of vectors and a query $q$, estimate the number of points within a specified distance threshold of $q$. We develop two estimators, LSH Count and MultiProbe Count that use locality sensitive hashing to preprocess the data to accurately and efficiently estimate the answers to such questions via importance sampling. A key innovation is the ability to maintain a small number of hash tables via preprocessing data structures and algorithms that sample from multiple buckets in each hash table. We give bounds on the space requirements and sample complexity of our schemes, and demonstrate their effectiveness in experiments on a standard word embedding dataset. 
LSTree  We study the problem of interpreting trained classification models in the setting of linguistic data sets. Leveraging a parse tree, we propose to assign leastsquares based importance scores to each word of an instance by exploiting syntactic constituency structure. We establish an axiomatic characterization of these importance scores by relating them to the Banzhaf value in coalitional game theory. Based on these importance scores, we develop a principled method for detecting and quantifying interactions between words in a sentence. We demonstrate that the proposed method can aid in interpretability and diagnostics for several widelyused language models. 
Lua  Lua is a powerful, efficient, lightweight, embeddable scripting language. It supports procedural programming, objectoriented programming, functional programming, datadriven programming, and data description. Lua combines simple procedural syntax with powerful data description constructs based on associative arrays and extensible semantics. Lua is dynamically typed, runs by interpreting bytecode with a registerbased virtual machine, and has automatic memory management with incremental garbage collection, making it ideal for configuration, scripting, and rapid prototyping. 
Ludwig  Ludwig is a toolbox that allows to train and test deep learning models without the need to write code. 
Luhn Algorithm  The Luhn algorithm or Luhn formula, also known as the ‘modulus 10’ or ‘mod 10’ algorithm, is a simple checksum formula used to validate a variety of identification numbers, such as credit card numbers, IMEI numbers, National Provider Identifier numbers in the United States, Canadian Social Insurance Numbers, Israel ID Numbers and Greek Social Security Numbers (????). It was created by IBM scientist Hans Peter Luhn and described in U.S. Patent No. 2,950,048, filed on January 6, 1954, and granted on August 23, 1960. The algorithm is in the public domain and is in wide use today. It is specified in ISO/IEC 78121. It is not intended to be a cryptographically secure hash function; it was designed to protect against accidental errors, not malicious attacks. Most credit cards and many government identification numbers use the algorithm as a simple method of distinguishing valid numbers from mistyped or otherwise incorrect numbers. checkLuhn 
Luigi  Luigi is a Python module that helps you build complex pipelines of batch jobs. It handles dependency resolution, workflow management, visualization etc. It also comes with Hadoop support built in. Luigi is an open source Pythonbased data framework for building data pipelines. Instead of using an XML/YAML configuration of some sort, all the jobs and their dependencies are written as Python programs. Because it’s Python, developers can backtrack to figure out exactly how data is processed. The framework makes it easier to build large data pipelines, with builtin checkpointing, failure recovery, parallel execution, command line integration, etc. Since it’s a Python program, any Python library assets can be reused. The Luigi framework itself is a couple of thousand lines, so it’s also easy to understand the entire mechanism. Facebook built a similar internal system called Dataswarm (Video), which allows developers to manage the entire data pipeline on Git + Python. While Luigi was originally invented for Spotify’s internal needs, companies such as Foursquare, Stripe, and Asana are using it in production. 
Lurking Variable  Lurking variables represent hidden information, and preclude a full understanding of phenomena of interest. Detection is usually based on serendipity — visual detection of unexplained, systematic variation. However, these approaches are doomed to fail if the lurking variables do not vary. 
Lyapunov Neural Network  Learning algorithms have shown considerable prowess in simulation by allowing robots to adapt to uncertain environments and improve their performance. However, such algorithms are rarely used in practice on safetycritical systems, since the learned policy typically does not yield any safety guarantees and thus the required exploration may cause physical harm to the robot or its environment. In this paper, we present a method to learn accurate safety certificates for nonlinear, closedloop dynamic systems. Specifically, we construct a neural network Lyapunov function and a training algorithm that adapts it to the shape of the largest safe region in the state space. The algorithm relies only on knowledge of inputs and outputs of the dynamics, rather than on any specific model structure. We demonstrate our method by learning the safe region of attraction for a simulated inverted pendulum. Furthermore, we discuss how our method can be used in safe learning algorithms together with statistical models of dynamic systems. 
LYRICS  In spite of the amazing results obtained by deep learning in many applications, a real intelligent behavior of an agent acting in a complex environment is likely to require some kind of higherlevel symbolic inference. Therefore, there is a clear need for the definition of a general and tight integration between lowlevel tasks, processing sensorial data that can be effectively elaborated using deep learning techniques, and the logic reasoning that allows humans to take decisions in complex environments. This paper presents LYRICS, a generic interface layer for AI, which is implemented in TersorFlow (TF). LYRICS provides an input language that allows to define arbitrary First Order Logic (FOL) background knowledge. The predicates and functions of the FOL knowledge can be bound to any TF computational graph, and the formulas are converted into a set of realvalued constraints, which participate to the overall optimization problem. This allows to learn the weights of the learners, under the constraints imposed by the prior knowledge. The framework is extremely general as it imposes no restrictions in terms of which models or knowledge can be integrated. In this paper, we show the generality of the approach showing some use cases of the presented language, including generative models, logic reasoning, model checking and supervised learning. 