Sonnet
It’s now nearly a year since DeepMind made the decision to switch the entire research organisation to using TensorFlow (TF). It’s proven to be a good choice – many of our models learn significantly faster, and the built-in features for distributed training have hugely simplified our code. Along the way, we found that the flexibility and adaptiveness of TF lends itself to building higher level frameworks for specific purposes, and we’ve written one for quickly building neural network modules with TF. We are actively developing this codebase, but what we have so far fits our research needs well, and we’re excited to announce that today we are open sourcing it. We call this framework Sonnet. …

Stochastic Proximal Langevin Algorithm (SPLA)
We propose a new algorithm—Stochastic Proximal Langevin Algorithm (SPLA)—for sampling from a log concave distribution. Our method is a generalization of the Langevin algorithm to potentials expressed as the sum of one stochastic smooth term and multiple stochastic nonsmooth terms. In each iteration, our splitting technique only requires access to a stochastic gradient of the smooth term and a stochastic proximal operator for each of the nonsmooth terms. We establish nonasymptotic sublinear and linear convergence rates under convexity and strong convexity of the smooth term, respectively, expressed in terms of the KL divergence and Wasserstein distance. We illustrate the efficiency of our sampling technique through numerical simulations on a Bayesian learning task. …

Smooth Imitation Learning (SIL)
In Smooth Imitation Learning for online sequence prediction is the goal is to train a policy that can smoothly imitate demonstrated behavior in a dynamic and continuous environment in response to online, sequential context input. …

Efficient Tuning of Lasso (ET-Lasso)
The L1 regularization (Lasso) has proven to be a versatile tool to select relevant features and estimate the model coefficients simultaneously. Despite its popularity, it is very challenging to guarantee the feature selection consistency of Lasso. One way to improve the feature selection consistency is to select an ideal tuning parameter. Traditional tuning criteria mainly focus on minimizing the estimated prediction error or maximizing the posterior model probability, such as cross-validation and BIC, which may either be time-consuming or fail to control the false discovery rate (FDR) when the number of features is extremely large. The other way is to introduce pseudo-features to learn the importance of the original ones. Recently, the Knockoff filter is proposed to control the FDR when performing feature selection. However, its performance is sensitive to the choice of the expected FDR threshold. Motivated by these ideas, we propose a new method using pseudo-features to obtain an ideal tuning parameter. In particular, we present the Efficient Tuning of Lasso (ET-Lasso) to separate active and inactive features by adding permuted features as pseudo-features in linear models. The pseudo-features are constructed to be inactive by nature, which can be used to obtain a cutoff to select the tuning parameter that separates active and inactive features. Experimental studies on both simulations and real-world data applications are provided to show that ET-Lasso can effectively and efficiently select active features under a wide range of different scenarios. …

Pyramid Mask Text Detector (PMTD)
Scene text detection, an essential step of scene text recognition system, is to locate text instances in natural scene images automatically. Some recent attempts benefiting from Mask R-CNN formulate scene text detection task as an instance segmentation problem and achieve remarkable performance. In this paper, we present a new Mask R-CNN based framework named Pyramid Mask Text Detector (PMTD) to handle the scene text detection. Instead of binary text mask generated by the existing Mask R-CNN based methods, our PMTD performs pixel-level regression under the guidance of location-aware supervision, yielding a more informative soft text mask for each text instance. As for the generation of text boxes, PMTD reinterprets the obtained 2D soft mask into 3D space and introduces a novel plane clustering algorithm to derive the optimal text box on the basis of 3D shape. Experiments on standard datasets demonstrate that the proposed PMTD brings consistent and noticeable gain and clearly outperforms state-of-the-art methods. Specifically, it achieves an F-measure of 80.13% on ICDAR 2017 MLT dataset. …

AFEL-REC
In this paper, we present preliminary results of AFEL-REC, a recommender system for social learning environments. AFEL-REC is build upon a scalable software architecture to provide recommendations of learning resources in near real-time. Furthermore, AFEL-REC can cope with any kind of data that is present in social learning environments such as resource metadata, user interactions or social tags. We provide a preliminary evaluation of three recommendation use cases implemented in AFEL-REC and we find that utilizing social data in form of tags is helpful for not only improving recommendation accuracy but also coverage. This paper should be valuable for both researchers and practitioners interested in providing resource recommendations in social learning environments. …

Superpixel-Based Patch (SuperPatch)
Superpixels have become very popular in many computer vision applications. Nevertheless, they remain underexploited since the superpixel decomposition may produce irregular and non stable segmentation results due to the dependency to the image content. In this paper, we first introduce a novel structure, a superpixel-based patch, called SuperPatch. The proposed structure, based on superpixel neighborhood, leads to a robust descriptor since spatial information is naturally included. The generalization of the PatchMatch method to SuperPatches, named SuperPatchMatch, is introduced. Finally, we propose a framework to perform fast segmentation and labeling from an image database, and demonstrate the potential of our approach since we outperform, in terms of computational cost and accuracy, the results of state-of-the-art methods on both face labeling and medical image segmentation. …

Multi-Resolution Flexible Irregular Time Series Network (Multi-FIT)
Missing values, irregularly collected samples, and multi-resolution signals commonly occur in multivariate time series data, making predictive tasks difficult. These challenges are especially prevalent in the healthcare domain, where patients’ vital signs and electronic records are collected at different frequencies and have occasionally missing information due to the imperfections in equipment or patient circumstances. Researchers have handled each of these issues differently, often handling missing data through mean value imputation and then using sequence models over the multivariate signals while ignoring the different resolution of signals. We propose a unified model named Multi-resolution Flexible Irregular Time series Network (Multi-FIT). The building block for Multi-FIT is the FIT network. The FIT network creates an informative dense representation at each time step using signal information such as last observed value, time difference since the last observed time stamp and overall mean for the signal. Vertical FIT (FIT-V) is a variant of FIT which also models the relationship between different temporal signals while creating the informative dense representations for the signal. The multi-FIT model uses multiple FIT networks for sets of signals with different resolutions, further facilitating the construction of flexible representations. Our model has three main contributions: a.) it does not impute values but rather creates informative representations to provide flexibility to the model for creating task-specific representations b.) it models the relationship between different signals in the form of support signals c.) it models different resolutions in parallel before merging them for the final prediction task. The FIT, FIT-V and Multi-FIT networks improve upon the state-of-the-art models for three predictive tasks, including the forecasting of patient survival. …

Bayesian Least-Squares Policy Iteration (BLSPI)
We introduce Bayesian least-squares policy iteration (BLSPI), an off-policy, model-free, policy iteration algorithm that uses the Bayesian least-squares temporal-difference (BLSTD) learning algorithm to evaluate policies. An online variant of BLSPI has been also proposed, called randomised BLSPI (RBLSPI), that improves its policy based on an incomplete policy evaluation step. In online setting, the exploration-exploitation dilemma should be addressed as we try to discover the optimal policy by using samples collected by ourselves. RBLSPI exploits the advantage of BLSTD to quantify our uncertainty about the value function. Inspired by Thompson sampling, RBLSPI first samples a value function from a posterior distribution over value functions, and then selects actions based on the sampled value function. The effectiveness and the exploration abilities of RBLSPI are demonstrated experimentally in several environments. …

Siamese Deep Forest (SDF)
A Siamese Deep Forest (SDF) is proposed in the paper. It is based on the Deep Forest or gcForest proposed by Zhou and Feng and can be viewed as a gcForest modification. It can be also regarded as an alternative to the well-known Siamese neural networks. The SDF uses a modified training set consisting of concatenated pairs of vectors. Moreover, it defines the class distributions in the deep forest as the weighted sum of the tree class probabilities such that the weights are determined in order to reduce distances between similar pairs and to increase them between dissimilar points. We show that the weights can be obtained by solving a quadratic optimization problem. The SDF aims to prevent overfitting which takes place in neural networks when only limited training data are available. The numerical experiments illustrate the proposed distance metric method. …

FlexNGIA
From virtual reality and telepresence, to augmented reality, holoportation, and remotely controlled robotics, these future network applications promise an unprecedented development for society, economics and culture by revolutionizing the way we live, learn, work and play. In order to deploy such futuristic applications and to cater to their performance requirements, recent trends stressed the need for the Tactile Internet, an Internet that, according to the International Telecommunication Union, combines ultra low latency with extremely high availability, reliability and security. Unfortunately, today’s Internet falls short when it comes to providing such stringent requirements due to several fundamental limitations in the design of the current network architecture and communication protocols. This brings the need to rethink the network architecture and protocols, and efficiently harness recent technological advances in terms of virtualization and network softwarization to design the Tactile Internet of the future. In this paper, we start by analyzing the characteristics and requirements of future networking applications. We then highlight the limitations of the traditional network architecture and protocols and their inability to cater to these requirements. Afterward, we put forward a novel network architecture adapted to the Tactile Internet called FlexNGIA, a Flexible Next-Generation Internet Architecture. We then describe some use-cases where we discuss the potential mechanisms and control loops that could be offered by FlexNGIA in order to ensure the required performance and reliability guarantees for future applications. Finally, we identify the key research challenges to further develop FlexNGIA towards a full-fledged architecture for the future Tactile Internet. …

Multi-Discriminator Generative Adversarial Network (MDGAN)
A recent technical breakthrough in the domain of machine learning is the discovery and the multiple applications of Generative Adversarial Networks (GANs). Those generative models are computationally demanding, as a GAN is composed of two deep neural networks, and because it trains on large datasets. A GAN is generally trained on a single server. In this paper, we address the problem of distributing GANs so that they are able to train over datasets that are spread on multiple workers. MD-GAN is exposed as the first solution for this problem: we propose a novel learning procedure for GANs so that they fit this distributed setup. We then compare the performance of MD-GAN to an adapted version of Federated Learning to GANs, using the MNIST and CIFAR10 datasets. MD-GAN exhibits a reduction by a factor of two of the learning complexity on each worker node, while providing better performances than federated learning on both datasets. We finally discuss the practical implications of distributing GANs. …

Diversity Index
A diversity index is a quantitative measure that reflects how many different types (such as species) there are in a dataset, and simultaneously takes into account how evenly the basic entities (such as individuals) are distributed among those types. The value of a diversity index increases both when the number of types increases and when evenness increases. For a given number of types, the value of a diversity index is maximized when all types are equally abundant. …

HornConcerto
Graph representations of large knowledge bases may comprise billions of edges. Usually built upon human-generated ontologies, several knowledge bases do not feature declared ontological rules and are far from being complete. Current rule mining approaches rely on schemata or store the graph in-memory, which can be unfeasible for large graphs. In this paper, we introduce HornConcerto, an algorithm to discover Horn clauses in large graphs without the need of a schema. Using a standard fact-based confidence score, we can mine close Horn rules having an arbitrary body size. We show that our method can outperform existing approaches in terms of runtime and memory consumption and mine high-quality rules for the link prediction task, achieving state-of-the-art results on a widely-used benchmark. Moreover, we find that rules alone can perform inference significantly faster than embedding-based methods and achieve accuracies on link prediction comparable to resource-demanding approaches such as Markov Logic Networks. …

Variational Collaborative Model (VCM)
Collaborative filtering (CF) has been successfully employed by many modern recommender systems. Conventional CF-based methods use the user-item interaction data as the sole information source to recommend items to users. However, CF-based methods are known for suffering from cold start problems and data sparsity problems. Hybrid models that utilize auxiliary information on top of interaction data have increasingly gained attention. A few ‘collaborative learning’-based models, which tightly bridges two heterogeneous learners through mutual regularization, are recently proposed for the hybrid recommendation. However, the ‘collaboration’ in the existing methods are actually asynchronous due to the alternative optimization of the two learners. Leveraging the recent advances in variational autoencoder~(VAE), we here propose a model consisting of two streams of mutual linked VAEs, named variational collaborative model (VCM). Unlike the mutual regularization used in previous works where two learners are optimized asynchronously, VCM enables a synchronous collaborative learning mechanism. Besides, the two stream VAEs setup allows VCM to fully leverages the Bayesian probabilistic representations in collaborative learning. Extensive experiments on three real-life datasets have shown that VCM outperforms several state-of-art methods. …

Human-eYe Perceptual Evaluation (HYPE)
Generative models often use human evaluations to determine and justify progress. Unfortunately, existing human evaluation methods are ad-hoc: there is currently no standardized, validated evaluation that: (1) measures perceptual fidelity, (2) is reliable, (3) separates models into clear rank order, and (4) ensures high-quality measurement without intractable cost. In response, we construct Human-eYe Perceptual Evaluation (HYPE), a human metric that is (1) grounded in psychophysics research in perception, (2) reliable across different sets of randomly sampled outputs from a model, (3) results in separable model performances, and (4) efficient in cost and time. We introduce two methods. The first, HYPE-Time, measures visual perception under adaptive time constraints to determine the minimum length of time (e.g., 250ms) that model output such as a generated face needs to be visible for people to distinguish it as real or fake. The second, HYPE-Infinity, measures human error rate on fake and real images with no time constraints, maintaining stability and drastically reducing time and cost. We test HYPE across four state-of-the-art generative adversarial networks (GANs) on unconditional image generation using two datasets, the popular CelebA and the newer higher-resolution FFHQ, and two sampling techniques of model outputs. By simulating HYPE’s evaluation multiple times, we demonstrate consistent ranking of different models, identifying StyleGAN with truncation trick sampling (27.6% HYPE-Infinity deception rate, with roughly one quarter of images being misclassified by humans) as superior to StyleGAN without truncation (19.0%) on FFHQ. See https://hype.stanford.edu for details. …

Subspace-CUSUM
We consider the sequential change-point detection problem of detecting changes that are characterized by a subspace structure. Such changes are frequent in high-dimensional streaming data altering the form of the corresponding covariance matrix. In this work we present a Subspace-CUSUM procedure and demonstrate its first-order asymptotic optimality properties for the case where the subspace structure is unknown and needs to be simultaneously estimated. To achieve this goal we develop a suitable analytical methodology that includes a proper parameter optimization for the proposed detection scheme. Numerical simulations corroborate our theoretical findings. …

Sliding Convolutional Attention Network (SCAN)
Scene text recognition has drawn great attentions in the community of computer vision and artificial intelligence due to its challenges and wide applications. State-of-the-art recurrent neural networks (RNN) based models map an input sequence to a variable length output sequence, but are usually applied in a black box manner and lack of transparency for further improvement, and the maintaining of the entire past hidden states prevents parallel computation in a sequence. In this paper, we investigate the intrinsic characteristics of text recognition, and inspired by human cognition mechanisms in reading texts, we propose a scene text recognition method with sliding convolutional attention network (SCAN). Similar to the eye movement during reading, the process of SCAN can be viewed as an alternation between saccades and visual fixations. Compared to the previous recurrent models, computations over all elements of SCAN can be fully parallelized during training. Experimental results on several challenging benchmarks, including the IIIT5k, SVT and ICDAR 2003/2013 datasets, demonstrate the superiority of SCAN over state-of-the-art methods in terms of both the model interpretability and performance. …

Coded TeraSort
We focus on sorting, which is the building block of many machine learning algorithms, and propose a novel distributed sorting algorithm, named Coded TeraSort, which substantially improves the execution time of the TeraSort benchmark in Hadoop MapReduce. The key idea of Coded TeraSort is to impose structured redundancy in data, in order to enable in-network coding opportunities that overcome the data shuffling bottleneck of TeraSort. We empirically evaluate the performance of CodedTeraSort algorithm on Amazon EC2 clusters, and demonstrate that it achieves 1.97x – 3.39x speedup, compared with TeraSort, for typical settings of interest. …

Multitask Soft Option Learning (MSOL)
We present Multitask Soft Option Learning (MSOL), a hierarchical multitask framework based on Planning as Inference. MSOL extends the concept of options, using separate variational posteriors for each task, regularized by a shared prior. This allows fine-tuning of options for new tasks without forgetting their learned policies, leading to faster training without reducing the expressiveness of the hierarchical policy. Additionally, MSOL avoids several instabilities during training in a multitask setting and provides a natural way to not only learn intra-option policies, but also their terminations. We demonstrate empirically that MSOL significantly outperforms both hierarchical and flat transfer-learning baselines in challenging multi-task environments. …

Robust Conditional GAN (RCGAN)
We study the problem of learning conditional generators from noisy labeled samples, where the labels are corrupted by random noise. A standard training of conditional GANs will not only produce samples with wrong labels, but also generate poor quality samples. We consider two scenarios, depending on whether the noise model is known or not. When the distribution of the noise is known, we introduce a novel architecture which we call Robust Conditional GAN (RCGAN). The main idea is to corrupt the label of the generated sample before feeding to the adversarial discriminator, forcing the generator to produce samples with clean labels. This approach of passing through a matching noisy channel is justified by corresponding multiplicative approximation bounds between the loss of the RCGAN and the distance between the clean real distribution and the generator distribution. This shows that the proposed approach is robust, when used with a carefully chosen discriminator architecture, known as projection discriminator. When the distribution of the noise is not known, we provide an extension of our architecture, which we call RCGAN-U, that learns the noise model simultaneously while training the generator. We show experimentally on MNIST and CIFAR-10 datasets that both the approaches consistently improve upon baseline approaches, and RCGAN-U closely matches the performance of RCGAN. …

Memory In Memory (MIM)
Natural spatiotemporal processes can be highly non-stationary in many ways, e.g. the low-level non-stationarity such as spatial correlations or temporal dependencies of local pixel values; and the high-level variations such as the accumulation, deformation or dissipation of radar echoes in precipitation forecasting. From Cramer’s Decomposition, any non-stationary process can be decomposed into deterministic, time-variant polynomials, plus a zero-mean stochastic term. By applying differencing operations appropriately, we may turn time-variant polynomials into a constant, making the deterministic component predictable. However, most previous recurrent neural networks for spatiotemporal prediction do not use the differential signals effectively, and their relatively simple state transition functions prevent them from learning too complicated variations in spacetime. We propose the Memory In Memory (MIM) networks and corresponding recurrent blocks for this purpose. The MIM blocks exploit the differential signals between adjacent recurrent states to model the non-stationary and approximately stationary properties in spatiotemporal dynamics with two cascaded, self-renewed memory modules. By stacking multiple MIM blocks, we could potentially handle higher-order non-stationarity. The MIM networks achieve the state-of-the-art results on three spatiotemporal prediction tasks across both synthetic and real-world datasets. We believe that the general idea of this work can be potentially applied to other time-series forecasting tasks. …

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 state-of-the-art methods. …

Propagation Map
Deep neural networks were shown to be vulnerable to single pixel modifications. However, the reason behind such phenomena has never been elucidated. Here, we propose Propagation Maps which show the influence of the perturbation in each layer of the network. Propagation Maps reveal that even in extremely deep networks such as Resnet, modification in one pixel easily propagates until the last layer. In fact, this initial local perturbation is also shown to spread becoming a global one and reaching absolute difference values that are close to the maximum value of the original feature maps in a given layer. Moreover, we do a locality analysis in which we demonstrate that nearby pixels of the perturbed one in the one-pixel attack tend to share the same vulnerability, revealing that the main vulnerability lies in neither neurons nor pixels but receptive fields. Hopefully, the analysis conducted in this work together with a new technique called propagation maps shall shed light into the inner workings of other adversarial samples and be the basis of new defense systems to come. …

Work Stealing Load Balancing Algorithm
A methodology for efficient load balancing of computational problems that can be easily decomposed into multiple tasks, but where it is hard to predict the computation cost of each task, and where new tasks are created dynamically during runtime. We present this methodology and its exploitation and feasibility in the context of graphics processors. Work-stealing allows an idle core to acquire tasks from a core that is overloaded, causing the total work to be distributed evenly among cores, while minimizing the communication costs, as tasks are only redistributed when required. This will often lead to higher throughput than using static partitioning.
Work Stealing with latency …

Nonlinear expectation
In probability theory, a nonlinear expectation is a nonlinear generalization of the expectation. Nonlinear expectations are useful in utility theory as they more closely match human behavior than traditional expectations. …

Turbo Smoothing
Recently, a novel method for developing filtering algorithms, based on the parallel concatenation of Bayesian filters and called turbo filtering, has been proposed. In this manuscript we show how the same conceptual approach can be exploited to devise a new smoothing method, called turbo smoothing. A turbo smoother combines a turbo filter, employed in its forward pass, with the parallel concatenation of two backward information filters used in its backward pass. As a specific application of our general theory, a detailed derivation of two turbo smoothing algorithms for conditionally linear Gaussian systems is illustrated. Numerical results for a specific dynamic system evidence that these algorithms can achieve a better complexity-accuracy tradeoff than other smoothing techniques recently appeared in the literature. …

Proof-of-Deep-Learning (PoDL)
An enormous amount of energy is wasted in Proofof-Work (PoW) mechanisms adopted by popular blockchain applications (e.g., PoW-based cryptocurrencies), because miners must conduct a large amount of computation. Owing to this, one serious rising concern is that the energy waste not only dilutes the value of the blockchain but also hinders its further application. In this paper, we propose a novel blockchain design that fully recycles the energy required for facilitating and maintaining it, which is re-invested to the computation of deep learning. We realize this by proposing Proof-of-Deep-Learning (PoDL) such that a valid proof for a new block can be generated if and only if a proper deep learning model is produced. We present a proof-of-concept design of PoDL that is compatible with the majority of the cryptocurrencies that are based on hash-based PoW mechanisms. Our benchmark and simulation results show that the proposed design is feasible for various popular cryptocurrencies such as Bitcoin, Bitcoin Cash, and Litecoin. …

Correlated Anomaly Detection (CAD)
Correlated anomaly detection (CAD) from streaming data is a type of group anomaly detection and an essential task in useful real-time data mining applications like botnet detection, financial event detection, industrial process monitor, etc. …

TaxoGen
Taxonomy construction is not only a fundamental task for semantic analysis of text corpora, but also an important step for applications such as information filtering, recommendation, and Web search. Existing pattern-based methods extract hypernym-hyponym term pairs and then organize these pairs into a taxonomy. However, by considering each term as an independent concept node, they overlook the topical proximity and the semantic correlations among terms. In this paper, we propose a method for constructing topic taxonomies, wherein every node represents a conceptual topic and is defined as a cluster of semantically coherent concept terms. Our method, TaxoGen, uses term embeddings and hierarchical clustering to construct a topic taxonomy in a recursive fashion. To ensure the quality of the recursive process, it consists of: (1) an adaptive spherical clustering module for allocating terms to proper levels when splitting a coarse topic into fine-grained ones; (2) a local embedding module for learning term embeddings that maintain strong discriminative power at different levels of the taxonomy. Our experiments on two real datasets demonstrate the effectiveness of TaxoGen compared with baseline methods. …

Hierarchical Navigable Small World (HNSW,Hierarchical NSW)
We present a new algorithm for the approximate nearest neighbor search based on navigable small world graphs with controllable hierarchy (Hierarchical NSW) admitting simple insertion, deletion and K-nearest neighbor queries. The Hierarchical NSW is a fully graph-based approach without a need for additional search structures (such as kd-trees or Cartesian concatenation) typically used at coarse search stage of the most proximity graph techniques. The algorithm incrementally builds a layered structure consisting from hierarchical set of proximity graphs (layers) for nested subsets of the stored elements. The maximum layer in which an element is present is selected randomly with exponentially decaying probability distribution. This allows producing graphs similar to the previously studied Navigable Small World (NSW) structures while additionally having the links separated by their characteristic distance scales. Starting search from the upper layer instead of random seeds together with utilizing the scale separation boosts the performance compared to the NSW and allows a logarithmic complexity scaling. Additional employment of a simple heuristic for selecting proximity graph neighbors increases performance at high recall and in case of highly clustered data. Performance evaluation on a large number of datasets has demonstrated that the proposed general metric space method is able to strongly outperform many previous state-of-art vector-only approaches such as FLANN, FALCONN and Annoy. Similarity of the algorithm to a well-known 1D skip list structure allows straightforward efficient and balanced distributed implementation.
A Comparative Study on Hierarchical Navigable Small World Graphs …

Imitation Network
In this paper, we propose imitation networks, a simple but effective method for training neural networks with a limited amount of training data. Our approach inherits the idea of knowledge distillation that transfers knowledge from a deep or wide reference model to a shallow or narrow target model. The proposed method employs this idea to mimic predictions of reference estimators that are much more robust against overfitting than the network we want to train. Different from almost all the previous work for knowledge distillation that requires a large amount of labeled training data, the proposed method requires only a small amount of training data. Instead, we introduce pseudo training examples that are optimized as a part of model parameters. Experimental results for several benchmark datasets demonstrate that the proposed method outperformed all the other baselines, such as naive training of the target model and standard knowledge distillation. …

Constrained Quantile Regression Averaging (CQRA)
Probabilistic load forecasts provide comprehensive information about future load uncertainties. In recent years, many methodologies and techniques have been proposed for probabilistic load forecasting. Forecast combination, a widely recognized best practice in point forecasting literature, has never been formally adopted to combine probabilistic load forecasts. This paper proposes a constrained quantile regression averaging (CQRA) method to create an improved ensemble from several individual probabilistic forecasts. We formulate the CQRA parameter estimation problem as a linear program with the objective of minimizing the pinball loss, with the constraints that the parameters are nonnegative and summing up to one. We demonstrate the effectiveness of the proposed method using two publicly available datasets, the ISO New England data and Irish smart meter data. Comparing with the best individual probabilistic forecast, the ensemble can reduce the pinball score by 4.39% on average. The proposed ensemble also demonstrates superior performance over nine other benchmark ensembles. …

Atomistic Structure Learning Algorithm (ASLA)
One endeavour of modern physical chemistry is to use bottom-up approaches to design materials and drugs with desired properties. Here we introduce an atomistic structure learning algorithm (ASLA) that utilizes a convolutional neural network to build 2D compounds and layered structures atom by atom. The algorithm takes no prior data or knowledge on atomic interactions but inquires a first-principles quantum mechanical program for physical properties. Using reinforcement learning, the algorithm accumulates knowledge of chemical compound space for a given number and type of atoms and stores this in the neural network, ultimately learning the blueprint for the optimal structural arrangement of the atoms for a given target property. ASLA is demonstrated to work on diverse problems, including grain boundaries in graphene sheets, organic compound formation and a surface oxide structure. This approach to structure prediction is a first step toward direct manipulation of atoms with artificially intelligent first principles computer codes. …

Grasp Quality Spatial Transformer Network (GQ-STN)
Grasping is a fundamental robotic task needed for the deployment of household robots or furthering warehouse automation. However, few approaches are able to perform grasp detection in real time (frame rate). To this effect, we present Grasp Quality Spatial Transformer Network (GQ-STN), a one-shot grasp detection network. Being based on the Spatial Transformer Network (STN), it produces not only a grasp configuration, but also directly outputs a depth image centered at this configuration. By connecting our architecture to an externally-trained grasp robustness evaluation network, we can train efficiently to satisfy a robustness metric via the backpropagation of the gradient emanating from the evaluation network. This removes the difficulty of training detection networks on sparsely annotated databases, a common issue in grasping. We further propose to use this robustness classifier to compare approaches, being more reliable than the traditional rectangle metric. Our GQ-STN is able to detect robust grasps on the depth images of the Dex-Net 2.0 dataset with 92.4 % accuracy in a single pass of the network. We finally demonstrate in a physical benchmark that our method can propose robust grasps more often than previous sampling-based methods, while being more than 60 times faster. …

Stacked Generalization (Stacking)
Stacked generalization (or stacking) (Wolpert, 1992) is a different way of combining multiple models, that introduces the concept of a meta learner. Although an attractive idea, it is less widely used than bagging and boosting. Unlike bagging and boosting, stacking may be (and normally is) used to combine models of different types. The procedure is as follows:
1. Split the training set into two disjoint sets.
2. Train several base learners on the first part.
3. Test the base learners on the second part.
4. Using the predictions from 3) as the inputs, and the correct responses as the outputs, train a higher level learner.
Note that steps 1) to 3) are the same as cross-validation, but instead of using a winner-takes-all approach, we combine the base learners, possibly nonlinearly. …

Rucio
Rucio is an open source software framework that provides scientific collaborations with the functionality to organize, manage, and access their volumes of data. The data can be distributed across heterogeneous data centers at widely distributed locations. Rucio has been originally developed to meet the requirements of the high-energy physics experiment ATLAS, and is continuously extended to support the LHC experiments and other diverse scientific communities. In this article we detail the fundamental concepts of Rucio, describe the architecture along with implementation details, and give operational experience from production usage. …

Transfer Learning
Machine learning and data mining techniques have been used in numerous real-world applications. An assumption of traditional machine learning methodologies is the training data and testing data are taken from the same domain, such that the input feature space and data distribution characteristics are the same. However, in some real-world machine learning scenarios, this assumption does not hold. There are cases where training data is expensive or difficult to collect. Therefore, there is a need to create high-performance learners trained with more easily obtained data from different domains. This methodology is referred to as transfer learning. This survey paper formally defines transfer learning, presents information on current solutions, and reviews applications applied to transfer learning. Lastly, there is information listed on software downloads for various transfer learning solutions and a discussion of possible future research work. The transfer learning solutions surveyed are independent of data size and can be applied to big data environments.
Recycling Deep Learning Models with Transfer Learning …

Multi-Stage Temporal Convolutional Network (MS-TCN)
Temporally locating and classifying action segments in long untrimmed videos is of particular interest to many applications like surveillance and robotics. While traditional approaches follow a two-step pipeline, by generating frame-wise probabilities and then feeding them to high-level temporal models, recent approaches use temporal convolutions to directly classify the video frames. In this paper, we introduce a multi-stage architecture for the temporal action segmentation task. Each stage features a set of dilated temporal convolutions to generate an initial prediction that is refined by the next one. This architecture is trained using a combination of a classification loss and a proposed smoothing loss that penalizes over-segmentation errors. Extensive evaluation shows the effectiveness of the proposed model in capturing long-range dependencies and recognizing action segments. Our model achieves state-of-the-art results on three challenging datasets: 50Salads, Georgia Tech Egocentric Activities (GTEA), and the Breakfast dataset. …

Low-Rank-Functional (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 non-convex quadratic functional, which will hence-forth be termed as the \textit{Low-Rank-Functional} (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. …