H2O  The Open Source InMemory Prediction Engine for Big Data Science. H2O is an awesome machine learning framework. It is really great for data scientists and business analysts ‘who need scalable and fast machine learning’. H2O is completely open source and what makes it important is that works right of the box. There seems to be no easier way to start with scalable machine learning. It hast support for R, Python, Scala, Java and also has a REST API and a own WebUI. So you can use it perfectly for research but also in production environments. H2O is based on Apache Hadoop and Apache Spark which gives it enormous power with inmemory parallel processing. Predict Social Network Influence with R and H2O Ensemble Learning 
Haar Scattering Network  
Habitat  We present Habitat, a new platform for research in embodied artificial intelligence (AI). Habitat enables training embodied agents (virtual robots) in highly efficient photorealistic 3D simulation, before transferring the learned skills to reality. Specifically, Habitat consists of the following: 1. HabitatSim: a flexible, highperformance 3D simulator with configurable agents, multiple sensors, and generic 3D dataset handling (with builtin support for SUNCG, Matterport3D, Gibson datasets). HabitatSim is fast — when rendering a scene from the Matterport3D dataset, HabitatSim achieves several thousand frames per second (fps) running singlethreaded, and can reach over 10,000 fps multiprocess on a single GPU, which is orders of magnitude faster than the closest simulator. 2. HabitatAPI: a modular highlevel library for endtoend development of embodied AI algorithms — defining embodied AI tasks (e.g. navigation, instruction following, question answering), configuring and training embodied agents (via imitation or reinforcement learning, or via classic SLAM), and benchmarking using standard metrics. These largescale engineering contributions enable us to answer scientific questions requiring experiments that were till now impracticable or `merely’ impractical. Specifically, in the context of pointgoal navigation (1) we revisit the comparison between learning and SLAM approaches from two recent works and find evidence for the opposite conclusion — that learning outperforms SLAM, if scaled to total experience far surpassing that of previous investigations, and (2) we conduct the first crossdataset generalization experiments {train, test} x {Matterport3D, Gibson} for multiple sensors {blind, RGB, RGBD, D} and find that only agents with depth (D) sensors generalize across datasets. We hope that our opensource platform and these findings will advance research in embodied AI. 
Hadamard Matrix Guided Online Hashing (HMOH) 
Online image hashing has received increasing research attention recently, which receives largescale data in a streaming manner to update the hash functions onthefly. Its key challenge lies in the difficulty in balancing the learning timeliness and model accuracy. To this end, most works exploit a supervised setting, i.e., using class labels to boost the hashing performance, which defects in two aspects: First, large amount of training batches are required to learn uptodate hash functions, which however largely increase the learning complexity. Second, strong constraints, e.g., orthogonal or similarity preserving, are used, which are however typically relaxed and lead to large accuracy drop. To handle the above challenges, in this paper, a novel supervised online hashing scheme termed Hadamard Matrix Guided Online Hashing (HMOH) is proposed. Our key innovation lies in the construction and usage of Hadamard matrix, which is an orthogonal binary matrix and is built via Sylvester method. To release the need of strong constraints, we regard each column of Hadamard matrix as the target code for each class label, which by nature satisfies several desired properties of hashing codes. To accelerate the online training, the LSH is first adopted to align the length of target code and the tobelearned binary code. And then, we treat the learning of hash functions as a set of binary classification problems to fit the assigned target code. Finally, we propose to ensemble the learned models in all rounds to maximally preserve the information of past streaming data. The superior accuracy and efficiency of the proposed method are demonstrated through extensive experiments on three widelyused datasets comparing to various stateoftheart methods. 
HalfLife of Data  Radioactive substances have a half life. The half life is the amount of time it takes for the substance to lose half of its radioactivity. Half life is used more generally in physics as a way to estimate the rate of decay. We can apply exactly the same principle – the rate of decay – to business information. Like natural materials, data is subject to deterioration over time. In science, the half life of a given substance could be milliseconds. It could be many thousands of years. The half life of data has been measured, and it may be shorter than you were expecting. http://…/infographicsthehalflifeofdata 
Halide  Halide is a computer programming language designed for writing digital image processing code that takes advantage of memory locality, vectorized computation and multicore CPUs and GPUs. Halide is implemented as an internal domainspecific language (DSL) in C++. The main innovation Halide brings is the separation of the algorithm being implemented from its execution schedule, i.e. code specifying the loop nesting, parallelization, loop unrolling and vector instruction. These two are usually interleaved together and experimenting with changing the schedule requires the programmer to rewrite large portions of the algorithm with every change. With Halide, changing the schedule does not require any changes to the algorithm and this allows the programmer to experiment with scheduling and finding the most efficient one. DNN Dataflow Choice Is Overrated 
Hamiltonian Flow Monte Carlo (HFMC) 

Hamiltonian Monte Carlo (HMC) 
The randomwalk behavior of many Markov Chain Monte Carlo (MCMC) algorithms makes Markov chain convergence to a target stationary distribution p(x) inefficient, resulting in slow mixing. Hamiltonian/Hybrid Monte Carlo (HMC), is a MCMC method that adopts physical system dynamics rather than a probability distribution to propose future states in the Markov chain. This allows the Markov chain to explore the target distribution much more efficiently, resulting in faster convergence. Here we introduce basic analytic and numerical concepts for simulation of Hamiltonian dynamics. We then show how Hamiltonian dynamics can be used as the Markov chain proposal function for an MCMC sampling algorithm (HMC). ➘ “Hybrid Monte Carlo” MCMC using Hamiltonian Dynamics 
Hamiltonian Variational AutoEncoder (HVAE) 
Variational AutoEncoders (VAEs) have become very popular techniques to perform inference and learning in latent variable models as they allow us to leverage the rich representational power of neural networks to obtain flexible approximations of the posterior of latent variables as well as tight evidence lower bounds (ELBOs). Combined with stochastic variational inference, this provides a methodology scaling to large datasets. However, for this methodology to be practically efficient, it is necessary to obtain lowvariance unbiased estimators of the ELBO and its gradients with respect to the parameters of interest. While the use of Markov chain Monte Carlo (MCMC) techniques such as Hamiltonian Monte Carlo (HMC) has been previously suggested to achieve this [23, 26], the proposed methods require specifying reverse kernels which have a large impact on performance. Additionally, the resulting unbiased estimator of the ELBO for most MCMC kernels is typically not amenable to the reparameterization trick. We show here how to optimally select reverse kernels in this setting and, by building upon Hamiltonian Importance Sampling (HIS) [17], we obtain a scheme that provides lowvariance unbiased estimators of the ELBO and its gradients using the reparameterization trick. This allows us to develop a Hamiltonian Variational AutoEncoder (HVAE). This method can be reinterpreted as a targetinformed normalizing flow [20] which, within our context, only requires a few evaluations of the gradient of the sampled likelihood and trivial Jacobian calculations at each iteration. 
HamiltonJacobi Reachability Analysis (HJRA) 
HamiltonJacobi (HJ) reachability analysis is an important formal verification method for guaranteeing performance and safety properties of dynamical systems; it has been applied to many smallscale systems in the past decade. Its advantages include compatibility with general nonlinear system dynamics, formal treatment of bounded disturbances, and the availability of welldeveloped numerical tools. The main challenge is addressing its exponential computational complexity with respect to the number of state variables. In this tutorial, we present an overview of basic HJ reachability theory and provide instructions for using the most recent numerical tools, including an efficient GPUparallelized implementation of a Level Set Toolbox for computing reachable sets. In addition, we review some of the current work in highdimensional HJ reachability to show how the dimensionality challenge can be alleviated via various general theoretical and applicationspecific insights. 
Hamming Distance  In information theory, the Hamming distance between two strings of equal length is the number of positions at which the corresponding symbols are different. In another way, it measures the minimum number of substitutions required to change one string into the other, or the minimum number of errors that could have transformed one string into the other. 
HANA  ➘ “SAP HANA” 
HANA Data Scientist Tool  The Application Function Modeler 2.0 (AFM 2) is a graphical editor for complex data analysis pipelines in the HANA Studio. This tool is based on the HANA Data Scientist prototype developed at the HANA Platform Innovation Center in Potsdam, Germany. It is planned to be the next generation of the existing HANA Studio Application Function Modeler which was developed at the TIP CE&SP Algorithm Labs in Shanghai, China. The AFM 2 team consists of original and new developers from both locations. 
HANA Graph Engine  The HANA Graph Engine implements graph data processing capabilities directly inside the Column Store Engine of the SAP HANA Database. 
HANA Sizing  Check the HANA sizing overview to find the appropriate sizing method. 
HANA Social Media Integration (HANASMI) 
HANASMI is a reusable component on HANA XS that enables XS application developers the integration of social media providers (with an initial focus on SAP Jam) into their business application. 
Handsontable  Handsontable is a data grid component with an Excellike appearance. Built in JavaScript, it integrates with any data source with peak efficiency. It comes with powerful features like data validation, sorting, grouping, data binding, formula support or column ordering. Built and actively supported by the Handsoncode team and the GitHub community ?, distributed free under the MIT license. rhandsontable 
HardELiSH  Deep Neural Networks have been shown to be beneficial for a variety of tasks, in particular allowing for endtoend learning and reducing the requirement for manual design decisions. However, still many parameters have to be chosen in advance, also raising the need to optimize them. One important, but often ignored system parameter is the selection of a proper activation function. Thus, in this paper we target to demonstrate the importance of activation functions in general and show that for different tasks different activation functions might be meaningful. To avoid the manual design or selection of activation functions, we build on the idea of genetic algorithms to learn the best activation function for a given task. In addition, we introduce two new activation functions, ELiSH and HardELiSH, which can easily be incorporated in our framework. In this way, we demonstrate for three different image classification benchmarks that different activation functions are learned, also showing improved results compared to typically used baselines. 
HardnessAware Deep Metric Learning  This paper presents a hardnessaware deep metric learning (HDML) framework. Most previous deep metric learning methods employ the hard negative mining strategy to alleviate the lack of informative samples for training. However, this mining strategy only utilizes a subset of training data, which may not be enough to characterize the global geometry of the embedding space comprehensively. To address this problem, we perform linear interpolation on embeddings to adaptively manipulate their hard levels and generate corresponding labelpreserving synthetics for recycled training, so that information buried in all samples can be fully exploited and the metric is always challenged with proper difficulty. Our method achieves very competitive performance on the widely used CUB2002011, Cars196, and Stanford Online Products datasets. 
HardtoFindData (HTFD) 
Well, really more of a 4letter acronym, but a powerful advantage of DaaS is the ability to source hardtofind data that has been aggregated from hundreds of Big Data sources. These data sets are highly targeted and go well beyond third party lists. 
Hardware Aware Knowledge Distillation (HAKD) 
Despite recent developments, deploying deep neural networks on resource constrained general purpose hardware remains a significant challenge. There has been much work in developing methods for reshaping neural networks, usually with a focus on minimising total parameter count. These methods are typically developed in a hardwareagnostic manner and do not exploit hardware behaviour. In this paper we propose a new approach, Hardware Aware Knowledge Distillation (HAKD) which uses empirical observations of hardware behaviour to design efficient student networks which are then trained with knowledge distillation. This allows the tradeoff between accuracy and performance to be managed explicitly. We have applied this approach across three platforms and evaluated it on two networks, MobileNet and DenseNet, on CIFAR10. We show that HAKD outperforms Deep Compression and Fisher pruning in terms of size, accuracy and performance. 
HardwareAware Automated Quantization (HAQ) 
Model quantization is a widely used technique to compress and accelerate deep neural network (DNN) inference. Emergent DNN hardware accelerators begin to support flexible bitwidth (18 bits) to further improve the computation efficiency, which raises a great challenge to find the optimal bitwidth for each layer: it requires domain experts to explore the vast design space trading off among accuracy, latency, power, and model size, which is both timeconsuming and suboptimal. Conventional quantization algorithm ignores the different hardware architectures and quantizes all the layers in an uniform way. In this paper, we introduce the HardwareAware Automated Quantization (HAQ) framework which leverages the reinforcement learning to automatically determine the quantization policy, and we take the hardware accelerator’s feedback in the design loop. Rather than relying on proxy signals such as FLOPs and model size, we employ a hardware simulator to generate direct feedback signals to the RL agent. Compared with conventional methods, our framework is fully automated and can specialize the quantization policy for different neural network architectures and hardware architectures. Our framework effectively reduced the latency by 1.41.95x and the energy consumption by 1.9x with negligible loss of accuracy compared with the fixed bitwidth (8 bits) quantization. Our framework reveals that the optimal policies on different hardware architectures (i.e., edge and cloud architectures) under different resource constraints (i.e., latency, power and model size) are drastically different. We interpreted the implication of different quantization policies, which offer insights for both neural network architecture design and hardware architecture design. 
Harmonia  Distributed storage employs replication to mask failures and improve availability. However, these systems typically exhibit a hard tradeoff between consistency and performance. Ensuring consistency introduces coordination overhead, and as a result the system throughput does not scale with the number of replicas. We present Harmonia, a replicated storage architecture that exploits the capability of newgeneration programmable switches to obviate this tradeoff by providing nearlinear scalability without sacrificing consistency. To achieve this goal, Harmonia detects readwrite conflicts in the network, which enables any replica to serve reads for objects with no pending writes. Harmonia implements this functionality at line rate, thus imposing no performance overhead. We have implemented a prototype of Harmonia on a cluster of commodity servers connected by a Barefoot Tofino switch, and have integrated it with Redis. We demonstrate the generality of our approach by supporting a variety of replication protocols, including primarybackup, chain replication, Viewstamped Replication, and NOPaxos. Experimental results show that Harmonia improves the throughput of these protocols by up to 10X for a replication factor of 10, providing nearlinear scalability up to the limit of our testbed. 
Harmonic Adversarial Attack Method (HAAM) 
Adversarial attacks find perturbations that can fool models into misclassifying images. Previous works had successes in generating noisy/edgerich adversarial perturbations, at the cost of degradation of image quality. Such perturbations, even when they are small in scale, are usually easily spottable by human vision. In contrast, we propose Harmonic Adversarial Attack Methods (HAAM), that generates edgefree perturbations by using harmonic functions. The property of edgefree guarantees that the generated adversarial images can still preserve visual quality, even when perturbations are of large magnitudes. Experiments also show that adversaries generated by HAAM often have higher rates of success when transferring between models. In addition, we find harmonic perturbations can simulate natural phenomena like natural lighting and shadows. It would then be possible to help find corner cases for given models, as a first step to improving them. 
Harmonic Coding  We consider the problem of distributedly computing a general class of functions, referred to as gradienttype computation, while maintaining the privacy of the input dataset. Gradienttype computation evaluates the sum of some `partial gradients’, defined as polynomials of subsets of the input. It underlies many algorithms in machine learning and data analytics. We propose Harmonic Coding, which universally computes any gradienttype function, while requiring the minimum possible number of workers. Harmonic Coding strictly improves computing schemes developed based on prior works, such as Shamir’s secret sharing and Lagrange Coded Computing, by injecting coded redundancy using harmonic progression. It enables the computing results of the workers to be interpreted as the sum of partial gradients and some redundant results, which then allows the cancellation of nongradient terms in the decoding process. By proving a matching converse, we demonstrate the optimality of Harmonic Coding, even compared to the schemes that are nonuniversal (i.e., can be designed based on a specific gradienttype function). 
Harmony Search Algorithm (HSA) 
In computer science and operations research, harmony search (HS) is a phenomenonmimicking algorithm (also known as metaheuristic algorithm, soft computing algorithm or evolutionary algorithm) inspired by the improvisation process of musicians proposed by Zong Woo Geem in 2001. In the HS algorithm, each musician (= decision variable) plays (= generates) a note (= a value) for finding a best harmony (= global optimum) all together. Proponents claim the following merits: · HS does not require differential gradients, thus it can consider discontinuous functions as well as continuous functions. · HS can handle discrete variables as well as continuous variables. · HS does not require initial value setting for the variables. · HS is free from divergence. · HS may escape local optima. · HS may overcome the drawback of GA’s building block theory which works well only if the relationship among variables in a chromosome is carefully considered. If neighbor variables in a chromosome have weaker relationship than remote variables, building block theory may not work well because of crossover operation. However, HS explicitly considers the relationship using ensemble operation. · HS has a novel stochastic derivative applied to discrete variables, which uses musician’s experiences as a searching direction. · Certain HS variants do not require algorithm parameters such as HMCR and PAR, thus novice users can easily use the algorithm. Harmony Search Algorithm Harmony Search Algorithm 
Hartley Spectral Pooling for Deep Learning  In most convolution neural networks (CNNs), downsampling hidden layers is adopted for increasing computation efficiency and the receptive field size. Such operation is commonly socalled pooling. Maximation and averaging over sliding windows (max/average pooling), and plain downsampling in the form of strided convolution are popular pooling methods. Since the pooling is a lossy procedure, a motivation of our work is to design a new pooling approach for less lossy in the dimensionality reduction. Inspired by the Fourier spectral pooling(FSP) proposed by Rippel et. al. [1], we present the Hartley transform based spectral pooling method in CNNs. Compared with FSP, the proposed spectral pooling avoids the use of complex arithmetic for frequency representation and reduces the computation. Spectral pooling preserves more structure features for network’s discriminability than max and average pooling. We empirically show that Hartley spectral pooling gives rise to the convergence of training CNNs on MNIST and CIFAR10 datasets. 
HARVEST Algorithm  Feature selection with highdimensional data and a very small proportion of relevant features poses a severe challenge to standard statistical methods. We have developed a new approach (HARVEST) that is straightforward to apply, albeit somewhat computerintensive. This algorithm can be used to prescreen a large number of features to identify those that are potentially useful. The basic idea is to evaluate each feature in the context of many random subsets of other features. HARVEST is predicated on the assumption that an irrelevant feature can add no real predictive value, regardless of which other features are included in the subset. Motivated by this idea, we have derived a simple statistical test for feature relevance. Empirical analyses and simulations produced so far indicate that the HARVEST algorithm is highly effective in predictive analytics, both in science and business. 
Harvest Classification Algorithm  A tree model will often provide good prediction relative to other methods. It is also relatively interpretable, which is key, since it is of interest to identify diverse chemical classes amongst the active compounds, to serve as leads for drug optimization. Interpretability of a tree is often reduced, however, by the sheer size and number of variables involved. We develop a ‘tree harvesting’ algorithm to reduce the complexity of the tree. Harvest.Tree 
HASBRAIN  Mobile video consumption is increasing and sophisticated video quality adaptation strategies are required to deal with mobile throughput fluctuations. These adaptation strategies have to keep the switching frequency low, the average quality high and prevent stalling occurrences to ensure customer satisfaction. This paper proposes a novel methodology for the design of machine learningbased adaptation logics named HASBRAIN. Furthermore, the performance of a trained neural network against two algorithms from the literature is evaluated. We first use a modified existing optimization formulation to calculate optimal adaptation paths with a minimum number of quality switches for a wide range of videos and for challenging mobile throughput patterns. Afterwards we use the resulting optimal adaptation paths to train and compare different machine learning models. The evaluation shows that an artificial neural networkbased model can reach a high average quality with a low number of switches in the mobile scenario. The proposed methodology is general enough to be extended for further designs of machine learningbased algorithms and the provided model can be deployed in ondemand streaming scenarios or be further refined using rewardbased mechanisms such as reinforcement learning. All tools, models and datasets created during the work are provided as opensource software. 
Hash2Vec  In this paper we propose the application of feature hashing to create word embeddings for natural language processing. Feature hashing has been used successfully to create document vectors in related tasks like document classification. In this work we show that feature hashing can be applied to obtain word embeddings in linear time with the size of the data. The results show that this algorithm, that does not need training, is able to capture the semantic meaning of words. We compare the results against GloVe showing that they are similar. As far as we know this is the first application of feature hashing to the word embeddings problem and the results indicate this is a scalable technique with practical results for NLP applications. 
Hashing Transformation Deep Neural Network (HashTranDNN) 
Adversarial machine learning in the context of image processing and related applications has received a large amount of attention. However, adversarial machine learning, especially adversarial deep learning, in the context of malware detection has received much less attention despite its apparent importance. In this paper, we present a framework for enhancing the robustness of Deep Neural Networks (DNNs) against adversarial malware samples, dubbed Hashing Transformation Deep Neural Networks} (HashTranDNN). The core idea is to use hash functions with a certain localitypreserving property to transform samples to enhance the robustness of DNNs in malware classification. The framework further uses a Denoising AutoEncoder (DAE) regularizer to reconstruct the hash representations of samples, making the resulting DNN classifiers capable of attaining the locality information in the latent space. We experiment with two concrete instantiations of the HashTranDNN framework to classify Android malware. Experimental results show that four known attacks can render standard DNNs useless in classifying Android malware, that known defenses can at most defend three of the four attacks, and that HashTranDNN can effectively defend against all of the four attacks. 
HashNet  Learning to hash has been widely applied to approximate nearest neighbor search for largescale multimedia retrieval, due to its computation efficiency and retrieval quality. Deep learning to hash, which improves retrieval quality by endtoend representation learning and hash encoding, has received increasing attention recently. Subject to the vanishing gradient difficulty in the optimization with binary activations, existing deep learning to hash methods need to first learn continuous representations and then generate binary hash codes in a separated binarization step, which suffer from substantial loss of retrieval quality. This paper presents HashNet, a novel deep architecture for deep learning to hash by continuation method, which learns exactly binary hash codes from imbalanced similarity data where the number of similar pairs is much smaller than the number of dissimilar pairs. The key idea is to attack the vanishing gradient problem in optimizing deep networks with nonsmooth binary activations by continuation method, in which we begin from learning an easier network with smoothed activation function and let it evolve during the training, until it eventually goes back to being the original, difficult to optimize, deep network with the sign activation function. Comprehensive empirical evidence shows that HashNet can generate exactly binary hash codes and yield stateoftheart multimedia retrieval performance on standard benchmarks. 
Haversine Distance  The haversine formula determines the greatcircle distance between two points on a sphere given their longitudes and latitudes. Important in navigation, it is a special case of a more general formula in spherical trigonometry, the law of haversines, that relates the sides and angles of spherical triangles. The first table of haversines in English was published by James Andrew in 1805, but Florian Cajori credits an earlier use by José de Mendoza y Ríos in 1801. The term haversine was coined in 1835 by James Inman. These names follow from the fact that they are customarily written in terms of the haversine function, given by haversin( ) = sin^2(theta/2). The formulas could equally be written in terms of any multiple of the haversine, such as the older versine function (twice the haversine). Prior to the advent of computers, the elimination of division and multiplication by factors of two proved convenient enough that tables of haversine values and logarithms were included in 19th and early 20th century navigation and trigonometric texts. These days, the haversine form is also convenient in that it has no coefficient in front of the sin^2 function. 
Hawkes Graph  This paper introduces the Hawkes skeleton and the Hawkes graph. These notions summarize the branching structure of a multivariate Hawkes point process in a compact and fertile way. In particular, we explain how the graph view is useful for the specification and estimation of Hawkes models from large, multitype event streams. Based on earlier work, we give a nonparametric statistical procedure to estimate the Hawkes skeleton and the Hawkes graph from data. We show how the graph estimation may then be used for choosing and fitting parametric Hawkes models. Our method avoids the a priori assumptions on the model from a straighforward MLEapproach and it is numerically more flexible than the latter. A simulation study confirms that the presented procedure works as desired. We give special attention to computational issues in the implementation. This makes our results applicable to highdimensional eventstream data, such as dozens of event streams and thousands of events per component. 
Hawkes Process  Hawkes processes are a particularly interesting class of stochastic process that have been applied in diverse areas, from earthquake modelling to financial analysis. They are point processes whose defining characteristic is that they ‘selfexcite’, meaning that each arrival increases the rate of future arrivals for some period of time. Hawkes processes are well established, particularly within the financial literature, yet many of the treatments are inaccessible to one not acquainted with the topic. This survey provides background, introduces the field and historical developments, and touches upon all major aspects of Hawkes processes. Hawkes Processes 
Hazard Function  The hazard function (also known as the failure rate, hazard rate, or force of mortality) h(x) is the ratio of the probability density function P(x) to the survival function S(x), given by h(x) = P(x)/S(x) = P(x)/(1 – D(x)), where D(x) is the distribution function. 
Hazard Ratio  In survival analysis, the hazard ratio (HR) is the ratio of the hazard rates corresponding to the conditions described by two levels of an explanatory variable. For example, in a drug study, the treated population may die at twice the rate per unit time as the control population. The hazard ratio would be 2, indicating higher hazard of death from the treatment. Or in another study, men receiving the same treatment may suffer a certain complication ten times more frequently per unit time than women, giving a hazard ratio of 10. Hazard ratios differ from relative risks in that the latter are cumulative over an entire study, using a defined endpoint, while the former represent instantaneous risk over the study time period, or some subset thereof. Hazard ratios suffer somewhat less from selection bias with respect to the endpoints chosen and can indicate risks that happen before the endpoint. 
Hazelcast  Hazelcast, a leading open source inmemory data grid (IMDG) with hundreds of thousands of installed clusters and over 17 million server starts per month, launched Hazelcast Jet – a distributed processing engine for big data streams. With Hazelcast’s IMDG providing storage functionality, Hazelcast Jet is a new Apache 2 licensed open source project that performs parallel execution to enable dataintensive applications to operate in near realtime. Using directed acyclic graphs (DAG) to model relationships between individual steps in the data processing pipeline, Hazelcast Jet is simple to deploy and can execute both batch and streambased data processing applications. Hazelcast Jet is appropriate for applications that require a near realtime experience such as sensor updates in IoT architectures (house thermostats, lighting systems), instore ecommerce systems and social media platforms. 
HCqa  Question Answering (QA) systems provide easy access to the vast amount of knowledge without having to know the underlying complex structure of the knowledge. The research community has provided ad hoc solutions to the key QA tasks, including named entity recognition and disambiguation, relation extraction and query building. Furthermore, some have integrated and composed these components to implement many tasks automatically and efficiently. However, in general, the existing solutions are limited to simple and short questions and still do not address complex questions composed of several subquestions. Exploiting the answer to complex questions is further challenged if it requires integrating knowledge from unstructured data sources, i.e., textual corpus, as well as structured data sources, i.e., knowledge graphs. In this paper, an approach (HCqa) is introduced for dealing with complex questions requiring federating knowledge from a hybrid of heterogeneous data sources (structured and unstructured). We contribute in developing (i) a decomposition mechanism which extracts subquestions from potentially long and complex input questions, (ii) a novel comprehensive schema, first of its kind, for extracting and annotating relations, and (iii) an approach for executing and aggregating the answers of subquestions. The evaluation of HCqa showed a superior accuracy in the fundamental tasks, such as relation extraction, as well as the federation task. 
hdetach  Recurrent neural networks are known for their notorious exploding and vanishing gradient problem (EVGP). This problem becomes more evident in tasks where the information needed to correctly solve them exist over long time scales, because EVGP prevents important gradient components from being backpropagated adequately over a large number of steps. We introduce a simple stochastic algorithm (\textit{h}detach) that is specific to LSTM optimization and targeted towards addressing this problem. Specifically, we show that when the LSTM weights are large, the gradient components through the linear path (cell state) in the LSTM computational graph get suppressed. Based on the hypothesis that these components carry information about long term dependencies (which we show empirically), their suppression can prevent LSTMs from capturing them. Our algorithm prevents gradients flowing through this path from getting suppressed, thus allowing the LSTM to capture such dependencies better. We show significant convergence and generalization improvements using our algorithm on various benchmark datasets. 
HDIdx  Fast Nearest Neighbor (NN) search is a fundamental challenge in largescale data processing and analytics, particularly for analyzing multimedia contents which are often of high dimensionality. Instead of using exact NN search, extensive research efforts have been focusing on approximate NN search algorithms. In this work, we present ‘HDIdx’, an efficient highdimensional indexing library for fast approximate NN search, which is opensource and written in Python. It offers a family of stateoftheart algorithms that convert input highdimensional vectors into compact binary codes, making them very efficient and scalable for NN search with very low space complexity. 
HDTCat  HDT (Header, Dictionary, Triples) is a serialization for RDF. HDT has become very popular in the last years because it allows to store RDF data with a small disk footprint, while remaining at the same time queriable. For this reason HDT is often used when scalability becomes an issue. Once RDF data is serialized into HDT, the disk footprint to store it and the memory footprint to query it are very low. However, generating HDT files from raw text RDF serializations (like NTriples) is a timeconsuming and (especially) memoryconsuming task. In this publication we present HDTCat, an algorithm and command line tool to join two HDT files with low memory footprint. HDTCat can be used in a divideandconquer strategy to generate HDT files from huge datasets using a lowmemory footprint. 
Header, Dictionary, Triples (HDT) 
Currently RDF data is stored and sent in very verbose textual serialization formats that waste a lot of bandwidth and are expensive to parse and index. If RDF is meant to be machine understandable, why not use an appropriate format for that? HDT (Header, Dictionary, Triples) is a compact data structure and binary serialization format for RDF that keeps big datasets compressed to save space while maintaining search and browse operations without prior decompression. This makes it an ideal format for storing and sharing RDF datasets on the Web. 
Heaped Data  
HeavyTailed Horseshoe Prior  Locally adaptive shrinkage in the Bayesian framework is achieved through the use of localglobal prior distributions that model both the global level of sparsity as well as individual shrinkage parameters for mean structure parameters. The most popular of these models is the Horseshoe prior and its variants due to their spike and slab behavior involving an asymptote at the origin and heavy tails. In this article, we present an alternative Horseshoe prior that exhibits both a sharper asymptote at the origin as well as heavier tails, which we call the Heavytailed Horseshoe prior. We prove that mixing on the shape parameters provides improved spike and slab behavior as well as better reconstruction properties than other Horseshoe variants. A simulation study is provided to show the advantage of the heavytailed Horseshoe in terms of absolute error to both the truth mean structure as well as the oracle. 
Heckman Correction  The Heckman correction (the twostage method, Heckman’s lambda or the Heckit method, Heckman Model) is any of a number of related statistical methods developed by James Heckman at the University of Chicago in 1976 to 1979 which allow the researcher to correct for selection bias. Selection bias problems are endemic to applied econometric problems, which make Heckman’s original technique, and subsequent refinements by both himself and others, indispensable to applied econometricians. Heckman received the Economics Nobel Prize in 2000 for this achievement. http://…/HeckmanSelectionModel.html 
Hedonic Regression  In economics, hedonic regression or hedonic demand theory is a revealed preference method of estimating demand or value. It decomposes the item being researched into its constituent characteristics, and obtains estimates of the contributory value of each characteristic. This requires that the composite good being valued can be reduced to its constituent parts and that the market values those constituent parts. Hedonic models are most commonly estimated using regression analysis, although more generalized models, such as sales adjustment grids, are special cases of hedonic models. An attribute vector, which may be a dummy or panel variable, is assigned to each characteristic or group of characteristics. Hedonic models can accommodate nonlinearity, variable interaction, or other complex valuation situations. Hedonic models are commonly used in real estate appraisal, real estate economics, and Consumer Price Index (CPI) calculations. In CPI calculations hedonic regression is used to control the effect of changes in product quality. Price changes that are due to substitution effects are subject to hedonic quality adjustments. 
Helix  Machine learning workflow development is a process of trialanderror: developers iterate on workflows by testing out small modifications until the desired accuracy is achieved. Unfortunately, existing machine learning systems focus narrowly on model training—a small fraction of the overall development time—and neglect to address iterative development. We propose Helix, a machine learning system that optimizes the execution across iterations—intelligently caching and reusing, or recomputing intermediates as appropriate. Helix captures a wide variety of application needs within its Scala DSL, with succinct syntax defining unified processes for data preprocessing, model specification, and learning. We demonstrate that the reuse problem can be cast as a MaxFlow problem, while the caching problem is NPHard. We develop effective lightweight heuristics for the latter. Empirical evaluation shows that Helix is not only able to handle a wide variety of use cases in one unified workflow but also much faster, providing run time reductions of up to 19x over stateoftheart systems, such as DeepDive or KeystoneML, on four realworld applications in natural language processing, computer vision, social and natural sciences. 
HellaSwag  Recent work by Zellers et al. (2018) introduced a new task of commonsense natural language inference: given an event description such as ‘A woman sits at a piano,’ a machine must select the most likely followup: ‘She sets her fingers on the keys.’ With the introduction of BERT, near humanlevel performance was reached. Does this mean that machines can perform human level commonsense inference? In this paper, we show that commonsense inference still proves difficult for even stateoftheart models, by presenting HellaSwag, a new challenge dataset. Though its questions are trivial for humans (>95% accuracy), stateoftheart models struggle (<48%). We achieve this via Adversarial Filtering (AF), a data collection paradigm wherein a series of discriminators iteratively select an adversarial set of machinegenerated wrong answers. AF proves to be surprisingly robust. The key insight is to scale up the length and complexity of the dataset examples towards a critical ‘Goldilocks’ zone wherein generated text is ridiculous to humans, yet often misclassified by stateoftheart models. Our construction of HellaSwag, and its resulting difficulty, sheds light on the inner workings of deep pretrained models. More broadly, it suggests a new path forward for NLP research, in which benchmarks coevolve with the evolving stateoftheart in an adversarial way, so as to present everharder challenges. 
Hellinger Correlation  In this paper, the defining properties of a valid measure of the dependence between two random variables are reviewed and complemented with two original ones, shown to be more fundamental than other usual postulates. While other popular choices are proved to violate some of these requirements, a class of dependence measures satisfying all of them is identified. One particular measure, that we call the Hellinger correlation, appears as a natural choice within that class due to both its theoretical and intuitive appeal. A simple and efficient nonparametric estimator for that quantity is proposed. Synthetic and realdata examples finally illustrate the descriptive ability of the measure, which can also be used as test statistic for exact independence testing. 
Hellinger Distance  In probability and statistics, the Hellinger distance (also called Bhattacharyya distance as this was originally introduced by Anil Kumar Bhattacharya) is used to quantify the similarity between two probability distributions. It is a type of fdivergence. The Hellinger distance is defined in terms of the Hellinger integral, which was introduced by Ernst Hellinger in 1909.[1][2] 
Helm Charts  Helm uses a packaging format called charts. A chart is a collection of files that describe a related set of Kubernetes resources. A single chart might be used to deploy something simple, like a memcached pod, or something complex, like a full web app stack with HTTP servers, databases, caches, and so on. Charts are created as files laid out in a particular directory tree, then they can be packaged into versioned archives to be deployed. 
HELP  Large crowdsourced datasets are widely used for training and evaluating neural models on natural language inference (NLI). Despite these efforts, neural models have a hard time capturing logical inferences, including those licensed by phrase replacements, socalled monotonicity reasoning. Since no large dataset has been developed for monotonicity reasoning, it is still unclear whether the main obstacle is the size of datasets or the model architectures themselves. To investigate this issue, we introduce a new dataset, called HELP, for handling entailments with lexical and logical phenomena. We add it to training data for the stateoftheart neural models and evaluate them on test sets for monotonicity phenomena. The results showed that our data augmentation improved the overall accuracy. We also find that the improvement is better on monotonicity inferences with lexical replacements than on downward inferences with disjunction and modification. This suggests that some types of inferences can be improved by our data augmentation while others are immune to it. 
Henge  We present Henge, a system to support intentbased multitenancy in modern stream processing applications. Henge supports multitenancy as a firstclass citizen: everyone inside an organization can now submit their stream processing jobs to a single, shared, consolidated cluster. Additionally, Henge allows each tenant (job) to specify its own intents (i.e., requirements) as a Service Level Objective (SLO) that captures latency and/or throughput. In a multitenant cluster, the Henge scheduler adapts continually to meet jobs’ SLOs in spite of limited cluster resources, and under dynamic input workloads. SLOs are soft and are based on utility functions. Henge continually tracks SLO satisfaction, and when jobs miss their SLOs, it wisely navigates the state space to perform resource allocations in real time, maximizing total system utility achieved by all jobs in the system. Henge is integrated in Apache Storm and we present experimental results using both production topologies and real datasets. 
Hereditary Independence Gap  The independence gap of a graph was introduced by Ekim et al. (2018) as a measure of how far a graph is from being wellcovered. It is defined as the difference between the maximum and minimum size of a maximal independent set. We investigate the independence gap of a graph from structural and algorithmic points of view, with a focus on classes of perfect graphs. Generalizing results on wellcovered graphs due to Dean and Zito (1994) and Hujdurovi\’c et al. (2018), we express the independence gap of a perfect graph in terms of clique partitions and use this characterization to develop a polynomialtime algorithm for recognizing graphs of constant independence gap in any class of perfect graphs of bounded clique number. Next, we introduce a hereditary variant of the parameter, which we call hereditary independence gap and which measures the maximum independence gap over all induced subgraphs of the graph. We show that determining whether a given graph has hereditary independence gap at most $k$ is polynomialtime solvable if $k$ is fixed and coNPcomplete if $k$ is part of input. We also investigate the complexity of the independent set problem in graph classes related to independence gap, showing that the problem is NPcomplete in the class of graphs of independence gap at most one and polynomialtime solvable in any class of graphs with bounded hereditary independence gap. Combined with some known results on clawfree graphs, our results imply that the independent domination problem is solvable in polynomial time. 
HerfindahlHirschman Index  Based on the aggregated shares retained by individual firms or actors within a market or space, the HerfindahlHirschman Index (HHI) measures the level of concentration in the market or space. It is often used as a measure of competition, where 0 equals perfect competition amongst firms or actors and 10,000 equals perfect monopoly. hhi 
Hessian Approximated Multiple Subsets Iteration (HAMSI) 
We propose HAMSI, a provably convergent incremental algorithm for solving largescale partially separable optimization problems that frequently emerge in machine learning and inferential statistics. The algorithm is based on a local quadratic approximation and hence allows incorporating a second order curvature information to speedup the convergence. Furthermore, HAMSI needs almost no tuning, and it is scalable as well as easily parallelizable. In largescale simulation studies with the MovieLens datasets, we illustrate that the method is superior to a stateoftheart distributed stochastic gradient descent method in terms of convergence behavior. This performance gain comes at the expense of using memory that scales only linearly with the total size of the optimization variables. We conclude that HAMSI may be considered as a viable alternative in many scenarios, where first order methods based on variants of stochastic gradient descent are applicable. 
Hessian AWare Quantization (HAWQ) 
Model size and inference speed/power have become a major challenge in the deployment of Neural Networks for many applications. A promising approach to address these problems is quantization. However, uniformly quantizing a model to ultra low precision leads to significant accuracy degradation. A novel solution for this is to use mixedprecision quantization, as some parts of the network may allow lower precision as compared to other layers. However, there is no systematic way to determine the precision of different layers. A brute force approach is not feasible for deep networks, as the search space for mixedprecision is exponential in the number of layers. Another challenge is a similar factorial complexity for determining blockwise finetuning order when quantizing the model to a target precision. Here, we introduce Hessian AWare Quantization (HAWQ), a novel secondorder quantization method to address these problems. HAWQ allows for the automatic selection of the relative quantization precision of each layer, based on the layer’s Hessian spectrum. Moreover, HAWQ provides a deterministic finetuning order for quantizing layers, based on secondorder information. We show the results of our method on Cifar10 using ResNet20, and on ImageNet using InceptionV3, ResNet50 and SqueezeNext models. Comparing HAWQ with stateoftheart shows that we can achieve similar/better accuracy with $8\times$ activation compression ratio on ResNet20, as compared to DNAS~\cite{wu2018mixed}, and up to $1\%$ higher accuracy with up to $14\%$ smaller models on ResNet50 and InceptionV3, compared to recently proposed methods of RVQuant~\cite{park2018value} and HAQ~\cite{wang2018haq}. Furthermore, we show that we can quantize SqueezeNext to just 1MB model size while achieving above $68\%$ top1 accuracy on ImageNet. 
Heterogeneous Autoregressive Realised Volatility (HARRV) 
We propose a heterogeneous simultaneous graphical dynamic linear model (HSGDLM), which extends the standard SGDLM framework to incorporate a heterogeneous autoregressive realised volatility (HARRV) model. This novel approach creates a GPUscalable multivariate volatility estimator, which decomposes multiple time series into economicallymeaningful variables to explain the endogenous and exogenous factors driving the underlying variability. This unique decomposition goes beyond the classic one step ahead prediction; indeed, we investigate inferences up to one month into the future using stocks, FX futures and ETF futures, demonstrating its superior performance according to accuracy of large moves, longerterm prediction and consistency over time. 
Heterogeneous Deep Diffusion (HDD) 
There are many realworld knowledge based networked systems with multitype interacting entities that can be regarded as heterogeneous networks including human connections and biological evolutions. One of the main issues in such networks is to predict information diffusion such as shape, growth and size of social events and evolutions in the future. While there exist a variety of works on this topic mainly using a thresholdbased approach, they suffer from the local viewpoint on the network and sensitivity to the threshold parameters. In this paper, information diffusion is considered through a latent representation learning of the heterogeneous networks to encode in a deep learning model. To this end, we propose a novel metapath representation learning approach, Heterogeneous Deep Diffusion(HDD), to exploit metapaths as main entities in networks. At first, the functional heterogeneous structures of the network are learned by a continuous latent representation through traversing metapaths with the aim of global endtoend viewpoint. Then, the wellknown deep learning architectures are employed on our generated features to predict diffusion processes in the network. The proposed approach enables us to apply it on different information diffusion tasks such as topic diffusion and cascade prediction. We demonstrate the proposed approach on benchmark network datasets through the wellknown evaluation measures. The experimental results show that our approach outperforms the earlier stateoftheart methods. 
Heterogeneous Deep Discriminative Model (HDDM) 
This paper presents a new deep learning approach for videobased scene classification. We design a Heterogeneous Deep Discriminative Model (HDDM) whose parameters are initialized by performing an unsupervised pretraining in a layerwise fashion using Gaussian Restricted Boltzmann Machines (GRBM). In order to avoid the redundancy of adjacent frames, we extract spatiotemporal variation patterns within frames and represent them sparsely using Sparse Cubic Symmetrical Pattern (SCSP). Then, a preinitialized HDDM is separately trained using the videos of each class to learn classspecific models. According to the minimum reconstruction error from the learnt classspecific models, a weighted voting strategy is employed for the classification. The performance of the proposed method is extensively evaluated on two action recognition datasets; UCF101 and Hollywood II, and three dynamic texture and dynamic scene datasets; DynTex, YUPENN, and Maryland. The experimental results and comparisons against stateoftheart methods demonstrate that the proposed method consistently achieves superior performance on all datasets. 
Heterogeneous Graph Neural Network  Graph neural network, as a powerful graph representation technique based on deep learning, has shown superior performance and attracted considerable research interest. However, it has not been fully considered in graph neural network for heterogeneous graph which contains different types of nodes and links. The heterogeneity and rich semantic information bring great challenges for designing a graph neural network for heterogeneous graph. Recently, one of the most exciting advancements in deep learning is the attention mechanism, whose great potential has been well demonstrated in various areas. In this paper, we first propose a novel heterogeneous graph neural network based on the hierarchical attention, including nodelevel and semanticlevel attentions. Specifically, the nodelevel attention aims to learn the importance between a node and its metapath based neighbors, while the semanticlevel attention is able to learn the importance of different metapaths. With the learned importance from both nodelevel and semanticlevel attention, the importance of node and metapath can be fully considered. Then the proposed model can generate node embedding by aggregating features from metapath based neighbors in a hierarchical manner. Extensive experimental results on three realworld heterogeneous graphs not only show the superior performance of our proposed model over the stateofthearts, but also demonstrate its potentially good interpretability for graph analysis. 
Heterogeneous Incremental Nearest Class Mean Random Forest (hiRF) 
In recent years, dynamically growing data and incrementally growing number of classes pose new challenges to largescale data classification research. Most traditional methods struggle to balance the precision and computational burden when data and its number of classes increased. However, some methods are with weak precision, and the others are timeconsuming. In this paper, we propose an incremental learning method, namely, heterogeneous incremental Nearest Class Mean Random Forest (hiRF), to handle this issue. It is a heterogeneous method that either replaces trees or updates trees leaves in the random forest adaptively, to reduce the computational time in comparable performance, when data of new classes arrive. Specifically, to keep the accuracy, one proportion of trees are replaced by new NCM decision trees; to reduce the computational load, the rest trees are updated their leaves probabilities only. Most of all, outofbag estimation and outofbag boosting are proposed to balance the accuracy and the computational efficiency. Fair experiments were conducted and demonstrated its comparable precision with much less computational time. 
Heterogeneous Information Network (HIN) 
Representation Learning for Heterogeneous Information Networks via Embedding Events 
Heterogeneous Information Network Learning (HINLearning) 
The explosive growth and increasing sophistication of Android malware call for new defensive techniques that are capable of protecting mobile users against novel threats. In this paper, we first extract the runtime Application Programming Interface (API) call sequences from Android apps, and then analyze higherlevel semantic relations within the ecosystem to comprehensively characterize the apps. To model different types of entities (i.e., app, API, IMEI, signature, affiliation) and the rich semantic relations among them, we then construct a structural heterogeneous information network (HIN) and present metapath based approach to depict the relatedness over apps. To efficiently classify nodes (e.g., apps) in the constructed HIN, we propose the HinLearning method to first obtain insample node embeddings and then learn representations of outofsample nodes without rerunning/adjusting HIN embeddings at the first attempt. Afterwards, we design a deep neural network (DNN) classifier taking the learned HIN representations as inputs for Android malware detection. A comprehensive experimental study on the largescale real sample collections from Tencent Security Lab is performed to compare various baselines. Promising experimental results demonstrate that our developed system AiDroid which integrates our proposed method outperforms others in realtime Android malware detection. AiDroid has already been incorporated into Tencent Mobile Security product that serves millions of users worldwide. 
Heterogeneous Information NetworkBased Text Clustering Framework (HINT) 
Currently, many intelligence systems contain the texts from multisources, e.g., bulletin board system (BBS) posts, tweets and news. These texts can be “comparative” since they may be semantically correlated and thus provide us with different perspectives toward the same topics or events. To better organize the multisourced texts and obtain more comprehensive knowledge, we propose to study the novel problem of Mutual Clustering on Comparative Texts (MCCT), which aims to cluster the comparative texts simultaneously and collaboratively. The MCCT problem is difficult to address because 1) comparative texts usually present different data formats and structures and thus they are hard to organize, and 2) there lacks an effective method to connect the semantically correlated comparative texts to facilitate clustering them in an unified way. To this aim, in this paper we propose a Heterogeneous Information Networkbased Text clustering framework HINT. HINT first models multisourced texts (e.g. news and tweets) as heterogeneous information networks by introducing the shared “anchor texts” to connect the comparative texts. Next, two similarity matrices based on HINT as well as a transition matrix for crosstextsource knowledge transfer are constructed. Comparative texts clustering are then conducted by utilizing the constructed matrices. Finally, a mutual clustering algorithm is also proposed to further unify the separate clustering results of the comparative texts by introducing a clustering consistency constraint. We conduct extensive experimental on three tweetsnews datasets, and the results demonstrate the effectiveness and robustness of the proposed method in addressing the MCCT problem. 
Heterogeneous MultiTask Metric Learning (HMTML) 
Distance metric learning (DML) plays a crucial role in diverse machine learning algorithms and applications. When the labeled information in target domain is limited, transfer metric learning (TML) helps to learn the metric by leveraging the sufficient information from other related domains. Multitask metric learning (MTML), which can be regarded as a special case of TML, performs transfer across all related domains. Current TML tools usually assume that the same feature representation is exploited for different domains. However, in realworld applications, data may be drawn from heterogeneous domains. Heterogeneous transfer learning approaches can be adopted to remedy this drawback by deriving a metric from the learned transformation across different domains. But they are often limited in that only two domains can be handled. To appropriately handle multiple domains, we develop a novel heterogeneous multitask metric learning (HMTML) framework. In HMTML, the metrics of all different domains are learned together. The transformations derived from the metrics are utilized to induce a common subspace, and the highorder covariance among the predictive structures of these domains is maximized in this subspace. There do exist a few heterogeneous transfer learning approaches that deal with multiple domains, but the highorder statistics (correlation information), which can only be exploited by simultaneously examining all domains, is ignored in these approaches. Compared with them, the proposed HMTML can effectively explore such highorder information, thus obtaining more reliable feature transformations and metrics. Effectiveness of our method is validated by the extensive and intensive experiments on text categorization, scene classification, and social image annotation. 
Heterogeneous Simultaneous Graphical Dynamic Linear Model (HSGDLM) 
We propose a heterogeneous simultaneous graphical dynamic linear model (HSGDLM), which extends the standard SGDLM framework to incorporate a heterogeneous autoregressive realised volatility (HARRV) model. This novel approach creates a GPUscalable multivariate volatility estimator, which decomposes multiple time series into economicallymeaningful variables to explain the endogenous and exogenous factors driving the underlying variability. This unique decomposition goes beyond the classic one step ahead prediction; indeed, we investigate inferences up to one month into the future using stocks, FX futures and ETF futures, demonstrating its superior performance according to accuracy of large moves, longerterm prediction and consistency over time. 
Heterogeneous Simultaneous Multiscale Change Point Estimator (HSMUCE) 
We propose, a heterogeneous simultaneous multiscale change point estimator called ‘HSMUCE’ for the detection of multiple change points of the signal in a heterogeneous Gaussian regression model. A piecewise constant function is estimated by minimizing the number of change points over the acceptance region of a multiscale test which locally adapts to changes in the variance. The multiscale test is a combination of local likelihood ratio tests which are properly calibrated by scaledependent critical values to keep a global nominal level a, even for finite samples. We show that HSMUCE controls the error of overestimation and underestimation of the number of change points. For this, new deviation bounds for Ftype statistics are derived. Moreover, we obtain confidence sets for the whole signal. All results are nonasymptotic and uniform over a large class of heterogeneous change point models. HSMUCE is fast to compute, achieves the optimal detection rate and estimates the number of change points at almost optimal accuracy for vanishing signals, while still being robust. We compare HSMUCE with several state of the art methods in simulations and analyse current recordings of a transmembrane protein in the bacterial outer membrane with pronounced heterogeneity for its states. An Rpackage is available on line. 
Heterogeneous Tasks on Homogeneous Cores (HTHC) 
A new generation of manycore processors is on the rise that offers dozens and more cores on a chip and, in a sense, fuses host processor and accelerator. In this paper we target the efficient training of generalized linear models on these machines. We propose a novel approach for achieving parallelism which we call Heterogeneous Tasks on Homogeneous Cores (HTHC). It divides the problem into multiple fundamentally different tasks, which themselves are parallelized. For evaluation, we design a detailed, architecturecognizant implementation of our scheme on a recent 72core Knights Landing processor that is adaptive to the cache, memory, and core structure. Experiments for Lasso and SVM with different data sets show a speedup of typically an order of magnitude compared to straightforward parallel implementations in C++. 
Heterogeneous Transactional Memory (HeTM) 
Modern heterogeneous computing architectures, which couple multicore CPUs with discrete manycore GPUs (or other specialized hardware accelerators), enable unprecedented peak performance and energy efficiency levels. Unfortunately, though, developing applications that can take full advantage of the potential of heterogeneous systems is a notoriously hard task. This work takes a step towards reducing the complexity of programming heterogeneous systems by introducing the abstraction of Heterogeneous Transactional Memory (HeTM). HeTM provides programmers with the illusion of a single memory region, shared among the CPUs and the (discrete) GPU(s) of a heterogeneous system, with support for atomic transactions. Besides introducing the abstract semantics and programming model of HeTM, we present the design and evaluation of a concrete implementation of the proposed abstraction, which we named Speculative HeTM (SHeTM). SHeTM makes use of a novel design that leverages on speculative techniques and aims at hiding the inherently large communication latency between CPUs and discrete GPUs and at minimizing interdevice synchronization overhead. SHeTM is based on a modular and extensible design that allows for easily integrating alternative TM implementations on the CPU’s and GPU’s sides, which allows the flexibility to adopt, on either side, the TM implementation (e.g., in hardware or software) that best fits the applications’ workload and the architectural characteristics of the processing unit. We demonstrate the efficiency of the SHeTM via an extensive quantitative study based both on synthetic benchmarks and on a porting of a popular object caching system. 
Heterogeneous Transfer Distance Metric Learning (HTDML) 
The goal of transfer learning is to improve the performance of target learning task by leveraging information (or transferring knowledge) from other related tasks. In this paper, we examine the problem of transfer distance metric learning (DML), which usually aims to mitigate the label information deficiency issue in the target DML. Most of the current Transfer DML (TDML) methods are not applicable to the scenario where data are drawn from heterogeneous domains. Some existing heterogeneous transfer learning (HTL) approaches can learn target distance metric by usually transforming the samples of source and target domain into a common subspace. However, these approaches lack flexibility in realworld applications, and the learned transformations are often restricted to be linear. This motivates us to develop a general flexible heterogeneous transfer distance metric learning (HTDML) framework. In particular, any (linear/nonlinear) DML algorithms can be employed to learn the source metric beforehand. Then the prelearned source metric is represented as a set of knowledge fragments to help target metric learning. We show how generalization error in the target domain could be reduced using the proposed transfer strategy, and develop novel algorithm to learn either linear or nonlinear target metric. Extensive experiments on various applications demonstrate the effectiveness of the proposed method. 
Heterogeneous UltraDense Network (HUDN) 
Machine Learning for Heterogeneous UltraDense Networks with Graphical Representations 
Heteroscedasticity  In statistics, a collection of random variables is heteroscedastic if there are subpopulations that have different variabilities from others. Here “variability” could be quantified by the variance or any other measure of statistical dispersion. Thus heteroscedasticity is the absence of homoscedasticity. 
Heteroskedastic PCA (HeteroPCA) 
Principal component analysis (PCA) and singular value decomposition (SVD) are widely used in statistics, machine learning, and applied mathematics. It has been well studied in the case of homoskedastic noise, where the noise levels of the contamination are homogeneous. In this paper, we consider PCA and SVD in the presence of heteroskedastic noise, which arises naturally in a range of applications. We introduce a general framework for heteroskedastic PCA and propose an algorithm called HeteroPCA, which involves iteratively imputing the diagonal entries to remove the bias due to heteroskedasticity. This procedure is computationally efficient and provably optimal under the generalized spiked covariance model. A key technical step is a deterministic robust perturbation analysis on the singular subspace, which can be of independent interest. The effectiveness of the proposed algorithm is demonstrated in a suite of applications, including heteroskedastic lowrank matrix denoising, Poisson PCA, and SVD based on heteroskedastic and incomplete data. 
Heuristic Analysis for NLI Systems (HANS) 
Machine learning systems can often achieve high performance on a test set by relying on heuristics that are effective for frequent example types but break down in more challenging cases. We study this issue within natural language inference (NLI), the task of determining whether one sentence entails another. Based on an analysis of the task, we hypothesize three fallible syntactic heuristics that NLI models are likely to adopt: the lexical overlap heuristic, the subsequence heuristic, and the constituent heuristic. To determine whether models have adopted these heuristics, we introduce a controlled evaluation set called HANS (Heuristic Analysis for NLI Systems), which contains many examples where the heuristics fail. We find that models trained on MNLI, including the stateoftheart model BERT, perform very poorly on HANS, suggesting that they have indeed adopted these heuristics. We conclude that there is substantial room for improvement in NLI systems, and that the HANS dataset can motivate and measure progress in this area. 
Heuristics Allied with Distant Supervision (HAnDS) 
Finegrained Entity Recognition (FgER) is the task of detecting and classifying entity mentions to a large set of types spanning diverse domains such as biomedical, finance and sports. We observe that when the type set spans several domains, detection of entity mention becomes a limitation for supervised learning models. The primary reason being lack of dataset where entity boundaries are properly annotated while covering a large spectrum of entity types. Our work directly addresses this issue. We propose Heuristics Allied with Distant Supervision (HAnDS) framework to automatically construct a quality dataset suitable for the FgER task. HAnDS framework exploits the high interlink among Wikipedia and Freebase in a pipelined manner, reducing annotation errors introduced by naively using distant supervision approach. Using HAnDS framework, we create two datasets, one suitable for building FgER systems recognizing up to 118 entity types based on the FIGER type hierarchy and another for up to 1115 entity types based on the TypeNet hierarchy. Our extensive empirical experimentation warrants the quality of the generated datasets. Along with this, we also provide a manually annotated dataset for benchmarking FgER systems. 
HexaGAN  Most deep learning classification studies assume clean data. However, dirty data is prevalent in real world, and this undermines the classification performance. The data we practically encounter has problems such as 1) missing data, 2) class imbalance, and 3) missing label. Preprocessing techniques assume one of these problems and mitigate it, but an algorithm that assumes all three problems and resolves them has not yet been proposed. Therefore, in this paper, we propose HexaGAN, a generative adversarial network (GAN) framework that shows good classification performance for all three problems. We interpret the three problems from a similar perspective to solve them jointly. To enable this, the framework consists of six components, which interact in an endtoend manner. We also devise novel loss functions corresponding to the architecture. The designed loss functions achieve stateoftheart imputation performance with up to a 14% improvement and highquality classconditional data. We evaluate the classification performance (F1score) of the proposed method with 20% missingness and confirm up to a 5% improvement in comparison with the combinations of stateoftheart methods. 
HGCaffe  Breakthroughs in the fields of deep learning and mobile systemonchips are radically changing the way we use our smartphones. However, deep neural networks inference is still a challenging task for edge AI devices due to the computational overhead on mobile CPUs and a severe drain on the batteries. In this paper, we present a deep neural network inference engine named HGCaffe, which supports GPUs with half precision. HGCaffe provides up to 20 times speedup with GPUs compared to the original implementations. In addition to the speedup, the peak memory usage is also reduced to about 80%. With HGCaffe, more innovative and fascinating mobile applications will be turned into reality. 
Hidden Factor Graph Models (HFM) 
Hidden Factor graph models generalise Hidden Markov Models to tree structured data. The distinctive feature of ‘treeHFM’ is that it learns a transition matrix for first order (sequential) and for second order (splitting) events. It can be applied to all discrete and continuous data that is structured as a binary tree. In the case of continuous observations, ‘treeHFM’ has Gaussian distributions as emissions. treeHFM 
Hidden Markov Model (HMM) 
Hidden Markov Models (HMMs) are powerful, flexible methods for representing and classifying data with trends over time. A hidden Markov model (HMM) is a statistical Markov model in which the system being modeled is assumed to be a Markov process with unobserved (hidden) states. A HMM can be considered the simplest dynamic Bayesian network. The mathematics behind the HMM was developed by L. E. Baum and coworkers. It is closely related to an earlier work on optimal nonlinear filtering problem (stochastic processes) by Ruslan L. Stratonovich, who was the first to describe the forwardbackward procedure. In simpler Markov models (like a Markov chain), the state is directly visible to the observer, and therefore the state transition probabilities are the only parameters. In a hidden Markov model, the state is not directly visible, but output, dependent on the state, is visible. Each state has a probability distribution over the possible output tokens. Therefore the sequence of tokens generated by an HMM gives some information about the sequence of states. Note that the adjective ‘hidden’ refers to the state sequence through which the model passes, not to the parameters of the model; the model is still referred to as a ‘hidden’ Markov model even if these parameters are known exactly. Hidden Markov models are especially known for their application in temporal pattern recognition such as speech, handwriting, gesture recognition, partofspeech tagging, musical score following, partial discharges and bioinformatics. 
Hidden Tree Markov Network (HTN) 
The paper introduces the Hidden Tree Markov Network (HTN), a neuroprobabilistic hybrid fusing the representation power of generative models for trees with the incremental and discriminative learning capabilities of neural networks. We put forward a modular architecture in which multiple generative models of limited complexity are trained to learn structural feature detectors whose outputs are then combined and integrated by neural layers at a later stage. In this respect, the model is both deep, thanks to the unfolding of the generative models on the input structures, as well as wide, given the potentially large number of generative modules that can be trained in parallel. Experimental results show that the proposed approach can outperform stateoftheart syntactic kernels as well as generative kernels built on the same probabilistic model as the HTN. 
HiddenLayer LSTM (HLSTM) 
Long shortterm memory (LSTM) has been widely used for sequential data modeling. Researchers have increased LSTM depth by stacking LSTM cells to improve performance. This incurs model redundancy, increases runtime delay, and makes the LSTMs more prone to overfitting. To address these problems, we propose a hiddenlayer LSTM (HLSTM) that adds hidden layers to LSTM’s original one level nonlinear control gates. HLSTM increases accuracy while employing fewer external stacked layers, thus reducing the number of parameters and runtime latency significantly. We employ growandprune (GP) training to iteratively adjust the hidden layers through gradientbased growth and magnitudebased pruning of connections. This learns both the weights and the compact architecture of HLSTM control gates. We have GPtrained HLSTMs for image captioning and speech recognition applications. For the NeuralTalk architecture on the MSCOCO dataset, our three models reduce the number of parameters by 38.7x [floatingpoint operations (FLOPs) by 45.5x], runtime latency by 4.5x, and improve the CIDEr score by 2.6. For the DeepSpeech2 architecture on the AN4 dataset, our two models reduce the number of parameters by 19.4x (FLOPs by 23.5x), runtime latency by 15.7%, and the word error rate from 12.9% to 8.7%. Thus, GPtrained HLSTMs can be seen to be compact, fast, and accurate. 
HideandSeek  We propose ‘HideandSeek’ a general purpose data augmentation technique, which is complementary to existing data augmentation techniques and is beneficial for various visual recognition tasks. The key idea is to hide patches in a training image randomly, in order to force the network to seek other relevant content when the most discriminative content is hidden. Our approach only needs to modify the input image and can work with any network to improve its performance. During testing, it does not need to hide any patches. The main advantage of HideandSeek over existing data augmentation techniques is its ability to improve object localization accuracy in the weaklysupervised setting, and we therefore use this task to motivate the approach. However, HideandSeek is not tied only to the image localization task, and can generalize to other forms of visual input like videos, as well as other recognition tasks like image classification, temporal action localization, semantic segmentation, emotion recognition, age/gender estimation, and person reidentification. We perform extensive experiments to showcase the advantage of HideandSeek on these various visual recognition problems. 
Hierarchical Attention Mechanism (Ham) 
Attention mechanisms in sequence to sequence models have shown great ability and wonderful performance in various natural language processing (NLP) tasks, such as sentence embedding, text generation, machine translation, machine reading comprehension, etc. Unfortunately, existing attention mechanisms only learn either highlevel or lowlevel features. In this paper, we think that the lack of hierarchical mechanisms is a bottleneck in improving the performance of the attention mechanisms, and propose a novel Hierarchical Attention Mechanism (Ham) based on the weighted sum of different layers of a multilevel attention. Ham achieves a stateoftheart BLEU score of 0.26 on Chinese poem generation task and a nearly 6.5% averaged improvement compared with the existing machine reading comprehension models such as BIDAF and MatchLSTM. Furthermore, our experiments and theorems reveal that Ham has greater generalization and representation ability than existing attention mechanisms. 
Hierarchical Attention Network (HAN) 
Knowledge Base (KB) completion, which aims to determine missing relation between entities, has raised increasing attention in recent years. Most existing methods either focus on the positional relationship between entity pair and single relation (1hop path) in semantic space or concentrate on the joint probability of Random Walks on multihop paths among entities. However, they do not fully consider the intrinsic relationships of all the links among entities. By observing that the single relation and multihop paths between the same entity pair generally contain shared/similar semantic information, this paper proposes a novel method to capture the shared features between them as the basis for inferring missing relations. To capture the shared features jointly, we develop Hierarchical Attention Networks (HANs) to automatically encode the inputs into lowdimensional vectors, and exploit two partial parametershared components, one for feature source discrimination and the other for determining missing relations. By joint Adversarial Training (AT) the entire model, our method minimizes the classification error of missing relations, and ensures the source of shared features are difficult to discriminate in the meantime. The AT mechanism encourages our model to extract features that are both discriminative for missing relation prediction and shareable between single relation and multihop paths. We extensively evaluate our method on several largescale KBs for relation completion. Experimental results show that our method consistently outperforms the baseline approaches. In addition, the hierarchical attention mechanism and the feature extractor in our model can be well interpreted and utilized in the related downstream tasks. 
Hierarchical AttentionBased Recurrent Highway Network (HRHN) 
Time series prediction has been studied in a variety of domains. However, it is still challenging to predict future series given historical observations and past exogenous data. Existing methods either fail to consider the interactions among different components of exogenous variables which may affect the prediction accuracy, or cannot model the correlations between exogenous data and target data. Besides, the inherent temporal dynamics of exogenous data are also related to the target series prediction, and thus should be considered as well. To address these issues, we propose an endtoend deep learning model, i.e., Hierarchical attentionbased Recurrent Highway Network (HRHN), which incorporates spatiotemporal feature extraction of exogenous variables and temporal dynamics modeling of target variables into a single framework. Moreover, by introducing the hierarchical attention mechanism, HRHN can adaptively select the relevant exogenous features in different semantic levels. We carry out comprehensive empirical evaluations with various methods over several datasets, and show that HRHN outperforms the state of the arts in time series prediction, especially in capturing sudden changes and sudden oscillations of time series. 
Hierarchical AttentionBased Temporal Convolutional Network (HATCN) 
Myotonia, which refers to delayed muscle relaxation after contraction, is the main symptom of myotonic dystrophy patients. We propose a hierarchical attentionbased temporal convolutional network (HATCN) for myotonic dystrohpy diagnosis from handgrip time series data, and introduce mechanisms that enable model explainability. We compare the performance of the HATCN model against that of benchmark TCN models, LSTM models with and without attention mechanisms, and SVM approaches with handcrafted features. In terms of classification accuracy and F1 score, we found all deep learning models have similar levels of performance, and they all outperform SVM. Further, the HATCN model outperforms its TCN counterpart with regards to computational efficiency regardless of network depth, and in terms of performance particularly when the number of hidden layers is small. Lastly, HATCN models can consistently identify relevant time series segments in the relaxation phase of the handgrip time series, and exhibit increased robustness to noise when compared to attentionbased LSTM models. 
Hierarchical Attentive Heterogeneous Information Network Embedding (HAHE) 
Given the intractability of large scale HIN, network embedding which learns low dimensional proximitypreserved representations for nodes in the new space becomes a natural way to analyse HIN. However, two challenges arise in HIN embedding. (1) Different HIN structures with different semantic meanings play different roles in capturing relationships among nodes in HIN, how can we learn personalized preferences over different metapaths for each individual node in HIN? (2) With the number of large scale HIN increasing dramatically in various web services, how can we update the embedding information of new nodes in an efficient way? To tackle these challenges, we propose a Hierarchical Attentive Heterogeneous information network Embedding (HAHE ) model which is capable of learning personalized metapath preferences for each node as well as updating the embedding information for each new node efficiently with only its neighbor node information. The proposed HAHE model extracts the semantic relationships among nodes in the semantic space based on different metapaths and adopts a neighborhood attention layer to conduct weighted aggregations of neighborhood structure features for each node, enabling the embedding information of each new node to be updated efficiently. Besides, a metapath attention layer is also employed to learn the personalized metapath preferences for each individual node. Extensive experiments on several realworld datasets show that our proposed HAHE model significantly outperforms the stateoftheart methods in terms of various evaluation metrics. 
Hierarchical Block Sparse Neural Network (HBsNN) 
Sparse deep neural networks(DNNs) are efficient in both memory and compute when compared to dense DNNs. But due to irregularity in computation of sparse DNNs, their efficiencies are much lower than that of dense DNNs on general purpose hardwares. This leads to poor/no performance benefits for sparse DNNs. Performance issue for sparse DNNs can be alleviated by bringing structure to the sparsity and leveraging it for improving runtime efficiency. But such structural constraints often lead to sparse models with suboptimal accuracies. In this work, we jointly address both accuracy and performance of sparse DNNs using our proposed class of neural networks called HBsNN ( Hierarchical Block Sparse Neural Networks). 
Hierarchical bMatching  A matching of a graph is a subset of edges no two of which share a common vertex, and a maximum matching is a matching of maximum cardinality. In a $b$matching every vertex $v$ has an associated bound $b_v$, and a maximum $b$matching is a maximum set of edges, such that every vertex $v$ appears in at most $b_v$ of them. We study an extension of this problem, termed {\em Hierarchical bMatching}. In this extension, the vertices are arranged in a hierarchical manner. At the first level the vertices are partitioned into disjoint subsets, with a given bound for each subset. At the second level the set of these subsets is again partitioned into disjoint subsets, with a given bound for each subset, and so on. In an {\em Hierarchical bmatching} we look for a maximum set of edges, that will obey all bounds (that is, no vertex $v$ participates in more than $b_v$ edges, then all the vertices in one subset do not participate in more that that subset’s bound of edges, and so on hierarchically). We propose a polynomialtime algorithm for this new problem, that works for any number of levels of this hierarchical structure. 
Hierarchical Clustering  In data mining, hierarchical clustering is a method of cluster analysis which seeks to build a hierarchy of clusters. Strategies for hierarchical clustering generally fall into two types: 1. Agglomerative: This is a “bottom up” approach: each observation starts in its own cluster, and pairs of clusters are merged as one moves up the hierarchy. 2. Divisive: This is a “top down” approach: all observations start in one cluster, and splits are performed recursively as one moves down the hierarchy. In general, the merges and splits are determined in a greedy manner. The results of hierarchical clustering are usually presented in a dendrogram. 
Hierarchical Clustering and Topic Modeling based on Fast Rank2 NMF (HierNMF2) 
The importance of unsupervised clustering and topic modeling is well recognized with everincreasing volumes of text data. In this paper, we propose a fast method for hierarchical clustering and topic modeling called HierNMF2. Our method is based on fast Rank2 nonnegative matrix factorization (NMF) that performs binary clustering and an efficient node splitting rule. Further utilizing the final leaf nodes generated in HierNMF2 and the idea of nonnegative least squares fitting, we propose a new clustering/topic modeling method called FlatNMF2 that recovers a flat clustering/topic modeling result in a very simple yet significantly more effective way than any other existing methods. We describe highly optimized open source software in C++ for both HierNMF2 and FlatNMF2 for hierarchical and partitional clustering/topic modeling of document data sets. Substantial experimental tests are presented that illustrate significant improvements both in computational time as well as quality of solutions. We compare our methods to other clustering methods including Kmeans, standard NMF, and CLUTO, and also topic modeling methods including latent Dirichlet allocation (LDA) and recently proposed algorithms for NMF with separability constraints. Overall, we present efficient tools for analyzing largescale data sets, and techniques that can be generalized to many other data analytics problem domains. 
Hierarchical Compartmental Model  A variety of trianglebased stochastic reserving techniques have been proposed for estimating future general insurance claims payments, ranging from generalized linear models (England and Verrall, 2002) to nonlinear hierarchical models (Guszcza, 2008). Methods incorporating both paid and incurred information have been explored (MartínezMiranda, Nielsen and Verrall, 2012; Quarg and Mack, 2004), which provide richer inference and improved interpretability. Furthermore, Bayesian methods (Zhang, Dukic and Guszcza, 2012; Meyers, 2007; England and Verrall, 2005; Verrall, 2004) have become increasingly ubiquitous; providing flexibility and the ability to robustly incorporate judgment into uncertainty projections. This paper explores a new trianglebased (and optionallyBayesian) stochastic reserving framework which considers the relationship between exposure, case reserves and paid claims. By doing so, it enables practitioners to build communicable models that are consistent with their understanding of the insurance claims process. Furthermore, it supports the identification and quantification of claims process characteristics to provide tangible business insights. Hierarchical compartmental reserving models 
Hierarchical Compositional Network (HCN) 
We introduce the hierarchical compositional network (HCN), a directed generative model able to discover and disentangle, without supervision, the building blocks of a set of binary images. The building blocks are binary features defined hierarchically as a composition of some of the features in the layer immediately below, arranged in a particular manner. At a high level, HCN is similar to a sigmoid belief network with pooling. Inference and learning in HCN are very challenging and existing variational approximations do not work satisfactorily. A main contribution of this work is to show that both can be addressed using maxproduct message passing (MPMP) with a particular schedule (no EM required). Also, using MPMP as an inference engine for HCN makes new tasks simple: adding supervision information, classifying images, or performing inpainting all correspond to clamping some variables of the model to their known values and running MPMP on the rest. When used for classification, fast inference with HCN has exactly the same functional form as a convolutional neural network (CNN) with linear activations and binary weights. However, HCN’s features are qualitatively very different. 
Hierarchical Configuration Model  We introduce a class of random graphs with a hierarchical community structure, which we call the hierarchical configuration model. On the intercommunity level, the graph is a configuration model, and on the intracommunity level, every vertex in the configuration model is replaced by a community: a small graph. These communities may have any shape, as long as they are connected. For these hierarchical graphs, we find the size of the largest component, the degree distribution and the clustering coefficient. Furthermore, we determine the conditions under which a giant percolation cluster exists, and find its size. 
Hierarchical Context Enabled Recurrent Neural Network (HCRNN) 
A long user history inevitably reflects the transitions of personal interests over time. The analyses on the user history require the robust sequential model to anticipate the transitions and the decays of user interests. The user history is often modeled by various RNN structures, but the RNN structures in the recommendation system still suffer from the longterm dependency and the interest drifts. To resolve these challenges, we suggest HCRNN with three hierarchical contexts of the global, the local, and the temporary interests. This structure is designed to withhold the global longterm interest of users, to reflect the local subsequence interests, and to attend the temporary interests of each transition. Besides, we propose a hierarchical contextbased gate structure to incorporate our \textit{interest drift assumption}. As we suggest a new RNN structure, we support HCRNN with a complementary \textit{bichannel attention} structure to utilize hierarchical context. We experimented the suggested structure on the sequential recommendation tasks with CiteULike, MovieLens, and LastFM, and our model showed the best performances in the sequential recommendations. 
Hierarchical Critics Assignment (HCA) 
In this paper, we investigate the use of global information to speed up the learning process and increase the cumulative rewards of multiagent reinforcement learning (MARL) tasks. Within the actorcritic MARL, we introduce multiple cooperative critics from two levels of the hierarchy and propose a hierarchical criticbased multiagent reinforcement learning algorithm. In our approach, the agent is allowed to receive information from local and global critics in a competition task. The agent not only receives lowlevel details but also consider coordination from high levels that receiving global information to increase operation skills. Here, we define multiple cooperative critics in the topbottom hierarchy, called the Hierarchical Critics Assignment (HCA) framework. Our experiment, a twoplayer tennis competition task in the Unity environment, tested HCA multiagent framework based on Asynchronous Advantage ActorCritic (A3C) with Proximal Policy Optimization (PPO) algorithm. The results showed that the HCA framework outperforms the nonhierarchical critics baseline method for MARL tasks. 
Hierarchical Data Format (HDF) 
Hierarchical Data Format (HDF, HDF4, or HDF5) is a set of file formats and libraries designed to store and organize large amounts of numerical data. Originally developed at the National Center for Supercomputing Applications, it is supported by the nonprofit HDF Group, whose mission is to ensure continued development of HDF5 technologies, and the continued accessibility of data stored in HDF. 
Hierarchical DCLSTMt Deep Learning  Advanced travel information and warning, if provided accurately, can help road users avoid traffic congestion through dynamic route planning and behavior change. It also enables traffic control centres mitigate the impact of congestion by activating Intelligent Transport System (ITS) proactively. Deep learning has become increasingly popular in recent years, following a surge of innovative GPU technology, highresolution, big datasets and thriving machine learning algorithms. However, there are few examples exploiting this emerging technology to develop applications for traffic prediction. This is largely due to the difficulty in capturing random, seasonal, nonlinear, and spatiotemporal correlated nature of traffic data. In this paper, we propose a datadriven modelling approach with a novel hierarchical DCLSTMt deep learning model for shortterm traffic speed prediction, a framework combined with convolutional neural network (CNN) and long shortterm memory (LSTM) models. A deep CNN model is employed to learn the spatiotemporal traffic patterns of the input graphs, which are then fed into a deep LSTM model for sequence learning. To capture traffic seasonal variations, time of the day and day of the week indicators are fused with trained features. The model is trained endtoend to predict travel speed in 15 to 90 minutes in the future. We compare the model performance against other baseline models including CNN, LGBM, LSTM, and traditional speedflow curves. Experiment results show that the DCLSTMt outperforms other models considerably. Model tests show that speed upstream also responds sensibly to a sudden accident occurring downstream. Our DCLSTMt model framework is also highly scalable for future extension such as for networkwide traffic prediction, which can also be improved by including additional features such as weather, long term seasonality and accident information. 
Hierarchical Deep Learning for Text Classification (HDLTex) 
The continually increasing number of documents produced each year necessitates ever improving information processing methods for searching, retrieving, and organizing text. Central to these information processing methods is document classification, which has become an important application for supervised learning. Recently the performance of these traditional classifiers has degraded as the number of documents has increased. This is because along with this growth in the number of documents has come an increase in the number of categories. This paper approaches this problem differently from current document classification methods that view the problem as multiclass classification. Instead we perform hierarchical classification using an approach we call Hierarchical Deep Learning for Text classification (HDLTex). HDLTex employs stacks of deep learning architectures to provide specialized understanding at each level of the document hierarchy. 
Hierarchical Deep Multiagent Reinforcement Learning (Hierarchical Deep MARL) 
Despite deep reinforcement learning has recently achieved great successes, however in multiagent environments, a number of challenges still remain. Multiagent reinforcement learning (MARL) is commonly considered to suffer from the problem of nonstationary environments and exponentially increasing policy space. It would be even more challenging to learn effective policies in circumstances where the rewards are sparse and delayed over long trajectories. In this paper, we study Hierarchical Deep Multiagent Reinforcement Learning (hierarchical deep MARL) in cooperative multiagent problems with sparse and delayed rewards, where efficient multiagent learning methods are desperately needed. We decompose the original MARL problem into hierarchies and investigate how effective policies can be learned hierarchically in synchronous/asynchronous hierarchical MARL frameworks. Several hierarchical deep MARL architectures, i.e., IndhDQN, hCom and hQmix, are introduced for different learning paradigms. Moreover, to alleviate the issues of sparse experiences in highlevel learning and nonstationarity in multiagent settings, we propose a new experience replay mechanism, named as Augmented Concurrent Experience Replay (ACER). We empirically demonstrate the effects and efficiency of our approaches in several classic Multiagent Trash Collection tasks, as well as in an extremely challenging team sports game, i.e., Fever Basketball Defense. 
Hierarchical Distribution Matching (HiDM) 
The implementation difficulties of combining distribution matching (DM) and dematching (invDM) for probabilistic shaping (PS) with softdecision forward error correction (FEC) coding can be relaxed by reverse concatenation, for which the FEC coding and decoding lies inside the shaping algorithms. PS can seemingly achieve performance close to the Shannon limit, although there are practical implementation challenges that need to be carefully addressed. We propose a hierarchical DM (HiDM) scheme, having fully parallelized input/output interfaces and a pipelined architecture that can efficiently perform the DM/invDM without the complex operations of previously proposed methods such as constant composition DM (CCDM). Furthermore, HiDM can operate at a significantly larger postFEC bit error rate (BER) for the same postinvDM BER performance, which facilitates simulations. These benefits come at the cost of a slightly larger rate loss and required signaltonoise ratio at a given postFEC BER. 
Hierarchical Dynamic Loop SelfScheduling Techniques (Hierarchical DLS) 
Computationallyintensive loops are the primary source of parallelism in scientific applications. Such loops are often irregular and a balanced execution of their loop iterations is critical for achieving high performance. However, several factors may lead to an imbalanced load execution, such as problem characteristics, algorithmic, and systemic variations. Dynamic loop selfscheduling (DLS) techniques are devised to mitigate these factors, and consequently, improve application performance. On distributedmemory systems, DLS techniques can be implemented using a hierarchical masterworker execution model and are, therefore, called hierarchical Dynamic loop selfscheduling techniques. These techniques selfschedule loop iterations at two levels of hardware parallelism: across and within compute nodes. Hybrid programming approaches that combine the message passing interface (MPI) with open multiprocessing (OpenMP) dominate the implementation of hierarchical DLS techniques. The MPI3 standard includes the feature of sharing memory regions among MPI processes. This feature introduced the MPI+MPI approach that simplifies the implementation of parallel scientific applications. The present work designs and implements hierarchical DLS techniques by exploiting the MPI+MPI approach. Four wellknown DLS techniques are considered in the evaluation proposed herein. The results indicate certain performance advantages of the proposed approach compared to the hybrid MPI+OpenMP approach. 
Hierarchical Extreme Learning Machine (HELM) 
Complex industrial systems are continuously monitored by a large number of heterogenous sensors. The diversity of their operating conditions and the possible fault types make it impossible to collect enough data for learning all the possible fault patterns. The paper proposes an integrated automatic unsupervised feature learning approach for fault detection that uses healthy conditions data only for its training. The approach is based on stacked Extreme Learning Machines (namely Hierarchical, or HELM) and comprises stacked autoencoders performing unsupervised feature learning, and a oneclass classifier monitoring the variations in the features to assess the health of the system. This study provides a comprehensive evaluation of HELM fault detection capability compared to other machine learning approaches, including Deep Belief Networks. The performance is first evaluated on a synthetic dataset with typical characteristics of condition monitoring data. Subsequently, the approach is evaluated on a real case study of a power plant fault. HELM demonstrates a better performance specifically in cases where several noninformative signals are included. 
Hierarchical Gated Recurrent Unit (HiGRU) 
In this paper, we address three challenges in utterancelevel emotion recognition in dialogue systems: (1) the same word can deliver different emotions in different contexts; (2) some emotions are rarely seen in general dialogues; (3) longrange contextual information is hard to be effectively captured. We therefore propose a hierarchical Gated Recurrent Unit (HiGRU) framework with a lowerlevel GRU to model the wordlevel inputs and an upperlevel GRU to capture the contexts of utterancelevel embeddings. Moreover, we promote the framework to two variants, HiGRU with individual features fusion (HiGRUf) and HiGRU with selfattention and features fusion (HiGRUsf), so that the word/utterancelevel individual inputs and the longrange contextual information can be sufficiently utilized. Experiments on three dialogue emotion datasets, IEMOCAP, Friends, and EmotionPush demonstrate that our proposed HiGRU models attain at least 8.7%, 7.5%, 6.0% improvement over the stateoftheart methods on each dataset, respectively. Particularly, by utilizing only the textual feature in IEMOCAP, our HiGRU models gain at least 3.8% improvement over the stateoftheart conversational memory network (CMN) with the trimodal features of text, video, and audio. 
Hierarchical Importance Weighted Autoencoder  Importance weighted variational inference (Burda et al., 2015) uses multiple i.i.d. samples to have a tighter variational lower bound. We believe a joint proposal has the potential of reducing the number of redundant samples, and introduce a hierarchical structure to induce correlation. The hope is that the proposals would coordinate to make up for the error made by one another to reduce the variance of the importance estimator. Theoretically, we analyze the condition under which convergence of the estimator variance can be connected to convergence of the lower bound. Empirically, we confirm that maximization of the lower bound does implicitly minimize variance. Further analysis shows that this is a result of negative correlation induced by the proposed hierarchical meta sampling scheme, and performance of inference also improves when the number of samples increases. 
Hierarchical Incremental GRAdient Descent (HiGrad) 
Hierarchical Incremental GRAdient Descent (HiGrad) algorithm, a firstorder algorithm for finding the minimizer of a function in online learning just like stochastic gradient descent (SGD). See Su and Zhu (2018) <arXiv:1802.04876> for details. higrad 
Hierarchical Inference Testing (HIT) 
hit 
Hierarchical Kernel Learning (HKL) 
http://…/jawanpuria15a.pdf 
Hierarchical Latent Dirichlet Allocation (HLDA, HLDA) 
An extension to LDA is the hierarchical LDA (hLDA), where topics are joined together in a hierarchy by using the nested Chinese restaurant process. http://…/automatictopicmodellingwithlda 
Hierarchical Latent Space Network Model (HLSM) 
HLSM 
Hierarchical Latent Tree Analysis (HLTA) 
In the LDA approach to topic detection, a topic is determined by identifying the words that are used with high frequency when writing about the topic. However, high frequency words in one topic may be also used with high frequency in other topics. Thus they may not be the best words to characterize the topic. In this paper, we propose a new method for topic detection, where a topic is determined by identifying words that appear with high frequency in the topic and low frequency in other topics. We model patterns of word cooccurrence and cooccurrences of those patterns using a hierarchy of discrete latent variables. The states of the latent variables represent clusters of documents and they are interpreted as topics. The words that best distinguish a cluster from other clusters are selected to characterize the topic. Empirical results show that the new method yields topics with clearer thematic characterizations than the alternative approaches. In this work, we introduce semantically higher level latent variables to model cooccurrence of those patterns, resulting in hierarchical latent tree models (HLTMs). The latent variables at higher levels of the hierarchy correspond to more general topics, while the latent variables at lower levels correspond to more specific topics. The proposed method for topic detection is therefore called hierarchical latent tree analysis (HLTA). 
Hierarchical Latent Tree Model (HLTM) 

Hierarchical Long ShortTerm Concurrent Memory (HLSTCM) 
In this paper, we aim to address the problem of human interaction recognition in videos by exploring the longterm interrelated dynamics among multiple persons. Recently, Long ShortTerm Memory (LSTM) has become a popular choice to model individual dynamic for singleperson action recognition due to its ability of capturing the temporal motion information in a range. However, existing RNN models focus only on capturing the dynamics of human interaction by simply combining all dynamics of individuals or modeling them as a whole. Such models neglect the interrelated dynamics of how human interactions change over time. To this end, we propose a novel Hierarchical Long ShortTerm Concurrent Memory (HLSTCM) to model the longterm interrelated dynamics among a group of persons for recognizing the human interactions. Specifically, we first feed each person’s static features into a SinglePerson LSTM to learn the singleperson dynamic. Subsequently, the outputs of all SinglePerson LSTM units are fed into a novel Concurrent LSTM (CoLSTM) unit, which mainly consists of multiple submemory units, a new cell gate and a new comemory cell. In a CoLSTM unit, each submemory unit stores individual motion information, while this CoLSTM unit selectively integrates and stores interrelated motion information between multiple interacting persons from multiple submemory units via the cell gate and comemory cell, respectively. Extensive experiments on four public datasets validate the effectiveness of the proposed HLSTCM by comparing against baseline and stateoftheart methods. 
Hierarchical LSTM With Adaptive Attention (hLSTMat) 
Recent progress has been made in using attention based encoderdecoder framework for image and video captioning. Most existing decoders apply the attention mechanism to every generated word including both visual words (e.g., ‘gun’ and ‘shooting’) and nonvisual words (e.g. ‘the’, ‘a’). However, these nonvisual words can be easily predicted using natural language model without considering visual signals or attention. Imposing attention mechanism on nonvisual words could mislead and decrease the overall performance of visual captioning. Furthermore, the hierarchy of LSTMs enables more complex representation of visual data, capturing information at different scales. To address these issues, we propose a hierarchical LSTM with adaptive attention (hLSTMat) approach for image and video captioning. Specifically, the proposed framework utilizes the spatial or temporal attention for selecting specific regions or frames to predict the related words, while the adaptive attention is for deciding whether to depend on the visual information or the language context information. Also, a hierarchical LSTMs is designed to simultaneously consider both lowlevel visual information and highlevel language context information to support the caption generation. We initially design our hLSTMat for video captioning task. Then, we further refine it and apply it to image captioning task. To demonstrate the effectiveness of our proposed framework, we test our method on both video and image captioning tasks. Experimental results show that our approach achieves the stateoftheart performance for most of the evaluation metrics on both tasks. The effect of important components is also well exploited in the ablation study. 
Hierarchical LSTMs for Contextual Emotion Detection (HRLCE) 
This paper describes the system submitted by ANA Team for the SemEval2019 Task 3: EmoContext. We propose a novel Hierarchical LSTMs for Contextual Emotion Detection (HRLCE) model. It classifies the emotion of an utterance given its conversational context. The results show that, in this task, our HRCLE outperforms the most recent stateoftheart text classification framework: BERT. We combine the results generated by BERT and HRCLE to achieve an overall score of 0.7709 which ranked 5th on the final leader board of the competition among 165 Teams. 
Hierarchical Measure Group and Approximate System (HMGAS) 
We present a formal measuretheoretical theory of neural networks (NN) built on probability coupling theory. Our main contributions are summarized as follows. * Built on the formalism of probability coupling theory, we derive an algorithm framework, named Hierarchical Measure Group and Approximate System (HMGAS), nicknamed SSystem, that is designed to learn the complex hierarchical, statistical dependency in the physical world. * We show that NNs are special cases of SSystem when the probability kernels assume certain exponential family distributions. Activation Functions are derived formally. We further endow geometry on NNs through information geometry, show that intermediate feature spaces of NNs are stochastic manifolds, and prove that ‘distance’ between samples is contracted as layers stack up. * SSystem shows NNs are inherently stochastic, and under a set of realistic boundedness and diversity conditions, it enables us to prove that for large size nonlinear deep NNs with a class of losses, including the hinge loss, all local minima are global minima with zero loss errors, and regions around the minima are flat basins where all eigenvalues of Hessians are concentrated around zero, using tools and ideas from mean field theory, random matrix theory, and nonlinear operator equations. * SSystem, the informationgeometry structure and the optimization behaviors combined completes the analog between Renormalization Group (RG) and NNs. It shows that a NN is a complex adaptive system that estimates the statistic dependency of microscopic object, e.g., pixels, in multiple scales. Unlike clearcut physical quantity produced by RG in physics, e.g., temperature, NNs renormalize/recompose manifolds emerging through learning/optimization that divide the sample space into highly semantically meaningful groups that are dictated by supervised labels (in supervised NNs). 
Hierarchical Methods of Moments  Spectral methods of moments provide a powerful tool for learning the parameters of latent variable models. Despite their theoretical appeal, the applicability of these methods to real data is still limited due to a lack of robustness to model misspecification. In this paper we present a hierarchical approach to methods of moments to circumvent such limitations. Our method is based on replacing the tensor decomposition step used in previous algorithms with approximate joint diagonalization. Experiments on topic modeling show that our method outperforms previous tensor decomposition methods in terms of speed and model quality. 
Hierarchical Mode Association Clustering / Mode Association Clustering (HMAC, MAC) 
Mode association clustering (MAC) can be conducted either hierarchically or at one level. MAC is similar to mixture model based clustering in the sense of characterizing clusters by smooth densities. However, MAC requires no model fitting and uses a nonparametric kernel density estimation. The density of a cluster is not restricted to be parametric, for instance, Gaussian, but ensures unimodality. The algorithm seems to combine the complementary merits of bottomup clustering such as linkage and topdown clustering such as mixture modeling and kmeans. It also tends to be robust against nonGaussian shaped clusters. 
Hierarchical Model  There isn’t a single authorative definition of a hierarchical model. Click for an overview. 
Hierarchical Multiagent Teaching  Heterogeneous knowledge naturally arises among different agents in cooperative multiagent reinforcement learning. As such, learning can be greatly improved if agents can effectively pass their knowledge on to other agents. Existing work has demonstrated that peertopeer knowledge transfer, a process referred to as action advising, improves teamwide learning. In contrast to previous frameworks that advise at the level of primitive actions, we aim to learn highlevel teaching policies that decide when and what highlevel action (e.g., subgoal) to advise a teammate. We introduce a new learning to teach framework, called hierarchical multiagent teaching (HMAT). The proposed framework solves difficulties faced by prior work on multiagent teaching when operating in domains with long horizons, delayed rewards, and continuous states/actions by leveraging temporal abstraction and deep function approximation. Our empirical evaluations show that HMAT accelerates teamwide learning progress in difficult environments that are more complex than those explored in previous work. HMAT also learns teaching policies that can be transferred to different teammates/tasks and can even teach teammates with heterogeneous action spaces. 
Hierarchical Multinomial Marginal Models (HMM) 
In the loglinear parametrization all the interactions are contrasts of logarithms of joint probabilities and this is the main reason why this parametrization is not convenient to express hypotheses on marginal distributions or to model ordered categorical data. On the contrary Hierarchical Multinomial Marginal models (HMM) (Bartolucci et al. 2007) are based on parameters, called generalized marginal interactions, which are contrasts of logarithms of sums of probabilities. HMM models allow great flexibility in choosing the marginal distributions, within which the interactions are defined, and they are a useful tool for modeling marginal distributions and for taking into proper account the presence of ordinal categorical variables. hmmm 
Hierarchical Multiscale LSTM  Hierarchical Multiscale LSTM (Chung et al., 2016a) is a stateoftheart language model that learns interpretable structure from characterlevel input. 
Hierarchical MultiTask Learning Model (HMTL) 
Much effort has been devoted to evaluate whether multitask learning can be leveraged to learn rich representations that can be used in various Natural Language Processing (NLP) downstream applications. However, there is still a lack of understanding of the settings in which multitask learning has a significant effect. In this work, we introduce a hierarchical model trained in a multitask learning setup on a set of carefully selected semantic tasks. The model is trained in a hierarchical fashion to introduce an inductive bias by supervising a set of low level tasks at the bottom layers of the model and more complex tasks at the top layers of the model. This model achieves stateoftheart results on a number of tasks, namely Named Entity Recognition, Entity Mention Detection and Relation Extraction without handengineered features or external NLP tools like syntactic parsers. The hierarchical training supervision induces a set of shared semantic representations at lower layers of the model. We show that as we move from the bottom to the top layers of the model, the hidden states of the layers tend to represent more complex semantic information. 
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 Knearest neighbor queries. The Hierarchical NSW is a fully graphbased approach without a need for additional search structures (such as kdtrees 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 stateofart vectoronly approaches such as FLANN, FALCONN and Annoy. Similarity of the algorithm to a wellknown 1D skip list structure allows straightforward efficient and balanced distributed implementation. A Comparative Study on Hierarchical Navigable Small World Graphs 
Hierarchical Navigation Reinforcement Network (HNRN) 
This paper proposes a navigation algorithm oriented to multiagent dynamic environment. The algorithm is expressed as a hierarchical framework which contains a Hidden Markov Model (HMM) and Deep Reinforcement Learning (DRL). For simplification, we term our method Hierarchical Navigation Reinforcement Network (HNRN). In highlevel architecture, we train an HMM to evaluate agents environment in order to obtain a score. According to this score, adaptive control action will be chosen. While in lowlevel architecture, two subsystems are introduced, one is a differential targetdriven system, which aims at heading to the target, the other is collision avoidance DRL system, which is used for avoiding obstacles in the dynamic environment. The advantage of this hierarchical system is to decouple the targetdriven and collision avoidance tasks, leading to a faster and easier model to be trained. As the experiments manifest, our algorithm has faster learning efficiency and a higher success rate than traditional Velocity Obstacle (VO) algorithms and hybrid DRL method. 
Hierarchical Nearest Neighbor Descent (HNND) 

Hierarchical Network  A hierarchical network is the type of network topology in which a central “root” node (the top level of the hierarchy) is connected to one or more other nodes that are one level lower in the hierarchy (i.e., the second level) with a pointtopoint link between each of the second level nodes and the top level central “root” node, while each of the second level nodes that are connected to the top level central “root” node will also have one or more other nodes that are one level lower in the hierarchy (i.e., the third level) connected to it, also with a pointtopoint link, the top level central “root” node being the only node that has no other node above it in the hierarchy. 
Hierarchical Network Model (HNM) 
Hierarchical network models are iterative algorithms for creating networks which are able to reproduce the unique properties of the scalefree topology and the high clustering of the nodes at the same time. These characteristics are widely observed in nature, from biology to language to some social networks. 
Hierarchical Planning and Reinforcement Learning (HIPRL) 
Longterm planning poses a major difficulty to many reinforcement learning algorithms. This problem becomes even more pronounced in dynamic visual environments. In this work we propose Hierarchical Planning and Reinforcement Learning (HIPRL), a method for merging the benefits and capabilities of Symbolic Planning with the learning abilities of Deep Reinforcement Learning. We apply HIPRL to the complex visual tasks of interactive question answering and visual semantic planning and achieve stateoftheart results on three challenging datasets all while taking fewer steps at test time and training in fewer iterations. Sample results can be found at youtu.be/0TtWJ_0mPfI 
Hierarchical Recurrent EncoderDecoder (HRED) 
As a generative model for building endtoend dialogue systems, Hierarchical Recurrent EncoderDecoder (HRED) consists of three layers of Gated Recurrent Unit (GRU), which from bottom to top are separately used as the wordlevel encoder, the sentencelevel encoder, and the decoder. Despite performing well on dialogue corpora, HRED is computationally expensive to train due to its complexity. To improve the training efficiency of HRED, we propose a new model, which is named as Simplified HRED (SHRED), by making each layer of HRED except the top one simpler than its upper layer. On the one hand, we propose Scalar Gated Unit (SGU), which is a simplified variant of GRU, and use it as the sentencelevel encoder. On the other hand, we use Fixedsize OrdinallyForgetting Encoding (FOFE), which has no trainable parameter at all, as the wordlevel encoder. The experimental results show that compared with HRED under the same word embedding size and the same hidden state size for each layer, SHRED reduces the number of trainable parameters by 25\%–35\%, and the training time by more than 50\%, but still achieves slightly better performance. 
Hierarchical Recurrent Neural Network (HRNN) 
Exploiting the temporal dependency among video frames or subshots is very important for the task of video summarization. Practically, RNN is good at temporal dependency modeling, and has achieved overwhelming performance in many videobased tasks, such as video captioning and classification. However, RNN is not capable enough to handle the video summarization task, since traditional RNNs, including LSTM, can only deal with short videos, while the videos in the summarization task are usually in longer duration. To address this problem, we propose a hierarchical recurrent neural network for video summarization, called HRNN in this paper. Specifically, it has two layers, where the first layer is utilized to encode short video subshots cut from the original video, and the final hidden state of each subshot is input to the second layer for calculating its confidence to be a key subshot. Compared to traditional RNNs, HRNN is more suitable to video summarization, since it can exploit long temporal dependency among frames, meanwhile, the computation operations are significantly lessened. The results on two popular datasets, including the Combined dataset and VTW dataset, have demonstrated that the proposed HRNN outperforms the stateofthearts. 
Hierarchical Reinforcement Learning (HRL) 

Hierarchical Reinforcement Learning Algorithm via MultiGoals Abstraction (HRLMG) 
The recommender system is an important form of intelligent application, which assists users to alleviate from information redundancy. Among the metrics used to evaluate a recommender system, the metric of conversion has become more and more important. The majority of existing recommender systems perform poorly on the metric of conversion due to its extremely sparse feedback signal. To tackle this challenge, we propose a deep hierarchical reinforcement learning based recommendation framework, which consists of two components, i.e., highlevel agent and lowlevel agent. The highlevel agent catches longterm sparse conversion signals, and automatically sets abstract goals for lowlevel agent, while the lowlevel agent follows the abstract goals and interacts with realtime environment. To solve the inherent problem in hierarchical reinforcement learning, we propose a novel deep hierarchical reinforcement learning algorithm via multigoals abstraction (HRLMG). Our proposed algorithm contains three characteristics: 1) the highlevel agent generates multiple goals to guide the lowlevel agent in different stages, which reduces the difficulty of approaching highlevel goals; 2) different goals share the same state encoder parameters, which increases the update frequency of the highlevel agent and thus accelerates the convergence of our proposed algorithm; 3) an appreciate benefit assignment function is designed to allocate rewards in each goal so as to coordinate different goals in a consistent direction. We evaluate our proposed algorithm based on a realworld ecommerce dataset and validate its effectiveness. 
Hierarchical Representation Learning on Heterogeneous Graph (HRLHG) 
While the volume of scholarly publications has increased at a frenetic pace, accessing and consuming the useful candidate papers, in very large digital libraries, is becoming an essential and challenging task for scholars. Unfortunately, because of language barrier, some scientists (especially the junior ones or graduate students who do not master other languages) cannot efficiently locate the publications hosted in a foreign language repository. In this study, we propose a novel solution, crosslanguage citation recommendation via Hierarchical Representation Learning on Heterogeneous Graph (HRLHG), to address this new problem. HRLHG can learn a representation function by mapping the publications, from multilingual repositories, to a lowdimensional joint embedding space from various kinds of vertexes and relations on a heterogeneous graph. By leveraging both global (task specific) plus local (task independent) information as well as a novel supervised hierarchical random walk algorithm, the proposed method can optimize the publication representations by maximizing the likelihood of locating the important crosslanguage neighborhoods on the graph. Experiment results show that the proposed method can not only outperform stateoftheart baseline models, but also improve the interpretability of the representation model for crosslanguage citation recommendation task. 
Hierarchical Routing Mixture of Experts (HRME) 
In regression tasks the distribution of the data is often too complex to be fitted by a single model. In contrast, partitionbased models are developed where data is divided and fitted by local models. These models partition the input space and do not leverage the inputoutput dependency of multimodaldistributed data, and strong local models are needed to make good predictions. Addressing these problems, we propose a binary treestructured hierarchical routing mixture of experts (HRME) model that has classifiers as nonleaf node experts and simple regression models as leaf node experts. The classifier nodes jointly softpartition the inputoutput space based on the natural separateness of multimodal data. This enables simple leaf experts to be effective for prediction. Further, we develop a probabilistic framework for the HRME model, and propose a recursive ExpectationMaximization (EM) based algorithm to learn both the tree structure and the expert models. Experiments on a collection of regression tasks validate the effectiveness of our method compared to a variety of other regression models. 
Hierarchical Semantic Embedding (HSE) 
Object categories inherently form a hierarchy with different levels of concept abstraction, especially for finegrained categories. For example, birds (Aves) can be categorized according to a fourlevel hierarchy of order, family, genus, and species. This hierarchy encodes rich correlations among various categories across different levels, which can effectively regularize the semantic space and thus make prediction less ambiguous. However, previous studies of finegrained image recognition primarily focus on categories of one certain level and usually overlook this correlation information. In this work, we investigate simultaneously predicting categories of different levels in the hierarchy and integrating this structured correlation information into the deep neural network by developing a novel Hierarchical Semantic Embedding (HSE) framework. Specifically, the HSE framework sequentially predicts the category score vector of each level in the hierarchy, from highest to lowest. At each level, it incorporates the predicted score vector of the higher level as prior knowledge to learn finergrained feature representation. During training, the predicted score vector of the higher level is also employed to regularize label prediction by using it as soft targets of corresponding subcategories. To evaluate the proposed framework, we organize the 200 bird species of the CaltechUCSD birds dataset with the fourlevel category hierarchy and construct a largescale butterfly dataset that also covers four level categories. Extensive experiments on these two and the newlyreleased VegFru datasets demonstrate the superiority of our HSE framework over the baseline methods and existing competitors. 
Hierarchical Spectral Merger (HSM) 
We present a new method for time series clustering which we call the Hierarchical Spectral Merger (HSM) method. This procedure is based on the spectral theory of time series and identifies series that share similar oscillations or waveforms. The extent of similarity between a pair of time series is measured using the total variation distance between their estimated spectral densities. At each step of the algorithm, every time two clusters merge, a new spectral density is estimated using the whole information present in both clusters, which is representative of all the series in the new cluster. The method is implemented in an R package HSMClust. We present two applications of the HSM method, one to data coming from waveheight measurements in oceanography and the other to electroencefalogram (EEG) data. 
Hierarchical Stochastic Clustering (HSC) 
Hierarchical clustering is one of the most powerful solutions to the problem of clustering, on the grounds that it performs a multi scale organization of the data. In recent years, research on hierarchical clustering methods has attracted considerable interest due to the demanding modern application domains. We present a novel divisive hierarchical clustering framework called Hierarchical Stochastic Clustering (HSC), that acts in two stages. In the first stage, it finds a primary hierarchy of clustering partitions in a dataset. In the second stage, feeds a clustering algorithm with each one of the clusters of the very detailed partition, in order to settle the final result. The output is a hierarchy of clusters. Our method is based on the previous research of Meyer and Weissel Stochastic Data Clustering and the theory of Simon and Ando on Variable Aggregation. Our experiments show that our framework builds a meaningful hierarchy of clusters and benefits consistently the clustering algorithm that acts in the second stage, not only computationally but also in terms of cluster quality. This result suggest that HSC framework is ideal for obtaining hierarchical solutions of large volumes of data. 
Hierarchical Temporal Convolutional Network (HierTCN) 
Recommender systems that can learn from crosssession data to dynamically predict the next item a user will choose are crucial for online platforms. However, existing approaches often use outofthebox sequence models which are limited by speed and memory consumption, are often infeasible for production environments, and usually do not incorporate crosssession information, which is crucial for effective recommendations. Here we propose Hierarchical Temporal Convolutional Networks (HierTCN), a hierarchical deep learning architecture that makes dynamic recommendations based on users’ sequential multisession interactions with items. HierTCN is designed for webscale systems with billions of items and hundreds of millions of users. It consists of two levels of models: The highlevel model uses Recurrent Neural Networks (RNN) to aggregate users’ evolving longterm interests across different sessions, while the lowlevel model is implemented with Temporal Convolutional Networks (TCN), utilizing both the longterm interests and the shortterm interactions within sessions to predict the next interaction. We conduct extensive experiments on a public XING dataset and a largescale Pinterest dataset that contains 6 million users with 1.6 billion interactions. We show that HierTCN is 2.5x faster than RNNbased models and uses 90% less data memory compared to TCNbased models. We further develop an effective data caching scheme and a queuebased minibatch generator, enabling our model to be trained within 24 hours on a single GPU. Our model consistently outperforms stateoftheart dynamic recommendation methods, with up to 18% improvement in recall and 10% in mean reciprocal rank. 
Hierarchical Temporal Memory (HTM) 
Hierarchical temporal memory (HTM) is a biologically constrained theory of machine intelligence originally described in the 2004 book On Intelligence by Jeff Hawkins with Sandra Blakeslee. HTM is based on neuroscience and the physiology and interaction of pyramidal neurons in the neocortex of the human brain. The technology has been tested and implemented in software through example applications from Numenta and commercial applications from Numenta’s partners. At the core of HTM are learning algorithms that can store, learn, infer and recall highorder sequences. Unlike most other machine learning methods, HTM learns timebased patterns in unlabeled data on a continuous basis. HTM is robust to noise and high capacity, meaning that it can learn multiple patterns simultaneously. When applied to computers, HTM is well suited for prediction, anomaly detection, classification and ultimately sensorimotor applications. 
Hierarchical Time Series / Grouped Time Series (HTS) 
Time series can often be naturally disaggregated in a hierarchical structure using attributes such as geographical location, product type, etc. For example, the total number of bicycles sold by a cycling warehouse can be disaggregated into a hierarchy of bicycle types. Such a warehouse will sell road bikes, mountain bikes, children bikes or hybrids. Each of these can be disaggregated into finer categories. Children’s bikes can be divided into balance bikes for children under 4 years old, single speed bikes for children between 4 and 6 and bikes for children over the age of 6. Hybrid bikes can be divided into city, commuting, comfort, and trekking bikes; and so on. Such disaggregation imposes a hierarchical structure. We refer to these as hierarchical time series. hts,gtop 
Hierarchical Topic Models  
Hierarchically Self Decomposing CNN  Conventional Convolutional neural networks (CNN) are trained on large domain datasets, and are hence typically overrepresented and inefficient in limited class applications. An efficient way to convert such large manyclass pretrained networks into small fewclass networks is through a hierarchical decomposition of its feature maps. To alleviate this issue, we propose an automated framework for such decomposition in Hierarchically Self Decomposing CNNs (HSDCNN), in four steps. HSDCNNs are derived automatically using a class specific filter sensitivity analysis that quantifies the impact of specific features on a class prediction. The decomposed and hierarchical network can be utilized and deployed directly to obtain subnetworks for subset of classes, and it is shown to perform better without the requirement of retraining these subnetworks. Experimental results show that HSDCNNs generally do not degrade accuracy if the full set of classes are used. However, when operating on known subsets of classes, HSDCNNs lead to an increased accuracy using a much smaller model size, requiring much less operations. HSDCNN flow is verified on the CIFAR10, CIFAR100 and CALTECH101 data sets. We report accuracies up to $85.6\%$ ( $94.75\%$ ) on scenarios with 13 ( 4 ) classes of CIFAR100, using a VGG16 network pretrained on the full data set. In this case, the used HSDCNN requires $3.97 \times$ fewer parameters and $3.56 \times$ fewer operations than the VGG16 baseline containing features for all 100 classes. 
Hierarchically Structured MetaLearning (HSML) 
In order to learn quickly with few samples, metalearning utilizes prior knowledge learned from previous tasks. However, a critical challenge in metalearning is task uncertainty and heterogeneity, which can not be handled via globally sharing knowledge among tasks. In this paper, based on gradientbased metalearning, we propose a hierarchically structured metalearning (HSML) algorithm that explicitly tailors the transferable knowledge to different clusters of tasks. Inspired by the way human beings organize knowledge, we resort to a hierarchical task clustering structure to cluster tasks. As a result, the proposed approach not only addresses the challenge via the knowledge customization to different clusters of tasks, but also preserves knowledge generalization among a cluster of similar tasks. To tackle the changing of task relationship, in addition, we extend the hierarchical structure to a continual learning environment. The experimental results show that our approach can achieve stateoftheart performance in both toyregression and fewshot image classification problems. 
Hierarchically Supervised Latent Dirichlet Allocation (HSLDA) 
We introduce hierarchically supervised latent Dirichlet allocation (HSLDA), a model for hierarchically and multiply labeled bagofword data. Examples of such data include web pages and their placement in directories, product descriptions and associated categories from product hierarchies, and freetext clinical records and their assigned diagnosis codes. Outofsample label prediction is the primary goal of this work, but improved lowerdimensional representations of the bagofword data are also of interest. We demonstrate HSLDA on largescale data from clinical document labeling and retail product categorization tasks. We show that leveraging the structure from hierarchical labels improves outofsample label prediction substantially when compared to models that do not. 
HierarchicallyClustered Representation Learning (HCRL) 
The joint optimization of representation learning and clustering in the embedding space has experienced a breakthrough in recent years. In spite of the advance, clustering with representation learning has been limited to flatlevel categories, which often involves cohesive clustering with a focus on instance relations. To overcome the limitations of flat clustering, we introduce hierarchicallyclustered representation learning (HCRL), which simultaneously optimizes representation learning and hierarchical clustering in the embedding space. Compared with a few prior works, HCRL firstly attempts to consider a generation of deep embeddings from every component of the hierarchy, not just leaf components. In addition to obtaining hierarchically clustered embeddings, we can reconstruct data by the various abstraction levels, infer the intrinsic hierarchical structure, and learn the levelproportion features. We conducted evaluations with image and text domains, and our quantitative analyses showed competent likelihoods and the best accuracies compared with the baselines. 
HierarchicalMeta Learning (HML) 
Meta learning is a promising solution to fewshot learning problems. However, existing meta learning methods are restricted to the scenarios where training and application tasks share the same output structure. To obtain a meta model applicable to the tasks with new structures, it is required to collect new training data and repeat the timeconsuming meta training procedure. This makes them inefficient or even inapplicable in learning to solve heterogeneous fewshot learning tasks. We thus develop a novel and principled HierarchicalMeta Learning (HML) method. Different from existing methods that only focus on optimizing the adaptability of a meta model to similar tasks, HML also explicitly optimizes its generalizability across heterogeneous tasks. To this end, HML first factorizes a set of similar training tasks into heterogeneous ones and trains the meta model over them at two levels to maximize adaptation and generalization performance respectively. The resultant model can then directly generalize to new tasks. Extensive experiments on fewshot classification and regression problems clearly demonstrate the superiority of HML over finetuning and stateoftheart meta learning approaches in terms of generalization across heterogeneous tasks. 
HierLPR  In this article we propose a novel ranking algorithm, referred to as HierLPR, for the multilabel classification problem when the candidate labels follow a known hierarchical structure. HierLPR is motivated by a new metric called eAUC that we design to assess the ranking of classification decisions. This metric, associated with the hit curve and local precision rate, emphasizes the accuracy of the first calls. We show that HierLPR optimizes eAUC under the tree constraint and some light assumptions on the dependency between the nodes in the hierarchy. We also provide a strategy to make calls for each node based on the ordering produced by HierLPR, with the intent of controlling FDR or maximizing Fscore. The performance of our proposed methods is demonstrated on synthetic datasets as well as a real example of disease diagnosis using NCBI GEO datasets. In these cases, HierLPR shows a favorable result over competing methods in the early part of the precisionrecall curve. 
High Dimensional Data Clustering (HDDC) 
Clustering in highdimensional spaces is a recurrent problem in many domains, for example in object recognition. Highdimensional data usually live in different lowdimensional subspaces hidden in the original space. HDDC is a clustering approach which estimates the specific subspace and the intrinsic dimension of each class. The approach adapts the Gaussian mixture model framework to highdimensional data and estimates the parameters which best fit the data. This results in a robust clustering method called High Dimensional Data Clustering (HDDC). HDDC is applied to locate objects in natural images in a probabilistic framework. Experiments on a recently proposed database demonstrate the effectiveness of our clustering method for category localization. 
High Dimensional Linear GMM  This paper proposes a desparsified GMM estimator for estimating highdimensional regression models allowing for, but not requiring, many more endogenous regressors than observations. We provide finite sample upper bounds on the estimation error of our estimator and show how asymptotically uniformly valid inference can be conducted in the presence of conditionally heteroskedastic error terms. We do not require the projection of the endogenous variables onto the linear span of the instruments to be sparse; that is we do not impose the instruments to be sparse for our inferential procedure to be asymptotically valid. Furthermore, the variables of the model are not required to be subgaussian and we also explain how our results carry over to the classic dynamic linear panel data model. Simulations show that our estimator has a low mean square error and does well in terms of size and power of the tests constructed based on the estimator. 
High Frequency Trading (HFT) 
Highfrequency trading (HFT) is a primary form of algorithmic trading in finance. Specifically, it is the use of sophisticated technological tools and computer algorithms to rapidly trade securities. HFT uses proprietary trading strategies carried out by computers to move in and out of positions in seconds or fractions of a second. It is estimated that as of 2009, HFT accounted for 6073% of all US equity trading volume, with that number falling to approximately 50% in 2012. Highfrequency traders move in and out of shortterm positions at high volumes aiming to capture sometimes a fraction of a cent in profit on every trade. HFT firms do not consume significant amounts of capital, accumulate positions or hold their portfolios overnight. As a result, HFT has a potential Sharpe ratio (a measure of risk and reward) tens of times higher than traditional buyandhold strategies. Highfrequency traders typically compete against other HFTs, rather than longterm investors. HFT firms make up the low margins with incredible high volumes of tradings, frequently numbering in the millions. It has been argued that a core incentive in much of the technological development behind highfrequency trading is essentially front running, in which the varying delays in the propagation of orders is taken advantage of by those who have earlier access to information. A substantial body of research argues that HFT and electronic trading pose new types of challenges to the financial system. Algorithmic and highfrequency traders were both found to have contributed to volatility in the May 6, 2010 Flash Crash, when highfrequency liquidity providers rapidly withdrew from the market. Several European countries have proposed curtailing or banning HFT due to concerns about volatility. Other complaints against HFT include the argument that some HFT firms scrape profits from investors when index funds rebalance their portfolios. Other financial analysts point to evidence of benefits that HFT has brought to the modern markets. Researchers have stated that HFT and automated markets improve market liquidity, reduce trading costs, and make stock prices more efficient. 
High Performance Analytics Toolkit (HPAT) 
Big data analytics requires high programmer productivity and high performance simultaneously on largescale clusters. However, current big data analytics frameworks (e.g. Apache Spark) have high runtime overheads since they are librarybased. Given the characteristics of the data analytics domain, we introduce the High Performance Analytics Toolkit (HPAT), which is a big data analytics framework that performs static compilation of highlevel scripting programs into high performance parallel code using novel domainspecific compilation techniques. HPAT provides scripting abstractions in the Julia language for analytics tasks, automatically parallelizes them, generates efficient MPI/C++ code, and provides resiliency. Since HPAT is compilerbased, it avoids overheads of librarybased systems such as dynamic task scheduling and masterexecutor coordination. In addition, it provides automatic optimizations for scripting programs, such as fusion of array operations. Therefore, HPAT is 14x to 400x faster than Spark on the Cori supercomputer at LBL/NERSC. Furthermore, HPAT is much more flexible in distributed data structures, which enables the use of existing libraries such as HDF5, ScaLAPACK, and Intel R DAAL. 
High Performance Computing (HPC) 
High Performance Computing most generally refers to the practice of aggregating computing power in a way that delivers much higher performance than one could get out of a typical desktop computer or workstation in order to solve large problems in science, engineering, or business. A supercomputer is a computer with a very highlevel computational capacity. As of 2015, there are supercomputers which could perform upto quadrillions of floating point operations per second. http://…/Supercomputer 
High Quality Bidirectional Generative Adversarial Network  Generative adversarial networks (GANs) have achieved outstanding success in generating the high quality data. Focusing on the generation process, existing GANs investigate unidirectional mapping from the latent vector to the data. Later, various studies point out that the latent space of GANs is semantically meaningful and can be utilized in advanced data analysis and manipulation. In order to analyze the real data in the latent space of GANs, it is necessary to investigate the inverse generation mapping from the data to the latent vector. To tackle this problem, the bidirectional generative models introduce an encoder to enable the inverse path of generation process. Unfortunately, this effort leads to the degradation of generation quality because the imperfect generator rather interferes the encoder training and vice versa. In this paper, we propose a new inference model that estimates the latent vector from the feature of GAN discriminator. While existing bidirectional models learns the image to latent translation, our algorithm formulates this inference mapping by the feature to latent translation. It is important to note that training of our model is independent of the GAN training. Owing to the attractive nature of this independency, the proposed algorithm can generate the high quality samples identical to those of unidirectional GANs and also reconstruct the original data faithfully. Moreover, our algorithm can be employed to any unidirectional GAN, even the pretraind GANs. 
High Utility Occupancy Pattern Mining (HUOPM) 
Mining useful patterns from varied types of databases is an important research topic, which has many reallife applications. Most studies have considered the frequency as sole interestingness measure for identifying high quality patterns. However, each object is different in nature. The relative importance of objects is not equal, in terms of criteria such as the utility, risk, or interest. Besides, another limitation of frequent patterns is that they generally have a low occupancy, i.e., they often represent small sets of items in transactions containing many items, and thus may not be truly representative of these transactions. To extract high quality patterns in real life applications, this paper extends the occupancy measure to also assess the utility of patterns in transaction databases. We propose an efficient algorithm named High Utility Occupancy Pattern Mining (HUOPM). It considers user preferences in terms of frequency, utility, and occupancy. A novel FrequencyUtility tree (FUtree) and two compact data structures, called the utilityoccupancy list and FUtable, are designed to provide global and partial downward closure properties for pruning the search space. The proposed method can efficiently discover the complete set of high quality patterns without candidate generation. Extensive experiments have been conducted on several datasets to evaluate the effectiveness and efficiency of the proposed algorithm. Results show that the derived patterns are intelligible, reasonable and acceptable, and that HUOPM with its pruning strategies outperforms the stateoftheart algorithm, in terms of runtime and search space, respectively. 
Highcharts  Highcharts is a charting library written in pure JavaScript, offering an easy way of adding interactive charts to your web site or web application. Highcharts currently supports line, spline, area, areaspline, column, bar, pie, scatter, angular gauges, arearange, areasplinerange, columnrange, bubble, box plot, error bars, funnel, waterfall and polar chart types. 
Higher Order Propagation Framework (HOPF) 
Given a graph wherein every node has certain attributes associated with it and some nodes have labels associated with them, Collective Classification (CC) is the task of assigning labels to every unlabeled node using information from the node as well as its neighbors. It is often the case that a node is not only influenced by its immediate neighbors but also by its higher order neighbors, multiple hops away. Recent stateoftheart models for CC use differentiable variations of WeisfeilerLehman kernels to aggregate multihop neighborhood information. However, in this work, we show that these models suffer from the problem of Node Information Morphing wherein the information of the node is morphed or overwhelmed by the information of its neighbors when considering multiple hops. Further, existing models are not scalable as the memory and computation needs grow exponentially with the number of hops considered. To circumvent these problems, we propose a generic Higher Order Propagation Framework (HOPF) which includes (i) a differentiable Node Information Preserving (NIP) kernel and (ii) a scalable iterative learning and inferencing mechanism to aggregate information over larger hops. We do an extensive evaluation using 11 datasets from different domains and show that unlike existing CC models, our NIP model with iterative inference is robust across all the datasets and can handle much larger neighborhoods in a scalable manner. 
HigherOrder Generalized Singular Value Decomposition  hogsvdR 
HigherOrder Graph Convolutional Network  ➘ “Motif Convolutional Network” 
HigherOrder KolmogorovSmirnov Test  We present an extension of the KolmogorovSmirnov (KS) twosample test, which can be more sensitive to differences in the tails. Our test statistic is an integral probability metric (IPM) defined over a higherorder total variation ball, recovering the original KS test as its simplest case. We give an exact representer result for our IPM, which generalizes the fact that the original KS test statistic can be expressed in equivalent variational and CDF forms. For small enough orders ($k \leq 5$), we develop a lineartime algorithm for computing our higherorder KS test statistic; for all others ($k \geq 6$), we give a nearly lineartime approximation. We derive the asymptotic null distribution for our test, and show that our nearly lineartime approximation shares the same asymptotic null. Lastly, we complement our theory with numerical studies. 
Highest Density Interval Regression Forest (HDIForest) 
By seeking the narrowest prediction intervals (PIs) that satisfy the specified coverage probability requirements, the recently proposed qualitybased PI learning principle can extract highquality PIs that better summarize the predictive certainty in regression tasks, and has been widely applied to solve many practical problems. Currently, the stateoftheart qualitybased PI estimation methods are based on deep neural networks or linear models. In this paper, we propose Highest Density Interval Regression Forest (HDIForest), a novel qualitybased PI estimation method that is instead based on Random Forest. HDIForest does not require additional model training, and directly reuses the trees learned in a standard Random Forest model. By utilizing the special properties of Random Forest, HDIForest could efficiently and more directly optimize the PI quality metrics. Extensive experiments on benchmark datasets show that HDIForest significantly outperforms previous approaches, reducing the average PI width by over 30\% while achieving the same or better coverage probability. 
Highest Density Regions (HDR) 

Highest Posterior Density (HPD) 
Highest Posterior Density – The x% highest posterior density interval is the shortest interval in parameter space that contains x% of the posterior probability. 
Highly Efficient Network (HENet) 
In order to enhance the realtime performance of convolutional neural networks(CNNs), more and more researchers are focusing on improving the efficiency of CNN. Based on the analysis of some CNN architectures, such as ResNet, DenseNet, ShuffleNet and so on, we combined their advantages and proposed a very efficient model called Highly Efficient Networks(HENet). The new architecture uses an unusual way to combine group convolution and channel shuffle which was mentioned in ShuffleNet. Inspired by ResNet and DenseNet, we also proposed a new way to use elementwise addition and concatenation connection with each block. In order to make greater use of feature maps, pooling operations are removed from HENet. The experiments show that our model’s efficiency is more than 1 times higher than ShuffleNet on many open source datasets, such as CIFAR10/100 and SVHN. 
Highly Missing Data Lasso (HMLasso) 
Sparse regression such as Lasso has achieved great success in dealing with high dimensional data for several decades. However, there are few methods applicable to missing data, which often occurs in high dimensional data. Recently, CoCoLasso was proposed to deal with high dimensional missing data, but it still suffers from highly missing data. In this paper, we propose a novel Lassotype regression technique for Highly Missing data, called `HMLasso’. We use the mean imputed covariance matrix, which is notorious in general due to its estimation bias for missing data. However, we effectively incorporate it into Lasso, by using a useful connection with the pairwise covariance matrix. The resulting optimization problem can be seen as a weighted modification of CoCoLasso with the missing ratios, and is quite effective for highly missing data. To the best of our knowledge, this is the first method that can efficiently deal with both high dimensional and highly missing data. We show that the proposed method is beneficial with regards to nonasymptotic properties of the covariance matrix. Numerical experiments show that the proposed method is highly advantageous in terms of estimation error and generalization error. 
HighResolution Deep Convolutional Generative Adversarial Network (HRDCGAN) 
Generative Adversarial Networks (GANs) convergence in a highresolution setting with a computational constrain of GPU memory capacity (from 12GB to 24 GB) has been beset with difficulty due to the known lack of convergence rate stability. In order to boost network convergence of DCGAN (Deep Convolutional Generative Adversarial Networks) and achieve goodlooking highresolution results we propose a new layered network structure, HRDCGAN, that incorporates current stateoftheart techniques for this effect. 
HighUtility Sequential Pattern (HUSP) 
➘ “HighUtility Sequential Pattern Mining” 
HighUtility Sequential Pattern Mining (HUSPM) 
Highutility sequential pattern mining is an emerging topic in the field of Knowledge Discovery in Databases. It consists of discovering subsequences having a high utility (importance) in sequences, referred to as highutility sequential patterns (HUSPs). HUSPs can be applied to many reallife applications, such as market basket analysis, Ecommerce recommendation, clickstream analysis and scenic route planning. For example, in economics and targeted marketing, understanding economic behavior of consumers is quite challenging, such as finding credible and reliable information on product profitability. Several algorithms have been proposed to address this problem by efficiently mining utilitybased useful sequential patterns. Nevertheless, the performance of these algorithms can be unsatisfying in terms of runtime and memory usage due to the combinatorial explosion of the search space for low utility threshold and large databases. Hence, this paper proposes a more efficient algorithm for the task of highutility sequential pattern mining, called HUSPULL. It utilizes a lexicographic sequence (LS)tree and a utilitylinked (UL)list structure to fast discover HUSPs. Furthermore, two pruning strategies are introduced in HUSPULL to obtain tight upperbounds on the utility of candidate sequences, and reduce the search space by pruning unpromising candidates early. Substantial experiments both on reallife and synthetic datasets show that the proposed algorithm can effectively and efficiently discover the complete set of HUSPs and outperforms the stateoftheart algorithms. 
HilbertSchmidt Independence Criterion (HSIC) 
‘Dependency Bottleneck’ in Autoencoding Architectures: an Empirical Study 
Hill Climbing  In computer science, hill climbing is a mathematical optimization technique which belongs to the family of local search. It is an iterative algorithm that starts with an arbitrary solution to a problem, then attempts to find a better solution by incrementally changing a single element of the solution. If the change produces a better solution, an incremental change is made to the new solution, repeating until no further improvements can be found. For example, hill climbing can be applied to the travelling salesman problem. It is easy to find an initial solution that visits all the cities but will be very poor compared to the optimal solution. The algorithm starts with such a solution and makes small improvements to it, such as switching the order in which two cities are visited. Eventually, a much shorter route is likely to be obtained. Hill climbing is good for finding a local optimum (a solution that cannot be improved by considering a neighbouring configuration) but it is not necessarily guaranteed to find the best possible solution (the global optimum) out of all possible solutions (the search space). In convex problems, hillclimbing is optimal. Examples of algorithms that solve convex problems by hillclimbing include the simplex algorithm for linear programming and binary search. The characteristic that only local optima are guaranteed can be cured by using restarts (repeated local search), or more complex schemes based on iterations, like iterated local search, on memory, like reactive search optimization and tabu search, or memoryless stochastic modifications, like simulated annealing. The relative simplicity of the algorithm makes it a popular first choice amongst optimizing algorithms. It is used widely in artificial intelligence, for reaching a goal state from a starting node. Choice of next node and starting node can be varied to give a list of related algorithms. Although more advanced algorithms such as simulated annealing or tabu search may give better results, in some situations hill climbing works just as well. Hill climbing can often produce a better result than other algorithms when the amount of time available to perform a search is limited, such as with realtime systems. It is an anytime algorithm: it can return a valid solution even if it’s interrupted at any time before it ends. 
Hindcasting  In oceanography and meteorology, backtesting is also known as hindcasting: a hindcast is a way of testing a mathematical model; known or closely estimated inputs for past events are entered into the model to see how well the output matches the known results. Hindcasting usually refers to a numerical model integration of a historical period where no observations have been assimilated. This distinguishes a hindcast run from a reanalysis. Oceanographic observations of salinity and temperature as well as observations of surface wave parameters such as the significant wave height are much scarcer than meteorological observations, making hindcasting more common in oceanography than in meteorology. Also, since surface waves represent a forced system where the wind is the only generating force, wave hindcasting is often considered adequate for generating a reasonable representation of the wave climate with little need for a full reanalysis. Hindcasting is also used in hydrology for model stream flows. 
HindmarshRose Neural Network  Spatial evolution of HindmarshRose neural network with time delays 
Hindsight Generative Adversarial Imitation Learning (HGAIL) 
Compared to reinforcement learning, imitation learning (IL) is a powerful paradigm for training agents to learn control policies efficiently from expert demonstrations. However, in most cases, obtaining demonstration data is costly and laborious, which poses a significant challenge in some scenarios. A promising alternative is to train agent learning skills via imitation learning without expert demonstrations, which, to some extent, would extremely expand imitation learning areas. To achieve such expectation, in this paper, we propose Hindsight Generative Adversarial Imitation Learning (HGAIL) algorithm, with the aim of achieving imitation learning satisfying no need of demonstrations. Combining hindsight idea with the generative adversarial imitation learning (GAIL) framework, we realize implementing imitation learning successfully in cases of expert demonstration data are not available. Experiments show that the proposed method can train policies showing comparable performance to current imitation learning methods. Further more, HGAIL essentially endows curriculum learning mechanism which is critical for learning policies. 
Hinted Network  We present Hinted Networks: a collection of architectural transformations for improving the accuracies of neural network models for regression tasks, through the injection of a prior for the output prediction (i.e. a hint). We ground our investigations within the camera relocalization domain, and propose two variants, namely the Hinted Embedding and Hinted Residual networks, both applied to the PoseNet base model for regressing camera pose from an image. Our evaluations show practical improvements in localization accuracy for standard outdoor and indoor localization datasets, without using additional information. We further assess the range of accuracy gains within an aerialview localization setup, simulated across vast areas at different times of the year. 
HIRO  Hierarchical reinforcement learning (HRL) is a promising approach to extend traditional reinforcement learning (RL) methods to solve more complex tasks. Yet, the majority of current HRL methods require careful taskspecific design and onpolicy training, making them difficult to apply in realworld scenarios. In this paper, we study how we can develop HRL algorithms that are general, in that they do not make onerous additional assumptions beyond standard RL algorithms, and efficient, in the sense that they can be used with modest numbers of interaction samples, making them suitable for realworld problems such as robotic control. For generality, we develop a scheme where lowerlevel controllers are supervised with goals that are learned and proposed automatically by the higherlevel controllers. To address efficiency, we propose to use offpolicy experience for both higher and lowerlevel training. This poses a considerable challenge, since changes to the lowerlevel behaviors change the action space for the higherlevel policy, and we introduce an offpolicy correction to remedy this challenge. This allows us to take advantage of recent advances in offpolicy modelfree RL to learn both higher and lowerlevel policies using substantially fewer environment interactions than onpolicy algorithms. We term the resulting HRL agent HIRO and find that it is generally applicable and highly sampleefficient. Our experiments show that HIRO can be used to learn highly complex behaviors for simulated robots, such as pushing objects and utilizing them to reach target locations, learning from only a few million samples, equivalent to a few days of realtime interaction. In comparisons with a number of prior HRL methods, we find that our approach substantially outperforms previous stateoftheart techniques. 
Histogram of Magnitude Optical Flow (HMOF) 
Anomaly detection is a challenging problem in intelligent video surveillance. Most existing methods are computation consuming, which cannot satisfy the realtime requirement. In this paper, we propose a realtime anomaly detection framework with low computational complexity and high efficiency. A new feature, named Histogram of Magnitude Optical Flow (HMOF), is proposed to capture the motion of video patches. Compared with existing feature descriptors, HMOF is more sensitive to motion magnitude and more efficient to distinguish anomaly information. The HMOF features are computed for foreground patches, and are reconstructed by the autoencoder for better clustering. Then, we use Gaussian Mixture Model (GMM) Classifiers to distinguish anomalies from normal activities in videos. Experimental results show that our framework outperforms stateoftheart methods, and can reliably detect anomalies in realtime. 
Histogram of Oriented Gradients (HOG) 
Histogram of Oriented Gradients (HOG) are feature descriptors used in computer vision and image processing for the purpose of object detection. The technique counts occurrences of gradient orientation in localized portions of an image. This method is similar to that of edge orientation histograms, scaleinvariant feature transform descriptors, and shape contexts, but differs in that it is computed on a dense grid of uniformly spaced cells and uses overlapping local contrast normalization for improved accuracy. Navneet Dalal and Bill Triggs, researchers for the French National Institute for Research in Computer Science and Control (INRIA), first described Histogram of Oriented Gradient descriptors in their June 2005 CVPR paper. In this work they focused their algorithm on the problem of pedestrian detection in static images, although since then they expanded their tests to include human detection in film and video, as well as to a variety of common animals and vehicles in static imagery. 
History PCA  In this paper we propose a new algorithm for streaming principal component analysis. With limited memory, small devices cannot store all the samples in the highdimensional regime. Streaming principal component analysis aims to find the $k$dimensional subspace which can explain the most variation of the $d$dimensional data points that come into memory sequentially. In order to deal with large $d$ and large $N$ (number of samples), most streaming PCA algorithms update the current model using only the incoming sample and then dump the information right away to save memory. However the information contained in previously streamed data could be useful. Motivated by this idea, we develop a new streaming PCA algorithm called History PCA that achieves this goal. By using $O(Bd)$ memory with $B\approx 10$ being the block size, our algorithm converges much faster than existing streaming PCA algorithms. By changing the number of inner iterations, the memory usage can be further reduced to $O(d)$ while maintaining a comparable convergence speed. We provide theoretical guarantees for the convergence of our algorithm along with the rate of convergence. We also demonstrate on synthetic and real world data sets that our algorithm compares favorably with other stateoftheart streaming PCA methods in terms of the convergence speed and performance. 
HiTMVAE  This work focuses on combining nonparametric topic models with AutoEncoding Variational Bayes (AEVB). Specifically, we first propose iTMVAE, where the topics are treated as trainable parameters and the documentspecific topic proportions are obtained by a stickbreaking construction. The inference of iTMVAE is modeled by neural networks such that it can be computed in a simple feedforward manner. We also describe how to introduce a hyperprior into iTMVAE so as to model the uncertainty of the prior parameter. Actually, the hyperprior technique is quite general and we show that it can be applied to other AEVB based models to alleviate the {\it collapsetoprior} problem elegantly. Moreover, we also propose HiTMVAE, where the documentspecific topic distributions are generated in a hierarchical manner. HiTMVAE is even more flexible and can generate topic distributions with better variability. Experimental results on 20News and Reuters RCV1V2 datasets show that the proposed models outperform the stateoftheart baselines significantly. The advantages of the hyperprior technique and the hierarchical model construction are also confirmed by experiments. 
HitNet  Neural networks designed for the task of classification have become a commodity in recent years. Many works target the development of better networks, which results in a complexification of their architectures with more layers, multiple subnetworks, or even the combination of multiple classifiers. In this paper, we show how to redesign a simple network to reach excellent performances, which are better than the results reproduced with CapsNet on several datasets, by replacing a layer with a HitorMiss layer. This layer contains activated vectors, called capsules, that we train to hit or miss a central capsule by tailoring a specific centripetal loss function. We also show how our network, named HitNet, is capable of synthesizing a representative sample of the images of a given class by including a reconstruction network. This possibility allows to develop a data augmentation step combining information from the data space and the feature space, resulting in a hybrid data augmentation process. In addition, we introduce the possibility for HitNet, to adopt an alternative to the true target when needed by using the new concept of ghost capsules, which is used here to detect potentially mislabeled images in the training data. 
Hitting Time  In the study of stochastic processes in mathematics, a hitting time (or first hit time) is the first time at which a given process “hits” a given subset of the state space. Exit times and return times are also examples of hitting times. 
HIVAE  Variational autoencoders (VAEs), as well as other generative models, have been shown to be efficient and accurate to capture the latent structure of vast amounts of complex highdimensional data. However, existing VAEs can still not directly handle data that are heterogenous (mixed continuous and discrete) or incomplete (with missing data at random), which is indeed common in realworld applications. In this paper, we propose a general framework to design VAEs, suitable for fitting incomplete heterogenous data. The proposed HIVAE includes likelihood models for realvalued, positive real valued, interval, categorical, ordinal and count data, and allows to estimate (and potentially impute) missing data accurately. Furthermore, HIVAE presents competitive predictive performance in supervised tasks, outperforming supervised models when trained on incomplete data. 
Hive Plot  The hive plot is a rational visualization method for drawing networks. Nodes are mapped to and positioned on radially distributed linear axes – this mapping is based on network structural properties. Edges are drawn as curved links. Simple and interpretable. The purpose of the hive plot is to establish a new baseline for visualization of large networks – a method that is both general and tunable and useful as a starting point in visually exploring network structure. 
HMPFSOM  kAnonymity by microaggregation is one of the most commonly used anonymization techniques. This success is owe to the achievement of a worth of interest tradeoff between information loss and identity disclosure risk. However, this method may have some drawbacks. On the disclosure limitation side, there is a lack of protection against attribute disclosure. On the data utility side, dealing with a real datasets is a challenging task to achieve. Indeed, the latter are characterized by their large number of attributes and the presence of noisy data, such that outliers or, even, data with missing values. Generating an anonymous individual data useful for data mining tasks, while decreasing the influence of noisy data is a compelling task to achieve. In this paper, we introduce a new microaggregation method, called HMPFSOM, based on fuzzy possibilistic clustering. Our proposed method operates through an hybrid manner. This means that the anonymization process is applied per block of similar data. Thus, we can help to decrease the information loss during the anonymization process. The HMPFSOM approach proposes to study the distribution of confidential attributes within each subdataset. Then, according to the latter distribution, the privacy parameter k is determined, in such a way to preserve the diversity of confidential attributes within the anonymized microdata. This allows to decrease the disclosure risk of confidential information. 
HodrickPrescott Filter (HP Filter) 
The HodrickPrescott filter (also known as HodrickPrescott decomposition) is a mathematical tool used in macroeconomics, especially in real business cycle theory, to remove the cyclical component of a time series from raw data. It is used to obtain a smoothedcurve representation of a time series, one that is more sensitive to longterm than to shortterm fluctuations. The adjustment of the sensitivity of the trend to shortterm fluctuations is achieved by modifying a multiplier \lambda. The filter was popularized in the field of economics in the 1990s by economists Robert J. Hodrick and Nobel Memorial Prize winner Edward C. Prescott. However, it was first proposed much earlier by E. T. Whittaker in 1923. The HP Filter and Unit Roots 
Hoeffding Anytime Tree  We introduce a novel incremental decision tree learning algorithm, Hoeffding Anytime Tree, that is statistically more efficient than the current stateoftheart, Hoeffding Tree. We demonstrate that an implementation of Hoeffding Anytime Tree—‘Extremely Fast Decision Tree’, a minor modification to the MOA implementation of Hoeffding Tree—obtains significantly superior prequential accuracy on most of the largest classification datasets from the UCI repository. Hoeffding Anytime Tree produces the asymptotic batch tree in the limit, is naturally resilient to concept drift, and can be used as a higher accuracy replacement for Hoeffding Tree in most scenarios, at a small additional computational cost. 
Hoeffding Tree (VFDT) 
A Hoeffding tree (VFDT) is an incremental, anytime decision tree induction algorithm that is capable of learning from massive data streams, assuming that the distribution generating examples does not change over time. Hoeffding trees exploit the fact that a small sample can often be enough to choose an optimal splitting attribute. This idea is supported mathematically by the Hoeffding bound, which quantifies the number of observations (in our case, examples) needed to estimate some statistics within a prescribed precision (in our case, the goodness of an attribute). A theoretically appealing feature of Hoeffding Trees not shared by otherincremental decision tree learners is that it has sound guarantees of performance. Using the Hoeffding bound one can show that its output is asymptotically nearly identical to that of a nonincremental learner using infinitely many examples. For more information see: Geoff Hulten, Laurie Spencer, Pedro Domingos: Mining timechanging data streams. In: ACM SIGKDD Intl. Conf. on Knowledge Discovery and Data Mining, 97106, 2001. 
Hogwild!  Stochastic Gradient Descent (SGD) is a popular algorithm that can achieve stateoftheart performance on a variety of machine learning tasks. Several researchers have recently pro posed schemes to parallelize SGD, but all require performancedestroying memory locking and synchronization. This work aims to show using novel theoretical analysis, algorithms, and im plementation that SGD can be implemented without any locking. We present an update scheme called Hogwild! which allows processors access to shared memory with the possibility of over writing each other’s work. 
Holdout Randomization Test (HRT) 
We consider the problem of feature selection using black box predictive models. For example, highthroughput devices in science are routinely used to gather thousands of features for each sample in an experiment. The scientist must then sift through the many candidate features to find explanatory signals in the data, such as which genes are associated with sensitivity to a prospective therapy. Often, predictive models are used for this task: the model is fit, error on held out data is measured, and strong performing models are assumed to have discovered some fundamental properties of the system. A modelspecific heuristic is then used to inspect the model parameters and rank important features, with top features reported as ‘discoveries.’ However, such heuristics provide no statistical guarantees and can produce unreliable results. We propose the holdout randomization test (HRT) as a principled approach to feature selection using black box predictive models. The HRT is similar to a permutation test, where each random reshuffling is a draw from the complete conditional distribution of the feature being tested. The HRT is model agnostic and produces a valid $p$value for each feature, enabling control over the false discovery rate (or Type I error) for any predictive model. Further, the HRT is computationally efficient and, in simulations, has greater power than a competing knockoffsbased approach. Code is available at https://…/hrt. 
HOList  We present an environment, benchmark, and deep learning driven automated theorem prover for higherorder logic. Higherorder interactive theorem provers enable the formalization of arbitrary mathematical theories and thereby present an interesting, openended challenge for deep learning. We provide an opensource framework based on the HOL Light theorem prover that can be used as a reinforcement learning environment. HOL Light comes with a broad coverage of basic mathematical theorems on calculus and the formal proof of the Kepler conjecture, from which we derive a challenging benchmark for automated reasoning. We also present a deep reinforcement learning driven automated theorem prover, DeepHOL, with strong initial results on this benchmark. 
Hollow Heap  We introduce the hollow heap, a very simple data structure with the same amortized efficiency as the classical Fibonacci heap. All heap operations except delete and deletemin take $O(1)$ time, worst case as well as amortized; delete and deletemin take $O(\log n)$ amortized time on a heap of $n$ items. Hollow heaps are by far the simplest structure to achieve this. Hollow heaps combine two novel ideas: the use of lazy deletion and reinsertion to do decreasekey operations, and the use of a dag (directed acyclic graph) instead of a tree or set of trees to represent a heap. Lazy deletion produces hollow nodes (nodes without items), giving the data structure its name. 
Holographic Neural Architecture (HNA) 
Representation learning is at the heart of what makes deep learning effective. In this work, we introduce a new framework for representation learning that we call ‘Holographic Neural Architectures’ (HNAs). In the same way that an observer can experience the 3D structure of a holographed object by looking at its hologram from several angles, HNAs derive Holographic Representations from the training set. These representations can then be explored by moving along a continuous bounded single dimension. We show that HNAs can be used to make generative networks, stateoftheart regression models and that they are inherently highly resistant to noise. Finally, we argue that because of their denoising abilities and their capacity to generalize well from very few examples, models based upon HNAs are particularly well suited for biological applications where training examples are rare or noisy. 
Holonomic Gradient Method (HGM) 
The holonomic gradient method introduced by Nakayama et al. (2011) presents a new methodology for evaluating normalizing constants of probability distributions and for obtaining the maximum likelihood estimate of a statistical model. The method utilizes partial differential equations satisfied by the normalizing constant and is based on the Grobner basis theory for the ring of differential operators. In this talk we give an introduction to this new methodology. The method has already proved to be useful for problems in directional statistics and in classical multivariate distribution theory involving hypergeometric functions of matrix arguments. hgm 
HoltWinters double exponential smoothing  This method is used when the data shows a trend. Exponential smoothing with a trend works much like simple smoothing except that two components must be updated each period – level and trend. The level is a smoothed estimate of the value of the data at the end of each period. The trend is a smoothed estimate of average growth at the end of each period. http://…–theholtwintersforecastingmethod.pdf 
HoltWinters Method (HW) 
Holt (1957) and Winters (1960) extended Holt’s method to capture seasonality. The HoltWinters seasonal method comprises the forecast equation and three smoothing equations – one for the level ℓ t , one for trend b t , and one for the seasonal component denoted by s t, with smoothing parameters α , β ∗ and γ. We use m to denote the period of the seasonality, i.e., the number of seasons in a year. For example, for quarterly data m=4 , and for monthly data m=12. There are two variations to this method that differ in the nature of the seasonal component. The additive method is preferred when the seasonal variations are roughly constant through the series, while the multiplicative method is preferred when the seasonal variations are changing proportional to the level of the series. With the additive method, the seasonal component is expressed in absolute terms in the scale of the observed series, and in the level equation the series is seasonally adjusted by subtracting the seasonal component. Within each year the seasonal component will add up to approximately zero. With the multiplicative method, the seasonal component is expressed in relative terms (percentages) and the series is seasonally adjusted by dividing through by the seasonal component. Within each year, the seasonal component will sum up to approximately m. 
Homebrew  Homebrew has made extensive use of GitHub to expand the support of several packages through user contributions. In 2010, Homebrew was the thirdmostforked repository on GitHub. In 2012, Homebrew had the largest number of new contributors on GitHub. In 2013, Homebrew had both the largest number of contributors and issues closed of any project on GitHub. Homebrew has spawned several subprojects such as Linuxbrew, which is a Linux port, Homebrew Cask, which builds upon Homebrew and focuses on the installation of GUI applications, and ‘taps’ dedicated to specific areas or programming languages like PHP. How to Install and Use Homebrew 
Homographic Adaptation  This paper presents a selfsupervised framework for training interest point detectors and descriptors suitable for a large number of multipleview geometry problems in computer vision. As opposed to patchbased neural networks, our fullyconvolutional model operates on fullsized images and jointly computes pixellevel interest point locations and associated descriptors in one forward pass. We introduce Homographic Adaptation, a multiscale, multihomography approach for boosting interest point detection accuracy and performing crossdomain adaptation (e.g., synthetictoreal). Our model, when trained on the MSCOCO generic image dataset using Homographic Adaptation, is able to repeatedly detect a much richer set of interest points than the initial preadapted deep model and any other traditional corner detector. The final system gives rise to strong interest point repeatability on the HPatches dataset and outperforms traditional descriptors such as ORB and SIFT on point matching accuracy and on the task of homography estimation. 
Homomorphic Instruction Set Architecture (HISA) 
Fully Homomorphic Encryption (FHE) refers to a set of encryption schemes that allow computations to be applied directly on encrypted data without requiring a secret key. This enables novel application scenarios where a client can safely offload storage and computation to a thirdparty cloud provider without having to trust the software and the hardware vendors with the decryption keys. Recent advances in both FHE schemes and implementations have moved such applications from theoretical possibilities into the realm of practicalities. This paper proposes a compact and wellreasoned interface called the Homomorphic Instruction Set Architecture (HISA) for developing FHE applications. Just as the hardware ISA interface enabled hardware advances to proceed independent of software advances in the compiler and language runtimes, HISA decouples compiler optimizations and runtimes for supporting FHE applications from advancements in the underlying FHE schemes. This paper demonstrates the capabilities of HISA by building an endtoend software stack for evaluating neural network models on encrypted data. Our stack includes an endtoend compiler, runtime, and a set of optimizations. Our approach shows generated code, on a set of popular neural network architectures, is faster than handoptimized implementations. 
Homomorphic Sensing  A recent line of research termed unlabeled sensing and shuffled linear regression has been exploring under great generality the recovery of signals from subsampled and permuted measurements; a challenging problem in diverse fields of data science and machine learning. In this paper we introduce an abstraction of this problem which we call homomorphic sensing. Given a linear subspace and a finite set of linear transformations we develop an algebraic theory which establishes conditions guaranteeing that points in the subspace are uniquely determined from their homomorphic image under some transformation in the set. As a special case, we recover known conditions for unlabeled sensing, as well as new results and extensions. On the algorithmic level we exhibit two dynamic programming based algorithms, which to the best of our knowledge are the first working solutions for the unlabeled sensing problem for small dimensions. One of them, additionally based on branchandbound, when applied to image registration under affine transformations, performs on par with or outperforms stateoftheart methods on benchmark datasets. 
Homoscedasticity  In statistics, a sequence or a vector of random variables is homoscedastic if all random variables in the sequence or vector have the same finite variance. This is also known as homogeneity of variance. The complementary notion is called heteroscedasticity. The spellings homoskedasticity and heteroskedasticity are also frequently used. 
Honda Research Institute Driving Dataset (HDD) 
Driving Scene understanding is a key ingredient for intelligent transportation systems. To achieve systems that can operate in a complex physical and social environment, they need to understand and learn how humans drive and interact with traffic scenes. We present the Honda Research Institute Driving Dataset (HDD), a challenging dataset to enable research on learning driver behavior in reallife environments. The dataset includes 104 hours of real human driving in the San Francisco Bay Area collected using an instrumented vehicle equipped with different sensors. We provide a detailed analysis of HDD with a comparison to other driving datasets. A novel annotation methodology is introduced to enable research on driver behavior understanding from untrimmed data sequences. As the first step, baseline algorithms for driver behavior detection are trained and tested to demonstrate the feasibility of the proposed task. 
Hopfield Network  A Hopfield network is a form of recurrent artificial neural network invented by John Hopfield in 1982. Hopfield nets serve as contentaddressable memory systems with binary threshold nodes. They are guaranteed to converge to a local minimum, but convergence to a false pattern (wrong local minimum) rather than the stored pattern (expected local minimum) can occur. Hopfield networks also provide a model for understanding human memory. 
HopRank  This paper introduces HopRank, an algorithm for modeling human navigation on semantic networks. HopRank leverages the assumption that users know or can see the whole structure of the network. Therefore, besides following links, they also follow nodes at certain distances (i.e., khop neighborhoods), and not at random as suggested by PageRank, which assumes only links are known or visible. We observe such preference towards khop neighborhoods on BioPortal, one of the leading repositories of biomedical ontologies on the Web. In general, users navigate within the vicinity of a concept. But they also ‘jump’ to distant concepts less frequently. We fit our model on 11 ontologies using the transition matrix of clickstreams, and show that semantic structure can influence teleportation in PageRank. This suggests that users–to some extent–utilize knowledge about the underlying structure of ontologies, and leverage it to reach certain pieces of information. Our results help the development and improvement of user interfaces for ontology exploration. 
HopsFS  Recent improvements in both the performance and scalability of sharednothing, transactional, inmemory NewSQL databases have reopened the research question of whether distributed metadata for hierarchical file systems can be managed using commodity databases. In this paper, we introduce HopsFS, a next generation distribution of the Hadoop Distributed File System (HDFS) that replaces HDFS single node inmemory metadata service, with a distributed metadata service built on a NewSQL database. By removing the metadata bottleneck, HopsFS improves capacity and throughput compared to HDFS. HopsFS can store 24 times more metadata than HDFS. We also provide public, fully reproducible experiments based on a workload trace from Spotify that show HopsFS has 2.6 times the throughput of Apache HDFS, lower latency for greater than 400 concurrent clients, and no downtime during failover. Finally, and most significantly, HopsFS allows metadata to be exported to external systems, analyzed or searched online, and easily extended. 
Horn  I introduce a new distributed system for effective training and regularizing of LargeScale Neural Networks on distributed computing architectures. The experiments demonstrate the effectiveness of flexible model partitioning and parallelization strategies based on neuroncentric computation model, with an implementation of the collective and parallel dropout neural networks training. Experiments are performed on MNIST handwritten digits classification including results. 
Horn Implication Counterexamples (HornICE) 
HornICE Learning for Synthesizing Invariants and Contracts 
HornConcerto  Graph representations of large knowledge bases may comprise billions of edges. Usually built upon humangenerated 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 inmemory, 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 factbased 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 highquality rules for the link prediction task, achieving stateoftheart results on a widelyused benchmark. Moreover, we find that rules alone can perform inference significantly faster than embeddingbased methods and achieve accuracies on link prediction comparable to resourcedemanding approaches such as Markov Logic Networks. 
Horovod  Training modern deep learning models requires large amounts of computation, often provided by GPUs. Scaling computation from one GPU to many can enable much faster training and research progress but entails two complications. First, the training library must support interGPU communication. Depending on the particular methods employed, this communication may entail anywhere from negligible to significant overhead. Second, the user must modify his or her training code to take advantage of interGPU communication. Depending on the training library’s API, the modification required may be either significant or minimal. Existing methods for enabling multiGPU training under the TensorFlow library entail nonnegligible communication overhead and require users to heavily modify their modelbuilding code, leading many researchers to avoid the whole mess and stick with slower singleGPU training. In this paper we introduce Horovod, an open source library that improves on both obstructions to scaling: it employs efficient interGPU communication via ring reduction and requires only a few lines of modification to user code, enabling faster, easier distributed training in TensorFlow. Horovod is available under the Apache 2.0 license at https://…/horovod. 
HorseRule  The HorseRule model is a flexible tree based Bayesian regression method for linear and nonlinear regression and classification described in Nalenz & Villani (2017) <arXiv:1702.05008>. horserule 
Horseshoe Estimator  This paper proposes a new approach to sparsesignal detection called the horseshoe estimator. We show that the horseshoe is a close cousin of the lasso in that it arises from the same class of multivariate scale mixtures of normals, but that it is almost universally superior to the doubleexponential prior at handling sparsity. A theoretical framework is proposed for understanding why the horseshoe is a better default ‘sparsity’ estimator than those that arise from poweredexponential priors. Comprehensive numerical evidence is presented to show that the difference in performance can often be large. Most importantly, we show that the horseshoe estimator corresponds quite closely to the answers one would get if one pursued a full Bayesian modelaveraging approach using a ‘twogroups’ model: a point mass at zero for noise, and a continuous density for signals. Surprisingly, this correspondence holds both for the estimator itself and for the classification rule induced by a simple threshold applied to the estimator. We show how the resulting thresholded horseshoe can also be viewed as a novel Bayes multipletesting procedure. horseshoe 
Horseshoe Regularization  Feature subset selection arises in many highdimensional applications in machine learning and statistics, such as compressed sensing and genomics. The $\ell_0$ penalty is ideal for this task, the caveat being it requires the NPhard combinatorial evaluation of all models. A recent area of considerable interest is to develop efficient algorithms to fit models with a nonconvex $\ell_\gamma$ penalty for $\gamma\in (0,1)$, which results in sparser models than the convex $\ell_1$ or lasso penalty, but is harder to fit. We propose an alternative, termed the horseshoe regularization penalty for feature subset selection, and demonstrate its theoretical and computational advantages. The distinguishing feature from existing nonconvex optimization approaches is a full probabilistic representation of the penalty as the negative of the logarithm of a suitable prior, which in turn enables an efficient expectationmaximization algorithm for optimization and MCMC for uncertainty quantification. In synthetic and real data, the resulting algorithm provides better statistical performance, and the computation requires a fraction of time of state of the art nonconvex solvers. 
Hospital Residents Problem  ➘ “Stable Marriage Problem” 
Hot Deck Imputation  This method sorts respondents and nonrespondents into a number of imputation subsets according to a userspecified set of covariates. An imputation subset comprises cases with the same values as those of the userspecified covariates. Missing values are then replaced with values taken from matching respondents (i.e. respondents that are similar with respect to the covariates). If there is more than one matching respondent for any particular nonrespondent, the user has two choices: 1. The first respondent’s value as counted from the missing entry downwards within the imputation subset is used to impute. The reason for this is that the first respondent’s value may be closer in time to the case that has the missing value. For example, if cases are entered according to the order in which they occur, there may possibly be some type of time effect in some studies. 2. A respondent’s value is randomly selected from within the imputation subset. If a matching respondent does not exist in the initial imputation class, the subset will be collapsed by one level starting with the last variable that was selected as a sort variable, or until a match can be found. Note that if no matching respondent is found, even after all of the sort variables have been collapsed, three options are available: 1. Respecify new sort variables: The user can specify up to five sort variables. 2. Perform random overall imputation: Where the missing value will be replaced with a value randomly selected from the observed values in that variable. 3. Do not impute the missing value: SOLAS will not impute any missing values for which no matching respondent is found. HotDeckImputation,hot.deck 
Hot Spot Analysis  Also known as GetisOrd Gi* – The resultant zscores and pvalues tell you where features with either high or low values cluster spatially. This tool works by looking at each feature within the context of neighboring features. A feature with a high value is interesting by may not be a statistically significant hot spot. To be a statistically significant hotspot, a feature will have a high value and be surrounded by other features with high values as well. The local sum for a feature and its neighbors is compared proportionally to the sum of all features; when the local sum is very different from the expected local sum, and that difference is too large to be the result of random choice, a statistically significant zscore results. The Gi* statistic returned for each feature in the dataset is a zscore. For statistically significant positive zscores, the larger the zscore is, the more intense clustering of high values (hot spot). For statistically significant negative zscores, the smaller the zscore is, the more intense the clustering of low values (cold spot). When to use: Results aren’t reliable with less than 30 features. Applications can be found in crime analysis, epidemiology, voting pattern analysis, economic geography, retail analysis, traffic incident analysis, and demographics. Examples: Where is the disease outbreak concentrated? – Where are kitchen fires a larger than expected proportion of all residential fires? – Where should the evacuation sites be located? – Where/When do peak intensities occur? How Hot Spot Analysis works 
Hotelling’s Law  ➘ “Principle of Minimum Differentiation” 
Houdini  Generating adversarial examples is a critical step for evaluating and improving the robustness of learning machines. So far, most existing methods only work for classification and are not designed to alter the true performance measure of the problem at hand. We introduce a novel flexible approach named Houdini for generating adversarial examples specifically tailored for the final performance measure of the task considered, be it combinatorial and nondecomposable. We successfully apply Houdini to a range of applications such as speech recognition, pose estimation and semantic segmentation. In all cases, the attacks based on Houdini achieve higher success rate than those based on the traditional surrogates used to train the models while using a less perceptible adversarial perturbation. 
Housing Allocation Problem  There is a set of agents and a set of houses. Each agent has a strict preference list for a subset of houses. We need to find a matching such that some criterion is optimized. One such criterion is Pareto Optimality. A matching is Pareto optimal if no coalition of agents can be strictly better off by exchanging houses among themselves. 
HPCBBO  Robot design is often a slow and difficult process requiring the iterative construction and testing of prototypes, with the goal of sequentially optimizing the design. For most robots, this process is further complicated by the need, when validating the capabilities of the hardware to solve the desired task, to already have an appropriate controller, which is in turn designed and tuned for the specific hardware. In this paper, we propose a novel approach, HPCBBO, to efficiently and automatically design hardware configurations, and evaluate them by also automatically tuning the corresponding controller. HPCBBO is based on a hierarchical Bayesian optimization process which iteratively optimizes morphology configurations (based on the performance of the previous designs during the controller learning process) and subsequently learns the corresponding controllers (exploiting the knowledge collected from optimizing for previous morphologies). Moreover, HPCBBO can select a ‘batch’ of multiple morphology designs at once, thus parallelizing hardware validation and reducing the number of timeconsuming production cycles. We validate HPCBBO on the design of the morphology and controller for a simulated 6legged microrobot. Experimental results show that HPCBBO outperforms multiple competitive baselines, and yields a $360\%$ reduction in production cycles over standard Bayesian optimization, thus reducing the hypothetical manufacturing time of our microrobot from 21 to 4 months. 
HSRL  The topological information is essential for studying the relationship between nodes in a network. Recently, Network Representation Learning (NRL), which projects a network into a lowdimensional vector space, has been shown their advantages in analyzing largescale networks. However, most existing NRL methods are designed to preserve the local topology of a network, they fail to capture the global topology. To tackle this issue, we propose a new NRL framework, named HSRL, to help existing NRL methods capture both the local and global topological information of a network. Specifically, HSRL recursively compresses an input network into a series of smaller networks using a communityawareness compressing strategy. Then, an existing NRL method is used to learn node embeddings for each compressed network. Finally, the node embeddings of the input network are obtained by concatenating the node embeddings from all compressed networks. Empirical studies for link prediction on five realworld datasets demonstrate the advantages of HSRL over stateoftheart methods. 
Huber Loss  In statistics, the Huber loss is a loss function used in robust regression, that is less sensitive to outliers in data than the squared error loss. A variant for classification is also sometimes used. hqreg 
Hubness  The tendency of highdimensional data to contain points (hubs) that frequently occur in knearestneighbor lists of other points. Semantically Aligned Bias Reducing Zero Shot Learning Semantically Aligned Bias Reducing Zero Shot Learning 
Hubs and Authorities  ➘ “HyperlinkInduced Topic Search” 
HuFu  Recently, Deep Learning (DL), especially Convolutional Neural Network (CNN), develops rapidly and is applied to many tasks, such as image classification, face recognition, image segmentation, and human detection. Due to its superior performance, DLbased models have a wide range of application in many areas, some of which are extremely safetycritical, e.g. intelligent surveillance and autonomous driving. Due to the latency and privacy problem of cloud computing, embedded accelerators are popular in these safetycritical areas. However, the robustness of the embedded DL system might be harmed by inserting hardware/software Trojans into the accelerator and the neural network model, since the accelerator and deploy tool (or neural network model) are usually provided by thirdparty companies. Fortunately, inserting hardware Trojans can only achieve inflexible attack, which means that hardware Trojans can easily break down the whole system or exchange two outputs, but can’t make CNN recognize unknown pictures as targets. Though inserting software Trojans has more freedom of attack, it often requires tampering input images, which is not easy for attackers. So, in this paper, we propose a hardwaresoftware collaborative attack framework to inject hidden neural network Trojans, which works as a backdoor without requiring manipulating input images and is flexible for different scenarios. We test our attack framework for image classification and face recognition tasks, and get attack success rate of 92.6% and 100% on CIFAR10 and YouTube Faces, respectively, while keeping almost the same accuracy as the unattacked model in the normal mode. In addition, we show a specific attack scenario in which a face recognition system is attacked and gives a specific wrong answer. 
Human Activity Knowledge Engine (HAKE) 
Human activity understanding is crucial for building automatic intelligent system. With the help of deep learning, activity understanding has made huge progress recently. But some challenges such as imbalanced data distribution, action ambiguity, complex visual patterns still remain. To address these and promote the activity understanding, we build a largescale Human Activity Knowledge Engine (HAKE) based on the human body part states. Upon existing activity datasets, we annotate the part states of all the active persons in all images, thus establish the relationship between instance activity and body part states. Furthermore, we propose a HAKE based part state recognition model with a knowledge extractor named Activity2Vec and a corresponding part state based reasoning network. With HAKE, our method can alleviate the learning difficulty brought by the longtail data distribution, and bring in interpretability. Now our HAKE has more than 7 M+ part state annotations and is still under construction. We first validate our approach on a part of HAKE in this preliminary paper, where we show 7.2 mAP performance improvement on HumanObject Interaction recognition, and 12.38 mAP improvement on the oneshot subsets. 
Human And Machine coLEarning Technique (HAMLET) 
Efficient label acquisition processes are key to obtaining robust classifiers. However, data labeling is often challenging and subject to high levels of label noise. This can arise even when classification targets are well defined, if instances to be labeled are more difficult than the prototypes used to define the class, leading to disagreements among the expert community. Here, we enable efficient training of deep neural networks. From lowconfidence labels, we iteratively improve their quality by simultaneous learning of machines and experts. We call it Human And Machine coLEarning Technique (HAMLET). Throughout the process, experts become more consistent, while the algorithm provides them with explainable feedback for confirmation. HAMLET uses a neural embedding function and a memory module filled with diverse reference embeddings from different classes. Its output includes classification labels and highly relevant reference embeddings as explanation. We took the study of brain monitoring at intensive care unit (ICU) as an application of HAMLET on continuous electroencephalography (cEEG) data. Although cEEG monitoring yields large volumes of data, labeling costs and difficulty make it hard to build a classifier. Additionally, while experts agree on the labels of clearcut examples of cEEG patterns, labeling many realworld cEEG data can be extremely challenging. Thus, a large minority of sequences might be mislabeled. HAMLET has shown significant performance gain against deep learning and other baselines, increasing accuracy from 7.03% to 68.75% on challenging inputs. Besides improved performance, clinical experts confirmed the interpretability of those reference embeddings in helping explaining the classification results by HAMLET. 
Human Group Optimizer (HGO) 
A large number of optimization algorithms have been developed by researchers to solve a variety of complex problems in operations management area. We present a novel optimization algorithm belonging to the class of swarm intelligence optimization methods. The algorithm mimics the decision making process of human groups and exploits the dynamics of this process as an optimization tool for combinatorial problems. In order to achieve this aim, a continuoustime Markov process is proposed to describe the behavior of a population of socially interacting agents, modelling how humans in a group modify their opinions driven by selfinterest and consensus seeking. As in the case of a collection of spins, the dynamics of such a system is characterized by a phase transition from low to high values of the overall consenus (magnetization). We recognize this phase transition as being associated with the emergence of a collective superior intelligence of the population. While this state being active, a cooling schedule is applied to make agents closer and closer to the optimal solution, while performing their random walk on the fitness landscape. A comparison with simulated annealing as well as with a multiagent version of the simulated annealing is presented in terms of efficacy in finding good solution on a NK – Kauffman landscape. In all cases our method outperforms the others, particularly in presence of limited knowledge of the agent. 
HumanCentered Artificial Intelligence  Humans are increasingly coming into contact with artificial intelligence and machine learning systems. Humancentered artificial intelligence is a perspective on AI and ML that algorithms must be designed with awareness that they are part of a larger system consisting of humans. We lay forth an argument that humancentered artificial intelligence can be broken down into two aspects: (1) AI systems that understand humans from a sociocultural perspective, and (2) AI systems that help humans understand them. We further argue that issues of social responsibility such as fairness, accountability, interpretability, and transparency. 
HumaneYe Perceptual Evaluation (HYPE) 
Generative models often use human evaluations to determine and justify progress. Unfortunately, existing human evaluation methods are adhoc: 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 highquality measurement without intractable cost. In response, we construct HumaneYe 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, HYPETime, 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, HYPEInfinity, 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 stateoftheart generative adversarial networks (GANs) on unconditional image generation using two datasets, the popular CelebA and the newer higherresolution 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% HYPEInfinity 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. 
HumanintheLoop (HITL) 

Humanintheloop Artificial Intelligence (HITAI) 
Little by little, newspapers are revealing the bright future that Artificial Intelligence (AI) is building. Intelligent machines will help everywhere. However, this bright future has a dark side: a dramatic job market contraction before its unpredictable transformation. Hence, in a near future, large numbers of job seekers will need financial support while catching up with these novel unpredictable jobs. This possible job market crisis has an antidote inside. In fact, the rise of AI is sustained by the biggest knowledge theft of the recent years. Learning AI machines are extracting knowledge from unaware skilled or unskilled workers by analyzing their interactions. By passionately doing their jobs, these workers are digging their own graves. In this paper, we propose Humanintheloop Artificial Intelligence (HITAI) as a fairer paradigm for Artificial Intelligence systems. HITAI will reward aware and unaware knowledge producers with a different scheme: decisions of AI systems generating revenues will repay the legitimate owners of the knowledge used for taking those decisions. As modern Robin Hoods, HITAI researchers should fight for a fairer Artificial Intelligence that gives back what it steals. 
HumanMachine Inference Network (HuMaIN) 
The emerging paradigm of HumanMachine Inference Networks (HuMaINs) combines complementary cognitive strengths of humans and machines in an intelligent manner to tackle various inference tasks and achieves higher performance than either humans or machines by themselves. While inference performance optimization techniques for humanonly or sensoronly networks are quite mature, HuMaINs require novel signal processing and machine learning solutions. In this paper, we present an overview of the HuMaINs architecture with a focus on three main issues that include architecture design, inference algorithms including security/privacy challenges, and application areas/use cases. 
HUPNU  Modern Internet of Things (IoT) applications generate massive amounts of data, much of it in the form of objects/items of readings, events, and log entries. Specifically, most of the objects in these IoT data contain rich embedded information (e.g., frequency and uncertainty) and different level of importance (e.g., unit utility of items, interestingness, cost, risk, or weight). Many existing approaches in data mining and analytics have limitations such as only the binary attribute is considered within a transaction, as well as all the objects/items having equal weights or importance. To solve these drawbacks, a novel utilitydriven data analytics algorithm named HUPNU is presented, to extract HighUtility patterns by considering both Positive and Negative unit utilities from Uncertain data. The qualified highutility patterns can be effectively discovered for risk prediction, manufacturing management, decisionmaking, among others. By using the developed vertical ProbabilityUtility list with the PositiveandNegative utilities structure, as well as several effective pruning strategies. Experiments showed that the developed HUPNU approach performed great in mining the qualified patterns efficiently and effectively. 
Hurst Coefficient  ➘ “Hurst Exponent” 
Hurst Exponent  The Hurst exponent is used as a measure of longterm memory of time series. It relates to the autocorrelations of the time series, and the rate at which these decrease as the lag between pairs of values increases. Studies involving the Hurst exponent were originally developed in hydrology for the practical matter of determining optimum dam sizing for the Nile river’s volatile rain and drought conditions that had been observed over a long period of time. The name ‘Hurst exponent’, or ‘Hurst coefficient’, derives from Harold Edwin Hurst (18801978), who was the lead researcher in these studies; the use of the standard notation H for the coefficient relates to his name also. In fractal geometry, the generalized Hurst exponent has been denoted by H or Hq in honor of both Harold Edwin Hurst and Ludwig Otto Hölder (18591937) by Benoît Mandelbrot (19242010). H is directly related to fractal dimension, D, and is a measure of a data series’ ‘mild’ or ‘wild’ randomness. The Hurst exponent is referred to as the ‘index of dependence’ or ‘index of longrange dependence’. It quantifies the relative tendency of a time series either to regress strongly to the mean or to cluster in a direction. A value H in the range 0.51 indicates a time series with longterm positive autocorrelation, meaning both that a high value in the series will probably be followed by another high value and that the values a long time into the future will also tend to be high. A value in the range 0 – 0.5 indicates a time series with longterm switching between high and low values in adjacent pairs, meaning that a single high value will probably be followed by a low value and that the value after that will tend to be high, with this tendency to switch between high and low values lasting a long time into the future. A value of H=0.5 can indicate a completely uncorrelated series, but in fact it is the value applicable to series for which the autocorrelations at small time lags can be positive or negative but where the absolute values of the autocorrelations decay exponentially quickly to zero. This in contrast to the typically power law decay for the 0.5 < H < 1 and 0 < H < 0.5 cases. 
HUSPULL  Highutility sequential pattern mining is an emerging topic in the field of Knowledge Discovery in Databases. It consists of discovering subsequences having a high utility (importance) in sequences, referred to as highutility sequential patterns (HUSPs). HUSPs can be applied to many reallife applications, such as market basket analysis, Ecommerce recommendation, clickstream analysis and scenic route planning. For example, in economics and targeted marketing, understanding economic behavior of consumers is quite challenging, such as finding credible and reliable information on product profitability. Several algorithms have been proposed to address this problem by efficiently mining utilitybased useful sequential patterns. Nevertheless, the performance of these algorithms can be unsatisfying in terms of runtime and memory usage due to the combinatorial explosion of the search space for low utility threshold and large databases. Hence, this paper proposes a more efficient algorithm for the task of highutility sequential pattern mining, called HUSPULL. It utilizes a lexicographic sequence (LS)tree and a utilitylinked (UL)list structure to fast discover HUSPs. Furthermore, two pruning strategies are introduced in HUSPULL to obtain tight upperbounds on the utility of candidate sequences, and reduce the search space by pruning unpromising candidates early. Substantial experiments both on reallife and synthetic datasets show that the proposed algorithm can effectively and efficiently discover the complete set of HUSPs and outperforms the stateoftheart algorithms. 
HVARX  The Vector AutoRegressive (VAR) model is fundamental to the study of multivariate time series. Although VAR models are intensively investigated by many researchers, practitioners often show more interest in analyzing VARX models that incorporate the impact of unmodeled exogenous variables (X) into the VAR. However, since the parameter space grows quadratically with the number of time series, estimation quickly becomes challenging. While several proposals have been made to sparsely estimate large VAR models, the estimation of large VARX models is underexplored. Moreover, typically these sparse proposals involve a lassotype penalty and do not incorporate lag selection into the estimation procedure. As a consequence, the resulting models may be difficult to interpret. In this paper, we propose a lagbased hierarchically sparse estimator, called ‘HVARX’, for large VARX models. We illustrate the usefulness of HVARX on a crosscategory management marketing application. Our results show how it provides a highly interpretable model, and improves outofsample forecast accuracy compared to a lassotype approach. 
Hy  Hy is a Lisp dialect that converts its structure into Python’s abstract syntax tree. It is to Python what LFE is to Erlang.This provides developers from many backgrounds with the following: · A lisp that feels very Pythonic · A great way to use Lisp’s crazy powers but in the wide world of Python’s libraries · A great way to start exploring Lisp, from the comfort of python · A pleasant language that has a lot of neat ideas 🙂 
Hybrid  We study the problem of personalized, interactive tag recommendation for Flickr: While a user enters/selects new tags for a particular picture, the system suggests related tags to her, based on the tags that she or other people have used in the past along with (some of) the tags already entered. The suggested tags are dynamically updated with every additional tag entered/selected. We describe a new algorithm, called Hybrid, which can be applied to this problem, and show that it outperforms previous algorithms. It has only a single tunable parameter, which we found to be very robust. 
Hybrid Ant Colony Optimization Algorithm (HACO) 
In this paper, we propose a Hybrid Ant Colony Optimization algorithm (HACO) for Next Release Problem (NRP). NRP, a NPhard problem in requirement engineering, is to balance customer requests, resource constraints, and requirement dependencies by requirement selection. Inspired by the successes of Ant Colony Optimization algorithms (ACO) for solving NPhard problems, we design our HACO to approximately solve NRP. Similar to traditional ACO algorithms, multiple artificial ants are employed to construct new solutions. During the solution construction phase, both pheromone trails and neighborhood information will be taken to determine the choices of every ant. In addition, a local search (first found hill climbing) is incorporated into HACO to improve the solution quality. Extensively wide experiments on typical NRP test instances show that HACO outperforms the existing algorithms (GRASP and simulated annealing) in terms of both solution uality and running time. 
Hybrid Artificial Intelligence  ➘ “Hybrid Intelligent System” 
Hybrid Artificial Intelligence Optimization  Book: Applications of Artificial Intelligence Techniques in Industry 4.0 
Hybrid AutoEncoder (HAE) 
Most existing visual search systems are deployed based upon fixed kinds of visual features, which prohibits the feature reusing across different systems or when upgrading systems with a new type of feature. Such a setting is obviously inflexible and time/memory consuming, which is indeed mendable if visual features can be ‘translated’ across systems. In this paper, we make the first attempt towards visual feature translation to break through the barrier of using features across different visual search systems. To this end, we propose a Hybrid AutoEncoder (HAE) to translate visual features, which learns a mapping by minimizing the translation and reconstruction errors. Based upon HAE, an Undirected Affinity Measurement (UAM) is further designed to quantify the affinity among different types of visual features. Extensive experiments have been conducted on several public datasets with 16 different types of widelyused features in visual search systems. Quantitative results show the encouraging possibility of feature translation. And for the first time, the affinity among widelyused features like SIFT and DELF is reported. 
Hybrid Automatic Repeat Request  Realtime remote estimation is critical for missioncritical applications including industrial automation, smart grid and tactile Internet. In this paper, we propose a hybrid automatic repeat request (HARQ)based realtime remote estimation framework for linear timeinvariant (LTI) dynamic systems. Considering the estimation quality of such a system, there is a fundamental tradeoff between the reliability and freshness of the sensor’s measurement transmission. We formulate a new problem to optimize the sensor’s online transmission control policy for static and Markov fading channels, which depends on both the current estimation quality of the remote estimator and the current number of retransmissions of the sensor, so as to minimize the longterm remote estimation mean squared error (MSE). This problem is nontrivial. In particular, it is challenging to derive the condition in terms of the communication channel quality and the LTI system parameters, to ensure a bounded longterm estimation MSE. We derive an elegant sufficient condition of the existence of a stationary and deterministic optimal policy that stabilizes the remote estimation system and minimizes the MSE. Also, we prove that the optimal policy has a switching structure, and accordingly derive a lowcomplexity suboptimal policy. Numerical results show that the proposed optimal policy significantly improves the performance of the remote estimation system compared to the conventional nonHARQ policy. 
Hybrid Consensus Alternating Direction Method of Multipliers (HCADMM) 
The present work introduces the hybrid consensus alternating direction method of multipliers (HCADMM), a novel framework for optimization over networks which unifies existing distributed optimization approaches, including the centralized and the decentralized consensus ADMM. HCADMM provides a flexible tool that leverages the underlying graph topology in order to achieve a desirable sweetspot between nodetonode communication overhead and rate of convergence — thereby alleviating known limitations of both CCADMM and DCADMM. A rigorous analysis of the novel method establishes linear convergence rate, and also guides the choice of parameters to optimize this rate. The novel hybrid update rules of HCADMM lend themselves to ‘innetwork acceleration’ that is shown to effect considerable — and essentially ‘freeofcharge’ — performance boost over the fully decentralized ADMM. Comprehensive numerical tests validate the analysis and showcase the potential of the method in tackling efficiently, widely useful learning tasks. 
Hybrid Contextualized Sentiment Classifier (HCSC) 
The use of user/product information in sentiment analysis is important, especially for coldstart users/products, whose number of reviews are very limited. However, current models do not deal with the coldstart problem which is typical in review websites. In this paper, we present Hybrid Contextualized Sentiment Classifier (HCSC), which contains two modules: (1) a fast word encoder that returns word vectors embedded with short and long range dependency features; and (2) ColdStart Aware Attention (CSAA), an attention mechanism that considers the existence of coldstart problem when attentively pooling the encoded word vectors. HCSC introduces shared vectors that are constructed from similar users/products, and are used when the original distinct vectors do not have sufficient information (i.e. coldstart). This is decided by a frequencyguided selective gate vector. Our experiments show that in terms of RMSE, HCSC performs significantly better when compared with on famous datasets, despite having less complexity, and thus can be trained much faster. More importantly, our model performs significantly better than previous models when the training data is sparse and has coldstart problems. 
Hybrid Cosine Based Convolution  Convolutional neural networks (CNNs) have demonstrated their capability to solve different kind of problems in a very huge number of applications. However, CNNs are limited for their computational and storage requirements. These limitations make difficult to implement these kind of neural networks on embedded devices such as mobile phones, smart cameras or advanced driving assistance systems. In this paper, we present a novel layer named Hybrid Cosine Based Convolution that replaces standard convolutional layers using cosine basis to generate filter weights. The proposed layers provide several advantages: faster convergence in training, the receptive field can be increased at no cost and substantially reduce the number of parameters. We evaluate our proposed layers on three competitive classification tasks where our proposed layers can achieve similar (and in some cases better) performances than VGG and ResNet architectures. 
Hybrid Deep MILP Planner (HDMILPPlan) 
In many realworld planning problems with factored, mixed discrete and continuous state and action spaces such as Reservoir Control, Heating Ventilation, and Air Conditioning, and Navigation domains, it is difficult to obtain a model of the complex nonlinear dynamics that govern state evolution. However, the ubiquity of modern sensors allows us to collect large quantities of data from each of these complex systems and build accurate, nonlinear deep neural network models of their state transitions. But there remains one major problem for the task of control — how can we plan with deep network learned transition models without resorting to Monte Carlo Tree Search and other blackbox transition model techniques that ignore model structure and do not easily extend to mixed discrete and continuous domains? In this paper, we introduce two types of nonlinear planning methods that can leverage deep neural network learned transition models: Hybrid Deep MILP Planner (HDMILPPlan) and Tensorflow Planner (TFPlan). In HDMILPPlan, we make the critical observation that the Rectified Linear Unit transfer function for deep networks not only allows faster convergence of model learning, but also permits a direct compilation of the deep network transition model to a MixedInteger Linear Program encoding. Further, we identify deep network specific optimizations for HDMILPPlan that improve performance over a base encoding and show that we can plan optimally with respect to the learned deep networks. In TFPlan, we take advantage of the efficiency of autodifferentiation tools and GPUbased computation where we encode a subclass of purely continuous planning problems as Recurrent Neural Networks and directly optimize the actions through backpropagation. We compare both planners and show that TFPlan is able to approximate the optimal plans found by HDMILPPlan in less computation time… 
Hybrid Density and PartitionBased Clustering (HyDaP) 
Clustering is an essential technique for discovering patterns in data. The steady increase in amount and complexity of data over the years led to improvements and development of new clustering algorithms. However, algorithms that can cluster data with mixed variable types (continuous and categorical) remain limited, despite the abundance of data with mixed types particularly in the medical field. Among existing methods for mixed data, some posit unverifiable distributional assumptions or that the contributions of different variable types are not well balanced. We propose a twostep hybrid density and partitionbased algorithm (HyDaP) that can detect clusters after variables selection. The first step involves both densitybased and partitionbased algorithms to identify the data structure formed by continuous variables and recognize the important variables for clustering; the second step involves partitionbased algorithm together with a novel dissimilarity measure we designed for mixed data to obtain clustering results. Simulations across various scenarios and data structures were conducted to examine the performance of the HyDaP algorithm compared to commonly used methods. We also applied the HyDaP algorithm on electronic health records to identify sepsis phenotypes. 
Hybrid Dictionary Learning Network (HDLN) 
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. 
Hybrid FilterWrapper Feature Selection Method  HybridFS 
Hybrid Forest  Nowadays with a growing number of online controlling systems in the organization and also a high demand of monitoring and stats facilities that uses data streams to log and control their subsystems, data stream mining becomes more and more vital. Hoeffding Trees (also called Very Fast Decision Trees a.k.a. VFDT) as a Big Data approach in dealing with the data stream for classification and regression problems showed good performance in handling facing challenges and making the possibility of anytime prediction. Although these methods outperform other methods e.g. Artificial Neural Networks (ANN) and Support Vector Regression (SVR), they suffer from high latency in adapting with new concepts when the statistical distribution of incoming data changes. In this article, we introduced a new algorithm that can detect and handle concept drift phenomenon properly. This algorithms also benefits from fast startup ability which helps systems to be able to predict faster than other algorithms at the beginning of data stream arrival. We also have shown that our approach will overperform other controversial approaches for classification and regression tasks. 
Hybrid Intelligent System  Hybrid intelligent system denotes a software system which employs, in parallel, a combination of methods and techniques from artificial intelligence subfields as: · Neurofuzzy systems · hybrid connectionistsymbolic models · Fuzzy expert systems · Connectionist expert systems · Evolutionary neural networks · Genetic fuzzy systems · Rough fuzzy hybridization · Reinforcement learning with fuzzy, neural, or evolutionary methods as well as symbolic reasoning methods. From the cognitive science perspective, every natural intelligent system is hybrid because it performs mental operations on both the symbolic and subsymbolic levels. For the past few years there has been an increasing discussion of the importance of A.I. Systems Integration. Based on notions that there have already been created simple and specific AI systems (such as systems for computer vision, speech synthesis, etc., or software that employs some of the models mentioned above) and now is the time for integration to create broad AI systems. Proponents of this approach are researchers such as Marvin Minsky, Ron Sun, Aaron Sloman, and Michael A. Arbib. An example hybrid is a hierarchical control system in which the lowest, reactive layers are subsymbolic. The higher layers, having relaxed time constraints, are capable of reasoning from an abstract world model and performing planning. Intelligent systems usually rely on hybrid reasoning systems, which include induction, deduction, abduction and reasoning by analogy. 
Hybrid Monte Carlo  In mathematics and physics, the hybrid Monte Carlo algorithm, also known as Hamiltonian Monte Carlo, is a Markov chain Monte Carlo method for obtaining a sequence of random samples from a probability distribution for which direct sampling is difficult. This sequence can be used to approximate the distribution (i.e., to generate a histogram), or to compute an integral (such as an expected value). It differs from the MetropolisHastings algorithm by reducing the correlation between successive sampled states by using a Hamiltonian evolution between states and additionally by targeting states with a higher acceptance criteria than the observed probability distribution. This causes it to converge more quickly to the absolute probability distribution. It was devised by Simon Duane, A.D. Kennedy, Brian Pendleton and Duncan Roweth in 1987. ➚ “Hamiltonian Monte Carlo” 
Hybrid Petri net (PN) 
This paper presents an approach to model an unknown Ladder Logic based Programmable Logic Controller (PLC) program consisting of Boolean logic and counters using Process Mining techniques. First, we tap the inputs and outputs of a PLC to create a data flow log. Second, we propose a method to translate the obtained data flow log to an event log suitable for Process Mining. In a third step, we propose a hybrid Petri net (PN) and neural network approach to approximate the logic of the actual underlying PLC program. We demonstrate the applicability of our proposed approach on a case study with three simulated scenarios. 
Hybrid Predictive Model (HPM) 
Interpretable machine learning has become a strong competitor for traditional blackbox models. However, the possible loss of the predictive performance for gaining interpretability is often inevitable, putting practitioners in a dilemma of choosing between high accuracy (blackbox models) and interpretability (interpretable models). In this work, we propose a novel framework for building a Hybrid Predictive Model (HPM) that integrates an interpretable model with any blackbox model to combine their strengths. The interpretable model substitutes the blackbox model on a subset of data where the blackbox is overkill or nearly overkill, gaining transparency at no or low cost of the predictive accuracy. We design a principled objective function that considers predictive accuracy, model interpretability, and model transparency (defined as the percentage of data processed by the interpretable substitute.) Under this framework, we propose two hybrid models, one substituting with association rules and the other with linear models, and we design customized training algorithms for both models. We test the hybrid models on structured data and text data where interpretable models collaborate with various stateoftheart blackbox models. Results show that hybrid models obtain an efficient tradeoff between transparency and predictive performance, characterized by our proposed efficient frontiers. 
Hybrid Rebeca  In cyberphysical systems like automotive systems, there are components like sensors, actuators, and controllers that communicate asynchronously with each other. The computational model of actor supports modeling distributed asynchronously communicating systems. We propose Hybrid Rebeca language to support modeling of cyberphysical systems. Hybrid Rebeca is an extension of actorbased language Rebeca. In this extension, physical actors are introduced as new computational entities to encapsulate physical behaviors. To support various means of communication among the entities, the network is explicitly modeled as a separate entity from actors. We derive hybrid automata as the basis for analysis of Hybrid Rebeca models. We demonstrate the applicability of our approach through a case study in the domain of automotive systems. We use SpaceEx framework for the analysis of the case study. 
Hybrid Spike Timing Dependent Plasticity (HybridSTDP) 
➘ “Spike Timing Dependent Plasticity” 
Hybrid Task Cascade (HTC) 
Cascade is a classic yet powerful architecture that has boosted performance on various tasks. However, how to introduce cascade to instance segmentation remains an open question. A simple combination of Cascade RCNN and Mask RCNN only brings limited gain. In exploring a more effective approach, we find that the key to a successful instance segmentation cascade is to fully leverage the reciprocal relationship between detection and segmentation. In this work, we propose a new framework, Hybrid Task Cascade (HTC), which differs in two important aspects: (1) instead of performing cascaded refinement on these two tasks separately, it interweaves them for a joint multistage processing; (2) it adopts a fully convolutional branch to provide spatial context, which can help distinguishing hard foreground from cluttered background. Overall, this framework can learn more discriminative features progressively while integrating complementary features together in each stage. Without bells and whistles, a single HTC obtains 38.4% and 1.5% improvement over a strong Cascade Mask RCNN baseline on MSCOCO dataset. More importantly, our overall system achieves 48.6 mask AP on the testchallenge dataset and 49.0 mask AP on testdev, which are the stateoftheart performance. 
Hybrid Transactional / Analytical Processing (HTAP) 
Hybrid Transactional/Analytical Processing (HTAP) is a term used to describe the capability of a single database that can perform both online transaction processing (OLTP) and online analytical processing (OLAP) for the purpose of realtime operational intelligence processing. The term was created by Gartner, Inc., a technology research firm. 
HybridMST  In this paper we present a hybrid active sampling strategy for pairwise preference aggregation, which aims at recovering the underlying rating of the test candidates from sparse and noisy pairwise labelling. Our method employs Bayesian optimization framework and BradleyTerry model to construct the utility function, then to obtain the Expected Information Gain (EIG) of each pair. For computational efficiency, GaussianHermite quadrature is used for estimation of EIG. In this work, a hybrid active sampling strategy is proposed, either using Global Maximum (GM) EIG sampling or Minimum Spanning Tree (MST) sampling in each trial, which is determined by the test budget. The proposed method has been validated on both simulated and realworld datasets, where it shows higher preference aggregation ability than the stateoftheart methods. 
HybridNet  In this paper, we introduce a new model for leveraging unlabeled data to improve generalization performances of image classifiers: a twobranch encoderdecoder architecture called HybridNet. The first branch receives supervision signal and is dedicated to the extraction of invariant classrelated representations. The second branch is fully unsupervised and dedicated to model information discarded by the first branch to reconstruct input data. To further support the expected behavior of our model, we propose an original training objective. It favors stability in the discriminative branch and complementarity between the learned representations in the two branches. HybridNet is able to outperform stateoftheart results on CIFAR10, SVHN and STL10 in various semisupervised settings. In addition, visualizations and ablation studies validate our contributions and the behavior of the model on both CIFAR10 and STL10 datasets. 
HybridSVD  We propose a hybrid algorithm for top$n$ recommendation task that allows to incorporate both user and item side information within the standard collaborative filtering approach. The algorithm extends PureSVD — one of the stateoftheart latent factor models — by exploiting a generalized formulation of the singular value decomposition. This allows to inherit key advantages of the classical algorithm such as highly efficient Lanczosbased optimization procedure, minimal parameter tuning during a model selection phase and a quick foldingin computation to generate recommendations instantly even in a highly dynamic online environment. Within the generalized formulation itself we provide an efficient scheme for side information fusion which avoids undesirable computational overhead and addresses the scalability question. Evaluation of the model is performed in both standard and coldstart scenarios using the datasets with different sparsity levels. We demonstrate in which cases our approach outperforms conventional methods and also provide some intuition on when it may give no significant improvement. 
Hydranet  Accurate estimates of rotation are crucial to visionbased motion estimation in augmented reality and robotics. In this work, we present a method to extract probabilistic estimates of rotation from deep regression models. First, we build on prior work and argue that a multiheaded network structure we name HydraNet provides better calibrated uncertainty estimates than methods that rely on stochastic forward passes. Second, we extend HydraNet to targets that belong to the rotation group, SO(3), by regressing unit quaternions and using the tools of rotation averaging and uncertainty injection onto the manifold to produce threedimensional covariances. Finally, we present results and analysis on a synthetic dataset, learn consistent orientation estimates on the 7Scenes dataset, and show how we can use our learned covariances to fuse deep estimates of relative orientation with classical stereo visual odometry to improve localization on the KITTI dataset. 
HyperAdam  Deep neural networks are traditionally trained using humandesigned stochastic optimization algorithms, such as SGD and Adam. Recently, the approach of learning to optimize network parameters has emerged as a promising research topic. However, these learned blackbox optimizers sometimes do not fully utilize the experience in humandesigned optimizers, therefore have limitation in generalization ability. In this paper, a new optimizer, dubbed as \textit{HyperAdam}, is proposed that combines the idea of ‘learning to optimize’ and traditional Adam optimizer. Given a network for training, its parameter update in each iteration generated by HyperAdam is an adaptive combination of multiple updates generated by Adam with varying decay rates. The combination weights and decay rates in HyperAdam are adaptively learned depending on the task. HyperAdam is modeled as a recurrent neural network with AdamCell, WeightCell and StateCell. It is justified to be stateoftheart for various network training, such as multilayer perceptron, CNN and LSTM. 
Hyperbolic Attention Network  We introduce hyperbolic attention networks to endow neural networks with enough capacity to match the complexity of data with hierarchical and powerlaw structure. A few recent approaches have successfully demonstrated the benefits of imposing hyperbolic geometry on the parameters of shallow networks. We extend this line of work by imposing hyperbolic geometry on the activations of neural networks. This allows us to exploit hyperbolic geometry to reason about embeddings produced by deep networks. We achieve this by reexpressing the ubiquitous mechanism of soft attention in terms of operations defined for hyperboloid and Klein models. Our method shows improvements in terms of generalization on neural machine translation, learning on graphs and visual question answering tasks while keeping the neural representations compact. 
Hyperbolic Bayesian Personalized Ranking (HyperBPR) 
Many wellestablished recommender systems are based on representation learning in Euclidean space. In these models, matching functions such as the Euclidean distance or inner product are typically used for computing similarity scores between user and item embeddings. This paper investigates the notion of learning user and item representations in Hyperbolic space. In this paper, we argue that Hyperbolic space is more suitable for learning useritem embeddings in the recommendation domain. Unlike Euclidean spaces, Hyperbolic spaces are intrinsically equipped to handle hierarchical structure, encouraged by its property of exponentially increasing distances away from origin. We propose HyperBPR (Hyperbolic Bayesian Personalized Ranking), a conceptually simple but highly effective model for the task at hand. Our proposed HyperBPR not only outperforms their Euclidean counterparts, but also achieves stateoftheart performance on multiple benchmark datasets, demonstrating the effectiveness of personalized recommendation in Hyperbolic space. 
Hyperbolic Embedding of ATributed networks (HEAT) 
Finding a low dimensional representation of hierarchical, structured data described by a network remains a challenging problem in the machine learning community. An emerging approach is embedding these networks into hyperbolic space because it can naturally represent a network’s hierarchical structure. However, existing hyperbolic embedding approaches cannot deal with attributed networks, in which nodes are annotated with additional attributes. These attributes might provide additional proximity information to constrain the representations of the nodes, which is important to learn high quality hyperbolic embeddings. To fill this gap, we introduce HEAT (Hyperbolic Embedding of ATributed networks), the first method for embedding attributed networks to a hyperbolic space. HEAT consists of 1) a modified random walk algorithm to obtain training samples that capture both topological and attribute similarity; and 2) a learning algorithm for learning hyperboloid embeddings from the obtained training samples. We show that by leveraging node attributes, HEAT can outperform a stateoftheart Hyperbolic embedding algorithm on several downstream tasks. As a general embedding method, HEAT opens the door to hyperbolic manifold learning on a wide range of attributed and unattributed networks. 
Hyperbolic Neural Network  Hyperbolic spaces have recently gained momentum in the context of machine learning due to their high capacity and treelikeliness properties. However, the representational power of hyperbolic geometry is not yet on par with Euclidean geometry, mostly because of the absence of corresponding hyperbolic neural network layers. This makes it hard to use hyperbolic embeddings in downstream tasks. Here, we bridge this gap in a principled manner by combining the formalism of M\’obius gyrovector spaces with the Riemannian geometry of the Poincar\’e model of hyperbolic spaces. As a result, we derive hyperbolic versions of important deep learning tools: multinomial logistic regression, feedforward and recurrent neural networks such as gated recurrent units. This allows to embed sequential data and perform classification in the hyperbolic space. Empirically, we show that, even if hyperbolic optimization tools are limited, hyperbolic sentence embeddings either outperform or are on par with their Euclidean variants on textual entailment and noisyprefix recognition tasks. 
Hyperbolic Recommender System  Many wellestablished recommender systems are based on representation learning in Euclidean space. In these models, matching functions such as the Euclidean distance or inner product are typically used for computing similarity scores between user and item embeddings. This paper investigates the notion of learning user and item representations in Hyperbolic space. In this paper, we argue that Hyperbolic space is more suitable for learning useritem embeddings in the recommendation domain. Unlike Euclidean spaces, Hyperbolic spaces are intrinsically equipped to handle hierarchical structure, encouraged by its property of exponentially increasing distances away from origin. We propose HyperBPR (Hyperbolic Bayesian Personalized Ranking), a conceptually simple but highly effective model for the task at hand. Our proposed HyperBPR not only outperforms their Euclidean counterparts, but also achieves stateoftheart performance on multiple benchmark datasets, demonstrating the effectiveness of personalized recommendation in Hyperbolic space. 
Hyperdata  Hyperdata indicates data objects linked to other data objects in other places, as hypertext indicates text linked to other text in other places. Hyperdata enables formation of a web of data, evolving from the “data on the Web” that is not interrelated (or at least, not linked). In the same way that hypertext usually refers to the World Wide Web but is a broader term, hyperdata usually refers to the Semantic Web, but may also be applied more broadly to other datalinking technologies such as Microformats – including XHTML Friends Network. 
HyperDenseNet  Recently, dense connections have attracted substantial attention in computer vision because they facilitate gradient flow and implicit deep supervision during training. Particularly, DenseNet, which connects each layer to every other layer in a feedforward fashion, has shown impressive performances in natural image classification tasks. We propose HyperDenseNet, a 3D fully convolutional neural network that extends the definition of dense connectivity to multimodal segmentation problems. Each imaging modality has a path, and dense connections occur not only between the pairs of layers within the same path, but also between those across different paths. This contrasts with the existing multimodal CNN approaches, in which modeling several modalities relies entirely on a single joint layer (or level of abstraction) for fusion, typically either at the input or at the output of the network. Therefore, the proposed network has total freedom to learn more complex combinations between the modalities, within and inbetween all the levels of abstraction, which increases significantly the learning representation. We report extensive evaluations over two different and highly competitive multimodal brain tissue segmentation challenges, iSEG 2017 and MRBrainS 2013, with the former focusing on 6month infant data and the latter on adult images. HyperDenseNet yielded significant improvements over many stateoftheart segmentation networks, ranking at the top on both benchmarks. We further provide a comprehensive experimental analysis of features reuse, which confirms the importance of hyperdense connections in multimodal representation learning. Our code is publicly available at https://…/HyperDenseNet. 
HyperFusionNet  Salient object detection (SOD), which aims to find the most important region of interest and segment the relevant object/item in that area, is an important yet challenging vision task. This problem is inspired by the fact that human seems to perceive main scene elements with high priorities. Thus, accurate detection of salient objects in complex scenes is critical for humancomputer interaction. In this paper, we present a novel feature learning framework for SOD, in which we cast the SOD as a pixelwise classification problem. The proposed framework utilizes a densely hierarchical feature fusion network, named HyperFusionNet, automatically predicts the most important area and segments the associated objects in an endtoend manner. Specifically, inspired by the human perception system and image reflection separation, we first decompose input images into reflective image pairs by contentpreserving transforms. Then, the complementary information of reflective image pairs is jointly extracted by an interweaved convolutional neural network (ICNN) and hierarchically combined with a hyperdense fusion mechanism. Based on the fused multiscale features, our method finally achieves a promising way of predicting SOD. As shown in our extensive experiments, the proposed method consistently outperforms other stateoftheart methods on seven public datasets with a large margin. 
HyperGAN  We introduce HyperGAN, a generative network that learns to generate all the weights within a deep neural network. HyperGAN employs a novel mixer to transform independent Gaussian noise into a latent space where dimensions are correlated, which is then transformed to generate weights in each layer of a deep neural network. We utilize an architecture that bears resemblance to generative adversarial networks, but we evaluate the likelihood of samples with a classification loss. This is equivalent to minimizing the KLdivergence between the generated network parameter distribution and an unknown true parameter distribution. We apply HyperGAN to classification, showing that HyperGAN can learn to generate parameters which solve the MNIST and CIFAR10 datasets with competitive performance to fully supervised learning, while learning a rich distribution of effective parameters. We also show that HyperGAN can also provide better uncertainty than standard ensembles. This is evaluated by the ability of HyperGAN generated ensembles to detect out of distribution data as well as adversarial examples. We see that in addition to being highly accurate on inlier data, HyperGAN can provide reasonable uncertainty estimates. 
HyperGCN  Graphbased semisupervised learning (SSL) is an important learning problem where the goal is to assign labels to initially unlabeled nodes in a graph. Graph Convolutional Networks (GCNs) have recently been shown to be effective for graphbased SSL problems. GCNs inherently assume existence of pairwise relationships in the graphstructured data. However, in many realworld problems, relationships go beyond pairwise connections and hence are more complex. Hypergraphs provide a natural modeling tool to capture such complex relationships. In this work, we explore the use of GCNs for hypergraphbased SSL. In particular, we propose HyperGCN, an SSL method which uses a layerwise propagation rule for convolutional neural networks operating directly on hypergraphs. To the best of our knowledge, this is the first principled adaptation of GCNs to hypergraphs. HyperGCN is able to encode both the hypergraph structure and hypernode features in an effective manner. Through detailed experimentation, we demonstrate HyperGCN’s effectiveness at hypergraphbased SSL. 
Hypergraph Neural Network (HGNN) 
In this paper, we present a hypergraph neural networks (HGNN) framework for data representation learning, which can encode highorder data correlation in a hypergraph structure. Confronting the challenges of learning representation for complex data in real practice, we propose to incorporate such data structure in a hypergraph, which is more flexible on data modeling, especially when dealing with complex data. In this method, a hyperedge convolution operation is designed to handle the data correlation during representation learning. In this way, traditional hypergraph learning procedure can be conducted using hyperedge convolution operations efficiently. HGNN is able to learn the hidden layer representation considering the highorder data structure, which is a general framework considering the complex data correlations. We have conducted experiments on citation network classification and visual object recognition tasks and compared HGNN with graph convolutional networks and other traditional methods. Experimental results demonstrate that the proposed HGNN method outperforms recent stateoftheart methods. We can also reveal from the results that the proposed HGNN is superior when dealing with multimodal data compared with existing methods. 
Hypergraph Null Model  Clustering on hypergraphs has been garnering increased attention with potential applications in network analysis, VLSI design and computer vision, among others. In this work, we generalize the framework of modularity maximization for clustering on hypergraphs. To this end, we introduce a hypergraph null model, analogous to the configuration model on undirected graphs, and a nodedegree preserving reduction to work with this model. This is used to define a modularity function that can be maximized using the popular and fast Louvain algorithm. We additionally propose a refinement over this clustering, by reweighting cut hyperedges in an iterative fashion. The efficacy and efficiency of our methods are demonstrated on several realworld datasets. 
Hypergraphbased Outlier Test for Categorical Data (HOT) 
As a widely used data mining technique, outlier detection is a process which aims to find anomalies while providing good explanations. Most existing detection methods are basically designed for numeric data, however, reallife data such as web pages, business transactions and bioinformatics records always contain categorical data. So it causes difficulty to find reasonable exceptions in the real world applications. In this paper, we introduce a novel outlier mining method based on hypergraph model for categorical data. Since hy pergraphs precisely capture the distribution characteristics in data subspaces, this method is effective in identifying anomalies in dense subspaces and presents good interpre tations for the local outlierness. By selecting the most rel evant subspaces, the problem of ‘curse of dimensionality’ in very large databases can also be ameliorated. Further more, the connectivity property is used to replace the dis tance metrics, so that the distancebased computation is not needed anymore, which enhances the robustness for han dling missingvalue data. The fact that connectivity com putation facilitates the aggregation operations supported by most SQLcompatible database systems, makes the mining process much efficient. Finally, we give experiments and analysis which show that our method can find outliers in categorical data with good performance and quality. 
HyperHeuristics  A hyperheuristic is a heuristic search method that seeks to automate, often by the incorporation of machine learning techniques, the process of selecting, combining, generating or adapting several simpler heuristics (or components of such heuristics) to efficiently solve computational search problems. One of the motivations for studying hyperheuristics is to build systems which can handle classes of problems rather than solving just one problem. There might be multiple heuristics from which one can choose for solving a problem, and each heuristic has its own strength and weakness. The idea is to automatically devise algorithms by combining the strength and compensating for the weakness of known heuristics. In a typical hyperheuristic framework there is a highlevel methodology and a set of lowlevel heuristics (either constructive or perturbative heuristics). Given a problem instance, the highlevel method selects which lowlevel heuristic should be applied at any given time, depending upon the current problem state, or search stage. 
HyperKron Graph  Graph models have long been used in lieu of real data which can be expensive and hard to come by. A common class of models constructs a matrix of probabilities, and samples an adjacency matrix by flipping a weighted coin for each entry. Examples include the Erd\H{o}sR\'{e}nyi model, ChungLu model, and the Kronecker model. Here we present the HyperKron Graph model: an extension of the Kronecker Model, but with a distribution over hyperedges. We prove that we can efficiently generate graphs from this model in order proportional to the number of edges times a small logfactor, and find that in practice the runtime is linear with respect to the number of edges. We illustrate a number of useful features of the HyperKron model including nontrivial clustering and highly skewed degree distributions. Finally, we fit the HyperKron model to realworld networks, and demonstrate the model’s flexibility with a complex application of the HyperKron model to networks with coherent feedforward loops. 
HyperlinkInduced Topic Search (HITS) 
HyperlinkInduced Topic Search (HITS; also known as hubs and authorities) is a link analysis algorithm that rates Web pages, developed by Jon Kleinberg. The idea behind Hubs and Authorities stemmed from a particular insight into the creation of web pages when the Internet was originally forming; that is, certain web pages, known as hubs, served as large directories that were not actually authoritative in the information that it held, but were used as compilations of a broad catalog of information that led users directly to other authoritative pages. In other words, a good hub represented a page that pointed to many other pages, and a good authority represented a page that was linked by many different hubs. The scheme therefore assigns two scores for each page: its authority, which estimates the value of the content of the page, and its hub value, which estimates the value of its links to other pages. Network Analysis for Wikipedia HITS Algorithm – Hubs and Authorities on the Internet 
HyperLogLog  HyperLogLog is an algorithm for the countdistinct problem, approximating the number of distinct elements in a multiset (the cardinality). Calculating the exact cardinality of a multiset requires an amount of memory proportional to the cardinality, which is impractical for very large data sets. Probabilistic cardinality estimators, such as the HyperLogLog algorithm, use significantly less memory than this, at the cost of obtaining only an approximation of the cardinality. The HyperLogLog algorithm is able to estimate cardinalities of with a typical accuracy of 2%, using 1.5kB of memory. HyperLogLog is an extension of the earlier LogLog algorithm. 
HyperMapper 2.0  Multiobjective optimization is a crucial matter in computer systems design space exploration because realworld applications often rely on a tradeoff between several objectives. Derivatives are usually not available or impractical to compute and the feasibility of an experiment can not always be determined in advance. These problems are particularly difficult when the feasible region is relatively small, and it may be prohibitive to even find a feasible experiment, let alone an optimal one. We introduce a new methodology and corresponding software framework, HyperMapper 2.0, which handles multiobjective optimization, unknown feasibility constraints, and categorical/ordinal variables. This new methodology also supports injection of user prior knowledge in the search when available. All of these features are common requirements in computer systems but rarely exposed in existing design space exploration systems. The proposed methodology follows a whitebox model which is simple to understand and interpret (unlike, for example, neural networks) and can be used by the user to better understand the results of the automatic search. We apply and evaluate the new methodology to automatic static tuning of hardware accelerators within the recently introduced Spatial programming language, with minimization of design runtime and compute logic under the constraint of the design fitting in a target field programmable gate array chip. Our results show that HyperMapper 2.0 provides better Pareto fronts compared to stateoftheart baselines, with better or competitive hypervolume indicator and with 8x improvement in sampling budget for most of the benchmarks explored. 
Hyperparameter  In Bayesian statistics, a hyperparameter is a parameter of a prior distribution; the term is used to distinguish them from parameters of the model for the underlying system under analysis. For example, if one is using a beta distribution to model the distribution of the parameter p of a Bernoulli distribution, then: · p is a parameter of the underlying system (Bernoulli distribution), and · alpha and beta are parameters of the prior distribution (beta distribution), hence hyperparameters One may take a single value for a given hyperparameter, or one can iterate and take a probability distribution on the hyperparameter itself, called a hyperprior. State of Hyperparameter Selection 
Hyperparameter Optimisation on the Fly  The performance of policy gradient methods is sensitive to hyperparameter settings that must be tuned for any new application. Widely used grid search methods for tuning hyperparameters are sample inefficient and computationally expensive. More advanced methods like Population Based Training that learn optimal schedules for hyperparameters instead of fixed settings can yield better results, but are also sample inefficient and computationally expensive. In this paper, we propose Hyperparameter Optimisation on the Fly (HOOF), a gradientfree metalearning algorithm that can automatically learn an optimal schedule for hyperparameters that affect the policy update directly through the gradient. The main idea is to use existing trajectories sampled by the policy gradient method to optimise a onestep improvement objective, yielding a sample and computationally efficient algorithm that is easy to implement. Our experimental results across multiple domains and algorithms show that using HOOF to learn these hyperparameter schedules leads to faster learning with improved performance. 
Hyperparameter Optimization  In the context of machine learning, hyperparameter optimization or model selection is the problem of choosing a set of hyperparameters for a learning algorithm, usually with the goal of obtaining good generalization. Hyperparameter optimization contrasts with actual learning problems, which are also often cast as optimization problems, but optimize a loss function on the training set alone. In effect, learning algorithms learn parameters that model/reconstruct their inputs well, while hyperparameter optimization is to ensure the model does not overfit its data by tuning, e.g., regularization. 
Hyperspectral Data Augmentation  Data augmentation is a popular technique which helps improve generalization capabilities of deep neural networks. It plays a pivotal role in remotesensing scenarios in which the amount of highquality ground truth data is limited, and acquiring new examples is costly or impossible. This is a common problem in hyperspectral imaging, where manual annotation of image data is difficult, expensive, and prone to human bias. In this letter, we propose online data augmentation of hyperspectral data which is executed during the inference rather than before the training of deep networks. This is in contrast to all other stateoftheart hyperspectral augmentation algorithms which increase the size (and representativeness) of training sets. Additionally, we introduce a new principal component analysis based augmentation. The experiments revealed that our data augmentation algorithms improve generalization of deep networks, work in realtime, and the online approach can be effectively combined with offline techniques to enhance the classification accuracy. 
Hyperspherical Convolution (SphereConv) 
Convolution as inner product has been the founding basis of convolutional neural networks (CNNs) and the key to endtoend visual representation learning. Benefiting from deeper architectures, recent CNNs have demonstrated increasingly strong representation abilities. Despite such improvement, the increased depth and larger parameter space have also led to challenges in properly training a network. In light of such challenges, we propose hyperspherical convolution (SphereConv), a novel learning framework that gives angular representations on hyperspheres. We introduce SphereNet, deep hyperspherical convolution networks that are distinct from conventional inner product based convolutional networks. In particular, SphereNet adopts SphereConv as its basic convolution operator and is supervised by generalized angular softmax loss – a natural loss formulation under SphereConv. We show that SphereNet can effectively encode discriminative representation and alleviate training difficulty, leading to easier optimization, faster convergence and comparable (even better) classification accuracy over convolutional counterparts. We also provide some theoretical insights for the advantages of learning on hyperspheres. In addition, we introduce the learnable SphereConv, i.e., a natural improvement over prefixed SphereConv, and SphereNorm, i.e., hyperspherical learning as a normalization method. Experiments have verified our conclusions. 
Hyperspherical Prototype Network (HPN) 
This paper introduces hyperspherical prototype networks, which unify regression and classification by prototypes on hyperspherical output spaces. Rather than defining prototypes as the mean output vector over training examples per class, we propose hyperspheres as output spaces to define class prototypes a priori with large margin separation. By doing so, we do not require any prototype updating, we can handle any training size, and the output dimensionality is no longer constrained to the number of classes. Furthermore, hyperspherical prototype networks generalize to regression, by optimizing outputs as an interpolation between two prototypes on the hypersphere. Since both tasks are now defined by the same loss function, they can be jointly optimized for multitask problems. Experimental evaluation shows the benefits of hyperspherical prototype networks for classification, regression, and their combination. 
Hyperspherical Variational AutoEncoder  The Variational AutoEncoder (VAE) is one of the most used unsupervised machine learning models. But although the default choice of a Gaussian distribution for both the prior and posterior represents a mathematically convenient distribution often leading to competitive results, we show that this parameterization fails to model data with a latent hyperspherical structure. To address this issue we propose using a von MisesFisher (vMF) distribution instead, leading to a hyperspherical latent space. Through a series of experiments we show how such a hyperspherical VAE, or $\mathcal{S}$VAE, is more suitable for capturing data with a hyperspherical latent structure, while outperforming a normal, $\mathcal{N}$VAE, in low dimensions on other data types. 
HyperTools  A python toolbox for gaining geometric insights into highdimensional data. 
HyperTrick  Training intelligent agents through reinforcement learning is a notoriously unstable procedure. Massive parallelization on GPUs and distributed systems has been exploited to generate a large amount of training experiences and consequently reduce instabilities, but the success of training remains strongly influenced by the choice of the hyperparameters. To overcome this issue, we introduce HyperTrick, a new metaoptimization algorithm, and show its effective application to tune hyperparameters in the case of deep reinforcement learning, while learning to play different Atari games on a distributed system. Our analysis provides evidence of the interaction between the identification of the optimal hyperparameters and the learned policy, that is typical of the case of metaoptimization for deep reinforcement learning. When compared with stateoftheart metaoptimization algorithms, HyperTrick is characterized by a simpler implementation and it allows learning similar policies, while making a more effective use of the computational resources in a distributed system. 
Hypervariate Data  Hypervariate data is Data with four or more dimensions in the dataset. Dartmouth College researchers have published a free Python software package called HyperTools that allows users to turn complex data into 3D shapes or animations. The tool allows users to visualize patterns in their data and compare the characteristics of different datasets, which in turn could inform researchers on how to train their machine learning algorithms by illuminating differences between groups of data. Additionally, the Dartmouth researchers have published tutorials for HyperTools and a gallery of examples, such as how to plot the text of State of the Union addresses, to help users create visualizations. 
Hypervolume Under Manifold (HUM) 
Paper: Jialiang Li (2008) <doi:10.1093/biostatistics/kxm050>. Jialiang Li (2014) <doi:10.3109/1354750X.2013.868516>. mcca 
Hypothesistestingbased Adaptive Spline Filtering (HASF) 
Trend Analysis of Fragmented Time Series for mHealth Apps: Hypothesis Testing Based Adaptive Spline Filtering Method with Importance Weighting 
Hypothesizing After the Results are Known (HARK) 
Recent advancements in machine learning research, i.e., deep learning, introduced methods that excel conventional algorithms as well as humans in several complex tasks, ranging from detection of objects in images and speech recognition to playing difficult strategic games. However, the current methodology of machine learning research and consequently, implementations of the realworld applications of such algorithms, seems to have a recurring HARKing (Hypothesizing After the Results are Known) issue. In this work, we elaborate on the algorithmic, economic and social reasons and consequences of this phenomenon. We present examples from current common practices of conducting machine learning research (e.g. avoidance of reporting negative results) and failure of generalization ability of the proposed algorithms and datasets in actual reallife usage. Furthermore, a potential future trajectory of machine learning research and development from the perspective of accountable, unbiased, ethical and privacyaware algorithmic decision making is discussed. We would like to emphasize that with this discussion we neither claim to provide an exhaustive argumentation nor blame any specific institution or individual on the raised issues. This is simply a discussion put forth by us, insiders of the machine learning field, reflecting on us. 
Advertisements