aid stringlengths 9 15 | mid stringlengths 7 10 | abstract stringlengths 78 2.56k | related_work stringlengths 92 1.77k | ref_abstract dict |
|---|---|---|---|---|
1905.09438 | 2946580944 | Reading Comprehension has received significant attention in recent years as high quality Question Answering (QA) datasets have become available. Despite state-of-the-art methods achieving strong overall accuracy, Multi-Hop (MH) reasoning remains particularly challenging. To address MH-QA specifically, we propose a Deep Reinforcement Learning based method capable of learning sequential reasoning across large collections of documents so as to pass a query-aware, fixed-size context subset to existing models for answer extraction. Our method is comprised of two stages: a linker, which decomposes the provided support documents into a graph of sentences, and an extractor, which learns where to look based on the current question and already-visited sentences. The result of the linker is a novel graph structure at the sentence level that preserves logical flow while still allowing rapid movement between documents. Importantly, we demonstrate that the sparsity of the resultant graph is invariant to context size. This translates to fewer decisions required from the Deep-RL trained extractor, allowing the system to scale effectively to large collections of documents. The importance of sequential decision making in the document traversal step is demonstrated by comparison to standard IE methods, and we additionally introduce a BM25-based IR baseline that retrieves documents relevant to the query only. We examine the integration of our method with existing models on the recently proposed QAngaroo benchmark and achieve consistent increases in accuracy across the board, as well as a 2-3x reduction in training time. | Graph-based document traversal has been applied in the QA context. @cite_5 utilise coreference relations to set the order in which tokens are presented to a recurrent network. Other systems typically construct a Knowledge Graph (KG) from text, through the use of a separate Information Extraction system and learn to reason about the types of links to explore @cite_9 . In this vein, Variational Reasoning Networks @cite_15 have recently being proposed specifically for the multi-hop case, however this work is implemented in a context where an external KG is available. | {
"cite_N": [
"@cite_5",
"@cite_9",
"@cite_15"
],
"mid": [
"2604685013",
"2251287417",
"2963068946"
],
"abstract": [
"Training recurrent neural networks to model long term dependencies is difficult. Hence, we propose to use external linguistic knowledge as an explicit signal to inform the model which memories it should utilize. Specifically, external knowledge is used to augment a sequence with typed edges between arbitrarily distant elements, and the resulting graph is decomposed into directed acyclic subgraphs. We introduce a model that encodes such graphs as explicit memory in recurrent neural networks, and use it to model coreference relations in text. We apply our model to several text comprehension tasks and achieve new state-of-the-art results on all considered benchmarks, including CNN, bAbi, and LAMBADA. On the bAbi QA tasks, our model solves 15 out of the 20 tasks with only 1000 training examples per task. Analysis of the learned representations further demonstrates the ability of our model to encode fine-grained entity information across a document.",
"Answering natural language questions over a knowledge base is an important and challenging task. Most of existing systems typically rely on hand-crafted features and rules to conduct question understanding and or answer ranking. In this paper, we introduce multi-column convolutional neural networks (MCCNNs) to understand questions from three different aspects (namely, answer path, answer context, and answer type) and learn their distributed representations. Meanwhile, we jointly learn low-dimensional embeddings of entities and relations in the knowledge base. Question-answer pairs are used to train the model to rank candidate answers. We also leverage question paraphrases to train the column networks in a multi-task learning manner. We use FREEBASE as the knowledge base and conduct extensive experiments on the WEBQUESTIONS dataset. Experimental results show that our method achieves better or comparable performance compared with baseline systems. In addition, we develop a method to compute the salience scores of question words in different column networks. The results help us intuitively understand what MCCNNs learn.",
""
]
} |
1905.09638 | 2945850646 | This paper demonstrates a novel method for separately estimating aleatoric risk and epistemic uncertainty in deep reinforcement learning. Aleatoric risk, which arises from inherently stochastic environments or agents, must be accounted for in the design of risk-sensitive algorithms. Epistemic uncertainty, which stems from limited data, is important both for risk-sensitivity and to efficiently explore an environment. We first present a Bayesian framework for learning the return distribution in reinforcement learning, which provides theoretical foundations for quantifying both types of uncertainty. Based on this framework, we show that the disagreement between only two neural networks is sufficient to produce a low-variance estimate of the epistemic uncertainty on the return distribution, thus providing a simple and computationally cheap uncertainty metric. We demonstrate experiments that illustrate our method and some applications. | There has been increasing interest in developing practical methods for uncertainty estimation in reinforcement learning, with prior work mostly focusing either on aleatoric risk or epistemic uncertainty. For the estimation of epistemic uncertainty in complex environments, several methods have been proposed. Pseudo-counts @cite_38 @cite_13 that approximate the number of visits of states or state-action pairs by an agent can be interpreted as an uncertainty measure. Bayesian inference techniques over the parameters that define the value function have been demonstrated @cite_23 @cite_5 @cite_36 . The disagreement between the predictions of an ensemble of neural networks has also been proposed as a way of estimating uncertainty @cite_15 @cite_27 @cite_9 @cite_28 @cite_31 @cite_17 @cite_40 . Dropout in neural networks has also been used @cite_43 @cite_30 . Our work builds on epistemic uncertainty estimation using dropout @cite_43 and anchored neural networks @cite_28 and proposes an expansion to the distributional setting, thus allowing us to consider both aleatoric and epistemic uncertainty in a single framework. | {
"cite_N": [
"@cite_30",
"@cite_38",
"@cite_31",
"@cite_36",
"@cite_28",
"@cite_9",
"@cite_43",
"@cite_27",
"@cite_40",
"@cite_23",
"@cite_5",
"@cite_15",
"@cite_13",
"@cite_17"
],
"mid": [
"2774915412",
"2963276097",
"2962723954",
"2963797557",
"2806909761",
"2963238274",
"2964059111",
"2963938771",
"2899205164",
"2962767126",
"2962996309",
"2129832613",
"2561776174",
"2898045695"
],
"abstract": [
"This paper studies directed exploration for reinforcement learning agents by tracking uncertainty about the value of each available action. We identify two sources of uncertainty that are relevant for exploration. The first originates from limited data (parametric uncertainty), while the second originates from the distribution of the returns (return uncertainty). We identify methods to learn these distributions with deep neural networks, where we estimate parametric uncertainty with Bayesian drop-out, while return uncertainty is propagated through the Bellman equation as a Gaussian distribution. Then, we identify that both can be jointly estimated in one network, which we call the Double Uncertain Value Network. The policy is directly derived from the learned distributions based on Thompson sampling. Experimental results show that both types of uncertainty may vastly improve learning in domains with a strong exploration challenge.",
"We consider an agent's uncertainty about its environment and the problem of generalizing this uncertainty across states. Specifically, we focus on the problem of exploration in non-tabular reinforcement learning. Drawing inspiration from the intrinsic motivation literature, we use density models to measure uncertainty, and propose a novel algorithm for deriving a pseudo-count from an arbitrary density model. This technique enables us to generalize count-based exploration algorithms to the non-tabular case. We apply our ideas to Atari 2600 games, providing sensible pseudo-counts from raw pixels. We transform these pseudo-counts into exploration bonuses and obtain significantly improved exploration in a number of hard games, including the infamously difficult MONTEZUMA'S REVENGE.",
"Dealing with uncertainty is essential for efficient reinforcement learning. There is a growing literature on uncertainty estimation for deep learning from fixed datasets, but many of the most popular approaches are poorly-suited to sequential decision problems. Other methods, such as bootstrap sampling, have no mechanism for uncertainty that does not come from the observed data. We highlight why this can be a crucial shortcoming and propose a simple remedy through addition of a randomized untrainable prior' network to each ensemble member. We prove that this approach is efficient with linear representations, provide simple illustrations of its efficacy with nonlinear representations and show that this approach scales to large-scale problems far better than previous attempts.",
"We propose Bayesian Deep Q-Network (BDQN), a practical Thompson sampling based Reinforcement Learning (RL) Algorithm. Thompson sampling allows for targeted exploration in high dimensions through posterior sampling but is usually computationally expensive. We address this limitation by introducing uncertainty only at the output layer of the network through a Bayesian Linear Regression (BLR) model, which can be trained with fast closed-form updates and its samples can be drawn efficiently through the Gaussian distribution. We apply our method to a wide range of Atari games in Arcade Learning Environments. Since BDQN carries out more efficient exploration, it is able to reach higher rewards substantially faster than a key baseline, double deep Q network DDQN.",
"",
"Deep neural networks (NNs) are powerful black box predictors that have recently achieved impressive performance on a wide spectrum of tasks. Quantifying predictive uncertainty in NNs is a challenging and yet unsolved problem. Bayesian NNs, which learn a distribution over weights, are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and are computationally expensive compared to standard (non-Bayesian) NNs. We propose an alternative to Bayesian NNs that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. Through a series of experiments on classification and regression benchmarks, we demonstrate that our method produces well-calibrated uncertainty estimates which are as good or better than approximate Bayesian NNs. To assess robustness to dataset shift, we evaluate the predictive uncertainty on test examples from known and unknown distributions, and show that our method is able to express higher uncertainty on out-of-distribution examples. We demonstrate the scalability of our method by evaluating predictive uncertainty estimates on ImageNet.",
"Deep learning tools have gained tremendous attention in applied machine learning. However such tools for regression and classification do not capture model uncertainty. In comparison, Bayesian models offer a mathematically grounded framework to reason about model uncertainty, but usually come with a prohibitive computational cost. In this paper we develop a new theoretical framework casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes. A direct result of this theory gives us tools to model uncertainty with dropout NNs - extracting information from existing models that has been thrown away so far. This mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy. We perform an extensive study of the properties of dropout's uncertainty. Various network architectures and nonlinearities are assessed on tasks of regression and classification, using MNIST as an example. We show a considerable improvement in predictive log-likelihood and RMSE compared to existing state-of-the-art methods, and finish by using dropout's uncertainty in deep reinforcement learning.",
"Efficient exploration remains a major challenge for reinforcement learning (RL). Common dithering strategies for exploration, such as '-greedy, do not carry out temporally-extended (or deep) exploration; this can lead to exponentially larger data requirements. However, most algorithms for statistically efficient RL are not computationally tractable in complex environments. Randomized value functions offer a promising approach to efficient exploration with generalization, but existing algorithms are not compatible with nonlinearly parameterized value functions. As a first step towards addressing such contexts we develop bootstrapped DQN. We demonstrate that bootstrapped DQN can combine deep exploration with deep neural networks for exponentially faster learning than any dithering strategy. In the Arcade Learning Environment bootstrapped DQN substantially improves learning speed and cumulative performance across most games.",
"We introduce an exploration bonus for deep reinforcement learning methods that is easy to implement and adds minimal overhead to the computation performed. The bonus is the error of a neural network predicting features of the observations given by a fixed randomly initialized neural network. We also introduce a method to flexibly combine intrinsic and extrinsic rewards. We find that the random network distillation (RND) bonus combined with this increased flexibility enables significant progress on several hard exploration Atari games. In particular we establish state of the art performance on Montezuma's Revenge, a game famously difficult for deep reinforcement learning methods. To the best of our knowledge, this is the first method that achieves better than average human performance on this game without using demonstrations or having access to the underlying state of the game, and occasionally completes the first level.",
"We propose randomized least-squares value iteration (RLSVI) - a new reinforcement learning algorithm designed to explore and generalize efficiently via linearly parameterized value functions. We explain why versions of least-squares value iteration that use Boltzmann or e-greedy exploration can be highly inefficient, and we present computational results that demonstrate dramatic efficiency gains enjoyed by RLSVI. Further, we establish an upper bound on the expected regret of RLSVI that demonstrates nearoptimality in a tabula rasa learning context. More broadly, our results suggest that randomized value functions offer a promising approach to tackling a critical challenge in reinforcement learning: synthesizing efficient exploration and effective generalization.",
"",
"We discuss a number of methods for estimating the standard error of predicted values from a multilayer perceptron. These methods include the delta method based on the Hessian, bootstrap estimators, and the “sandwich” estimator. The methods are described and compared in a number of examples. We find that the bootstrap methods perform best, partly because they capture variability due to the choice of starting weights.",
"Count-based exploration algorithms are known to perform near-optimally when used in conjunction with tabular reinforcement learning (RL) methods for solving small discrete Markov decision processes (MDPs). It is generally thought that count-based methods cannot be applied in high-dimensional state spaces, since most states will only occur once. Recent deep RL exploration strategies are able to deal with high-dimensional continuous state spaces through complex heuristics, often relying on optimism in the face of uncertainty or intrinsic motivation. @PARASPLIT In this work, we describe a surprising finding: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and or continuous deep RL benchmarks. States are mapped to hash codes, which allows to count their occurrences with a hash table. These counts are then used to compute a reward bonus according to the classic count-based exploration theory. We find that simple hash functions can achieve surprisingly good results on many challenging tasks. Furthermore, we show that a domain-dependent learned hash code may further improve these results. @PARASPLIT Detailed analysis reveals important aspects of a good hash function: 1) having appropriate granularity and 2) encoding information relevant to solving the MDP. This exploration strategy achieves near state-of-the-art performance on both continuous control tasks and Atari 2600 games, hence providing a simple yet powerful baseline for solving MDPs that require considerable exploration.",
"Many current autonomous systems are being designed with a strong reliance on black box predictions from deep neural networks (DNNs). However, DNNs tend to be overconfident in predictions on unseen data and can give unpredictable results for far-from-distribution test data. The importance of predictions that are robust to this distributional shift is evident for safety-critical applications, such as collision avoidance around pedestrians. Measures of model uncertainty can be used to identify unseen data, but the state-of-the-art extraction methods such as Bayesian neural networks are mostly intractable to compute. This paper uses MC-Dropout and Bootstrapping to give computationally tractable and parallelizable uncertainty estimates. The methods are embedded in a Safe Reinforcement Learning framework to form uncertainty-aware navigation around pedestrians. The result is a collision avoidance policy that and cautiously avoids pedestrians that exhibit unseen behavior. The policy is demonstrated in simulation to be more robust to novel observations and take safer actions than an uncertainty-unaware baseline."
]
} |
1905.09638 | 2945850646 | This paper demonstrates a novel method for separately estimating aleatoric risk and epistemic uncertainty in deep reinforcement learning. Aleatoric risk, which arises from inherently stochastic environments or agents, must be accounted for in the design of risk-sensitive algorithms. Epistemic uncertainty, which stems from limited data, is important both for risk-sensitivity and to efficiently explore an environment. We first present a Bayesian framework for learning the return distribution in reinforcement learning, which provides theoretical foundations for quantifying both types of uncertainty. Based on this framework, we show that the disagreement between only two neural networks is sufficient to produce a low-variance estimate of the epistemic uncertainty on the return distribution, thus providing a simple and computationally cheap uncertainty metric. We demonstrate experiments that illustrate our method and some applications. | Prior work on aleatoric risk in reinforcement learning has focused on the stochastic nature of the returns, which can be caused either by randomness inherent to the environment or randomness in the agent's actions. Several approaches have been developed to estimate higher-order terms of the return distribution @cite_1 @cite_34 @cite_6 . More recently, distributional reinforcement learning algorithms have been developed that aim to learn the entire distribution of the returns @cite_16 @cite_7 . Our work builds on this distributional approach by proposing a method for estimating epistemic uncertainty as well in this setting. | {
"cite_N": [
"@cite_7",
"@cite_1",
"@cite_6",
"@cite_16",
"@cite_34"
],
"mid": [
"2963423916",
"2313791856",
"2356031020",
"2102288976",
"2141203641"
],
"abstract": [
"In this paper we argue for the fundamental importance of the value distribution: the distribution of the random return received by a reinforcement learning agent. This is in contrast to the common approach to reinforcement learning which models the expectation of this return, or value. Although there is an established body of literature studying the value distribution, thus far it has always been used for a specific purpose such as implementing risk-aware behaviour. We begin with theoretical results in both the policy evaluation and control settings, exposing a significant distributional instability in the latter. We then use the distributional perspective to design a new algorithm which applies Bellman's equation to the learning of approximate value distributions. We evaluate our algorithm using the suite of games from the Arcade Learning Environment. We obtain both state-of-the-art results and anecdotal evidence demonstrating the importance of the value distribution in approximate reinforcement learning. Finally, we combine theoretical and empirical evidence to highlight the ways in which the value distribution impacts learning in the approximate setting.",
"Formulae are presented for the variance and higher moments of the present value of single-stage rewards in a finite Markov decision process. Similar formulae are exhibited for a semi-Markov decision process. There is a short discussion of the obstacles to using the variance formula in algorithms to maximize the mean minus a multiple of the standard deviation.",
"In Markov decision processes (MDPs), the variance of the reward-to-go is a natural measure of uncertainty about the long term performance of a policy, and is important in domains such as finance, resource allocation, and process control. Currently however, there is no tractable procedure for calculating it in large scale MDPs. This is in contrast to the case of the expected reward-to-go, also known as the value function, for which effective simulation-based algorithms are known, and have been used successfully in various domains. In this paper we extend temporal difference (TD) learning algorithms to estimating the variance of the reward-to-go for a fixed policy. We propose variants of both TD(0) and LSTD(λ) with linear function approximation, prove their convergence, and demonstrate their utility in an option pricing problem. Our results show a dramatic improvement in terms of sample efficiency over standard Monte-Carlo methods, which are currently the state-of-the-art.",
"Standard Reinforcement Learning (RL) aims to optimize decision-making rules in terms of the expected return. However, especially for risk-management purposes, other criteria such as the expected shortfall are sometimes preferred. Here, we describe a method of approximating the distribution of returns, which allows us to derive various kinds of information about the returns. We first show that the Bellman equation, which is a recursive formula for the expected return, can be extended to the cumulative return distribution. Then we derive a nonparametric return distribution estimator with particle smoothing based on this extended Bellman equation. A key aspect of the proposed algorithm is to represent the recursion relation in the extended Bellman equation by a simple replacement procedure of particles associated with a state by using those of the successor state. We show that our algorithm leads to a risk-sensitive RL paradigm. The usefulness of the proposed approach is demonstrated through numerical experiments.",
"In many sequential decision-making problems we may want to manage risk by minimizing some measure of variability in rewards in addition to maximizing a standard criterion. Variance-related risk measures are among the most common risk-sensitive criteria in finance and operations research. However, optimizing many such criteria is known to be a hard problem. In this paper, we consider both discounted and average reward Markov decision processes. For each formulation, we first define a measure of variability for a policy, which in turn gives us a set of risk-sensitive criteria to optimize. For each of these criteria, we derive a formula for computing its gradient. We then devise actor-critic algorithms for estimating the gradient and updating the policy parameters in the ascent direction. We establish the convergence of our algorithms to locally risk-sensitive optimal policies. Finally, we demonstrate the usefulness of our algorithms in a traffic signal control application."
]
} |
1905.09768 | 2946348289 | Performing knowledge transfer from a large teacher network to a smaller student is a popular task in modern deep learning applications. However, due to growing dataset sizes and stricter privacy regulations, it is increasingly common not to have access to the data that was used to train the teacher. We propose a novel method which trains a student to match the predictions of its teacher without using any data or metadata. We achieve this by training an adversarial generator to search for images on which the student poorly matches the teacher, and then using them to train the student. Our resulting student closely approximates its teacher for simple datasets like SVHN, and on CIFAR10 we improve on the state-of-the-art for few-shot distillation (with 100 images per class), despite using no data. Finally, we also propose a metric to quantify the degree of belief matching between teacher and student in the vicinity of decision boundaries, and observe a significantly higher match between our zero-shot student and the teacher, than between a student distilled with real data and the teacher. Code available at: this https URL | The idea of using the outputs of a network to train another was first proposed by @cite_0 as a way to compress a large ensemble into a single network, and was later made popular by @cite_5 and then @cite_9 , who proposed smoothing the teacher's probability outputs. Since then, the focus has mostly been on improving distillation efficiency by designing better students , or getting small performance gains with extra loss terms, such as attention transfer (AT) . Since the term (KD) has become intertwined with the loss function introduced by @cite_9 , we refer to our task more generally as zero-shot knowledge transfer (KT) for clarity. | {
"cite_N": [
"@cite_0",
"@cite_5",
"@cite_9"
],
"mid": [
"27961112",
"2952881492",
"1821462560"
],
"abstract": [
"We introduce the problem of zero-data learning, where a model must generalize to classes or tasks for which no training data are available and only a description of the classes or tasks are provided. Zero-data learning is useful for problems where the set of classes to distinguish or tasks to solve is very large and is not entirely covered by the training data. The main contributions of this work lie in the presentation of a general formalization of zero-data learning, in an experimental analysis of its properties and in empirical evidence showing that generalization is possible and significant in this context. The experimental work of this paper addresses two classification problems of character recognition and a multitask ranking problem in the context of drug discovery. Finally, we conclude by discussing how this new framework could lead to a novel perspective on how to extend machine learning towards AI, where an agent can be given a specification for a learning problem before attempting to solve it (with very few or even zero examples).",
"Currently, deep neural networks are the state of the art on problems such as speech recognition and computer vision. In this extended abstract, we show that shallow feed-forward networks can learn the complex functions previously learned by deep nets and achieve accuracies previously only achievable with deep models. Moreover, in some cases the shallow neural nets can learn these deep functions using a total number of parameters similar to the original deep model. We evaluate our method on the TIMIT phoneme recognition task and are able to train shallow fully-connected nets that perform similarly to complex, well-engineered, deep convolutional architectures. Our success in training shallow neural nets to mimic deeper models suggests that there probably exist better algorithms for training shallow feed-forward nets than those currently available.",
"A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel."
]
} |
1905.09434 | 2946033872 | Turning is the most commonly available and least expensive machining operation, in terms of both machine-hour rates and tool insert prices. A practical CNC process planner has to maximize the utilization of turning, not only to attain precision requirements for turnable surfaces, but also to minimize the machining cost, while non-turnable features can be left for other processes such as milling. Most existing methods rely on separation of surface features and lack guarantees when analyzing complex parts with interacting features. In a previous study, we demonstrated successful implementation of a feature-free milling process planner based on configuration space methods used for spatial reasoning and AI search for planning. This paper extends the feature-free method to include turning process planning. It opens up the opportunity for seamless integration of turning actions into a mill-turn process planner that can handle arbitrarily complex shapes with or without a priori knowledge of feature semantics. | Over the past few decades, extensive research has been done on feature recognition and process planning for turning. Here we restrict ourselves to a brief overview that is by no means comprehensive. See @cite_9 @cite_20 @cite_25 @cite_11 @cite_18 @cite_26 for thorough reviews of feature-based CAPP methods. | {
"cite_N": [
"@cite_18",
"@cite_26",
"@cite_9",
"@cite_25",
"@cite_20",
"@cite_11"
],
"mid": [
"2131470105",
"2797813920",
"2034273119",
"2078980193",
"2049469357",
"2042830635"
],
"abstract": [
"For the past three decades, computer-aided process planning (CAPP) has attracted a large amount of research interest. A huge volume of literature has been published on this subject. Today, CAPP research faces new challenges owing to the dynamic markets and business globalisation. Thus, there is an urgent need to ascertain the current status and identify future trends of CAPP. Covering articles published on the subjects of CAPP in the past 10 years or so, this article aims to provide an up-to-date review of the CAPP research works, a critical analysis of journals that publish CAPP research works, and an understanding of the future direction in the field. First, general information is provided on CAPP. The past reviews are summarised. Discussions about the recent CAPP research are presented in a number of categories, i.e. feature-based technologies, knowledge-based systems, artificial neural networks, genetic algorithms, fuzzy set theory and fuzzy logic, Petri nets, agent-based technology, Internet-based technology, STEP-compliant CAPP and other emerging technologies. Research on some specific aspects of CAPP is also provided. Discussions and analysis of the methods are then presented based on the data gathered from the Elsevier's Scopus abstract and citation database. The concepts of 'Subject Strength' of a journal and 'technology impact factor' are introduced and used for discussions based on the publication data. The former is used to gauge the level of focus of a journal on a particular research subject domain, whereas the latter is used to assess the level of impact of a particular technology, in terms of citation counts. Finally, a discussion on the future development is presented.",
"The concept of smart manufacturing has become an important issue in the manufacturing industry since the start of the twenty-first century in terms of time and production cost. In addition to high production quality, a quick response could determine the success or failure of many companies and factories. One the most effective concepts for achieving a smart manufacturing industry is the use of computer-aided process planning (CAPP) techniques. Computer-aided process planning refers to key technology that connects the computer-aided design (CAD) and the computer-aided manufacturing (CAM) processes. Researchers have used many approaches as an interface between CAD and CAPP systems. In this field of research, a lot of effort has been spent to take CAPP systems to the next level in the form of automatic computer-aided process planning (ACAPP). This is to provide complete information about the product, in a way that is automated, fast, and accurate. Moreover, automatic feature recognition (AFR) techniques are considered one of the most important tasks to create an ACAPP system. This article presents a comprehensive survey about two main aspects: the degree of automation in each required input and expected output of computer-aided process planning systems as well as the benefits and the limitations of the different automatic feature recognition techniques. The aim is to demonstrate the missing aspects in smart ACAPP generation, the limitations of current systems in recognising new features, and justifying the process of selection.",
"Summary The present survey is intended to give a comprehensive picture of the actual knowledge 1n the field of computer-aided process planning systems. In a first part, a methodic analysis of the process planning function 1s carried out on the basis of a typical example. Possible areas of computer utilization are identified and problems in computer Introduction are analyzed. In a second part, a systematic review of typical existing systems is performed with an effort to classify systems according to their basic characteristics. Some examples of interesting outputs obtained with the help of computer-aided process planning are reviewed in more details.",
"Abstract Modern manufacturing is characterized by low volume, high variety production and close tolerance high quality products. Computer Integrated Manufacturing (CIM) is recognized as an effective platform for increasing manufacturing competitiveness. Computer Aided Process Planning is an essential key for achieving CIM. The integration of design, computer aided process planning (CAPP) and production planning and control (PPC) is becoming essential especially in a concurrent engineering environment where many product life cycle factors are of concern. An overview of the major development thrust in CAPP is presented along with some of the evolving trends and challenges such as rapid, generic, dynamic and or distributed process planning. Related issued of quality and evolving standards are also discussed.",
"Methods for automated process planning and NC programming are surveyed and classified into generic categories. Aspects of CAPP investigated include part definition methods, reasoning and search techniques used to generate process plans, both Al-based and traditional, and process planning languages. NC programming techniques have been categorised into those based on cell decomposition, volume delta volume decomposition, sectioning and geometric reasoning. Methods for simulation and verification of NC programs are also surveyed. The current state of the technology and the advantages and shortcomings of each of the CAPP and NC programming generic methods are also discussed.",
"Automated feature recognition (AFR) has provided the greatest contribution to fully automated CAPP system development. The objective of this paper is to review various approaches for solving three major AFR problems: (i) extraction of geometric primitives from a CAD model; (ii) defining a suitable part representation for form feature identification; and (iii) feature pattern matching recognition. A novel, detailed classification of developed AFR systems has been introduced. This paper also provides a thorough investigation of methods for geometric feature extraction, emphasizing STEP standard application and, finally, a review of recent research reports in the field of AFR with rule-based feature pattern recognition. We discuss potentials and limitations of these approaches and emphasize directions for further research work."
]
} |
1905.09418 | 2951528897 | Multi-head self-attention is a key component of the Transformer, a state-of-the-art architecture for neural machine translation. In this work we evaluate the contribution made by individual attention heads in the encoder to the overall performance of the model and analyze the roles played by them. We find that the most important and confident heads play consistent and often linguistically-interpretable roles. When pruning heads using a method based on stochastic gates and a differentiable relaxation of the L0 penalty, we observe that specialized heads are last to be pruned. Our novel pruning method removes the vast majority of heads without seriously affecting performance. For example, on the English-Russian WMT dataset, pruning 38 out of 48 encoder heads results in a drop of only 0.15 BLEU. | A line of work studying the ability of language models to capture hierarchical information @cite_22 @cite_20 was extended by and to machine translation models. | {
"cite_N": [
"@cite_22",
"@cite_20"
],
"mid": [
"2952344559",
"2963751529"
],
"abstract": [
"Model pruning seeks to induce sparsity in a deep neural network's various connection matrices, thereby reducing the number of nonzero-valued parameters in the model. Recent reports (, 2015; , 2017) prune deep networks at the cost of only a marginal loss in accuracy and achieve a sizable reduction in model size. This hints at the possibility that the baseline models in these experiments are perhaps severely over-parameterized at the outset and a viable alternative for model compression might be to simply reduce the number of hidden units while maintaining the model's dense connection structure, exposing a similar trade-off in model size and accuracy. We investigate these two distinct paths for model compression within the context of energy-efficient inference in resource-constrained environments and propose a new gradual pruning technique that is simple and straightforward to apply across a variety of models datasets with minimal tuning and can be seamlessly incorporated within the training process. We compare the accuracy of large, but pruned models (large-sparse) and their smaller, but dense (small-dense) counterparts with identical memory footprint. Across a broad range of neural network architectures (deep CNNs, stacked LSTM, and seq2seq LSTM models), we find large-sparse models to consistently outperform small-dense models and achieve up to 10x reduction in number of non-zero parameters with minimal loss in accuracy.",
"Recurrent neural networks (RNNs) have achieved impressive results in a variety of linguistic processing tasks, suggesting that they can induce non-trivial properties of language. We investigate here to what extent RNNs learn to track abstract hierarchical syntactic structure. We test whether RNNs trained with a generic language modeling objective in four languages (Italian, English, Hebrew, Russian) can predict long-distance number agreement in various constructions. We include in our evaluation nonsensical sentences where RNNs cannot rely on semantic or lexical cues (\"The colorless green ideas I ate with the chair sleep furiously\"), and, for Italian, we compare model performance to human intuitions. Our language-model-trained RNNs make reliable predictions about long-distance agreement, and do not lag much behind human performance. We thus bring support to the hypothesis that RNNs are not just shallow-pattern extractors, but they also acquire deeper grammatical competence."
]
} |
1905.09523 | 2944967640 | Many tasks such as retrieval and recommendations can significantly benefit from structuring the data, commonly in a hierarchical way. To achieve this through annotations of high dimensional data such as images or natural text can be significantly labor intensive. We propose an approach for uncovering the hierarchical structure of data based on efficient discriminative testing rather than annotations of individual datapoints. Using two-alternative-forced-choice (2AFC) testing and deep metric learning we achieve embedding of the data in semantic space where we are able to successfully hierarchically cluster. We actively select triplets for the 2AFC test such that the modeling process is highly efficient with respect to the number of tests presented to the annotator. We empirically demonstrate the feasibility of the method by confirming the shape bias on synthetic data and extract hierarchical structure on the Fashion-MNIST dataset to a finer granularity than the original labels. | Fine-grained image recognition (FGIR) tasks are also closely related to extracting hierarchical structure of image data. In @cite_8 the authors introduce a bi-linear model in order to create high-order image representations which are able to compute local pairwise interactions between features of two independent sub-networks. Such approaches have been enabled by the hierarchical representation learning present in modern convolutional neural network models @cite_34 @cite_3 . However, due to the high dimensionality of the features it becomes impractical for subsequent analysis. In order to reduce the high dimensionality of bilinear model features, @cite_22 introduced a model that approximates such bilinear feature by using polynomial kernels. @cite_12 went a step further and introduced a classifier co-decomposition to further restrict a bilinear model. | {
"cite_N": [
"@cite_22",
"@cite_8",
"@cite_3",
"@cite_34",
"@cite_12"
],
"mid": [
"2963066927",
"2104657103",
"",
"2270657321",
"2554320282"
],
"abstract": [
"Bilinear models has been shown to achieve impressive performance on a wide range of visual tasks, such as semantic segmentation, fine grained recognition and face recognition. However, bilinear features are high dimensional, typically on the order of hundreds of thousands to a few million, which makes them impractical for subsequent analysis. We propose two compact bilinear representations with the same discriminative power as the full bilinear representation but with only a few thousand dimensions. Our compact representations allow back-propagation of classification errors enabling an end-to-end optimization of the visual recognition system. The compact bilinear representations are derived through a novel kernelized analysis of bilinear pooling which provide insights into the discriminative power of bilinear pooling, and a platform for further research in compact pooling methods. Experimentation illustrate the utility of the proposed representations for image classification and few-shot learning across several datasets.",
"We propose bilinear models, a recognition architecture that consists of two feature extractors whose outputs are multiplied using outer product at each location of the image and pooled to obtain an image descriptor. This architecture can model local pairwise feature interactions in a translationally invariant manner which is particularly useful for fine-grained categorization. It also generalizes various orderless texture descriptors such as the Fisher vector, VLAD and O2P. We present experiments with bilinear models where the feature extractors are based on convolutional neural networks. The bilinear form simplifies gradient computation and allows end-to-end training of both networks using image labels only. Using networks initialized from the ImageNet dataset followed by domain specific fine-tuning we obtain 84.1 accuracy of the CUB-200-2011 dataset requiring only category labels at training time. We present experiments and visualizations that analyze the effects of fine-tuning and the choice two networks on the speed and accuracy of the models. Results show that the architecture compares favorably to the existing state of the art on a number of fine-grained datasets while being substantially simpler and easier to train. Moreover, our most accurate model is fairly efficient running at 8 frames sec on a NVIDIA Tesla K40 GPU. The source code for the complete system will be made available at http: vis-www.cs.umass.edu bcnn.",
"",
"Salient object detection increasingly receives attention as an important component or step in several pattern recognition and image processing tasks. Although a variety of powerful saliency models have been intensively proposed, they usually involve heavy feature (or model) engineering based on priors (or assumptions) about the properties of objects and backgrounds. Inspired by the effectiveness of recently developed feature learning, we provide a novel deep image saliency computing (DISC) framework for fine-grained image saliency computing. In particular, we model the image saliency from both the coarse-and fine-level observations, and utilize the deep convolutional neural network (CNN) to learn the saliency representation in a progressive manner. In particular, our saliency model is built upon two stacked CNNs. The first CNN generates a coarse-level saliency map by taking the overall image as the input, roughly identifying saliency regions in the global context. Furthermore, we integrate superpixel-based local context information in the first CNN to refine the coarse-level saliency map. Guided by the coarse saliency map, the second CNN focuses on the local context to produce fine-grained and accurate saliency map while preserving object details. For a testing image, the two CNNs collaboratively conduct the saliency computing in one shot. Our DISC framework is capable of uniformly highlighting the objects of interest from complex background while preserving well object details. Extensive experiments on several standard benchmarks suggest that DISC outperforms other state-of-the-art methods and it also generalizes well across data sets without additional training. The executable version of DISC is available online: http: vision.sysu.edu.cn projects DISC .",
"Pooling second-order local feature statistics to form a high-dimensional bilinear feature has been shown to achieve state-of-the-art performance on a variety of fine-grained classification tasks. To address the computational demands of high feature dimensionality, we propose to represent the covariance features as a matrix and apply a low-rank bilinear classifier. The resulting classifier can be evaluated without explicitly computing the bilinear feature map which allows for a large reduction in the compute time as well as decreasing the effective number of parameters to be learned. To further compress the model, we propose a classifier co-decomposition that factorizes the collection of bilinear classifiers into a common factor and compact per-class terms. The co-decomposition idea can be deployed through two convolutional layers and trained in an end-to-end architecture. We suggest a simple yet effective initialization that avoids explicitly first training and factorizing the larger bilinear classifiers. Through extensive experiments, we show that our model achieves state-of-the-art performance on several public datasets for fine-grained classification trained with only category labels. Importantly, our final model is an order of magnitude smaller than the recently proposed compact bilinear model [8], and three orders smaller than the standard bilinear CNN model [19]."
]
} |
1905.09068 | 2944876351 | Supervised machine learning applications in the health domain often face the problem of insufficient training datasets. The quantity of labelled data is small due to privacy concerns and the cost of data acquisition and labelling by a medical expert. Furthermore, it is quite common that collected data are unbalanced and getting enough data to personalize models for individuals is very expensive or even infeasible. This paper addresses these problems by (1) designing a recurrent Generative Adversarial Network to generate realistic synthetic data and to augment the original dataset, (2) enabling the generation of balanced datasets based on heavily unbalanced dataset, and (3) to control the data generation in such a way that the generated data resembles data from specific individuals. We apply these solutions for sleep apnea detection and study in the evaluation the performance of four well-known techniques, i.e., K-Nearest Neighbour, Random Forest, Multi-Layer Perceptron, and Support Vector Machine. All classifiers exhibit in the experiments a consistent increase in sensitivity and a kappa statistic increase by between 0.007 and 0.182. | We note that most of the related work with the exception of @cite_10 focuses individually on the synthetic data generation and evaluation, and not how to use these data to augment the original dataset to potentially improve the generalization capability of other classifiers. To the best of our knowledge only few works @cite_26 @cite_25 @cite_24 exist that examine the potential application of GANs to produce realistic synthetic data for class rebalancing of a training dataset. Only one of them uses specifically a recurrent GAN architecture. Finally, we did not find any relevant work that depicts the data distribution as a mixture of different recording distributions, with the end-goal of producing more personalized synthetic data. | {
"cite_N": [
"@cite_24",
"@cite_26",
"@cite_10",
"@cite_25"
],
"mid": [
"2795144650",
"2756182389",
"2963034797",
"2911583416"
],
"abstract": [
"Image classification datasets are often imbalanced, characteristic that negatively affects the accuracy of deep-learning classifiers. In this work we propose balancing GAN (BAGAN) as an augmentation tool to restore balance in imbalanced datasets. This is challenging because the few minority-class images may not be enough to train a GAN. We overcome this issue by including during the adversarial training all available images of majority and minority classes. The generative model learns useful features from majority classes and uses these to generate images for minority classes. We apply class conditioning in the latent space to drive the generation process towards a target class. The generator in the GAN is initialized with the encoder module of an autoencoder that enables us to learn an accurate class-conditioning in the latent space. We compare the proposed methodology with state-of-the-art GANs and demonstrate that BAGAN generates images of superior quality when trained with an imbalanced dataset.",
"Application of conditional Generative Adversarial Networks as oversampling method.Generates minority class samples by recovering the training data distribution.Outperforms various standard oversampling algorithms.Performance advantage of the proposed method remains stable with higher imbalance ratios. Learning from imbalanced datasets is a frequent but challenging task for standard classification algorithms. Although there are different strategies to address this problem, methods that generate artificial data for the minority class constitute a more general approach compared to algorithmic modifications. Standard oversampling methods are variations of the SMOTE algorithm, which generates synthetic samples along the line segment that joins minority class samples. Therefore, these approaches are based on local information, rather on the overall minority class distribution. Contrary to these algorithms, in this paper the conditional version of Generative Adversarial Networks (cGAN) is used to approximate the true data distribution and generate data for the minority class of various imbalanced datasets. The performance of cGAN is compared against multiple standard oversampling algorithms. We present empirical results that show a significant improvement in the quality of the generated data when cGAN is used as an oversampling algorithm.",
"The rapid growth of Electronic Health Records (EHRs), as well as the accompanied opportunities in Data-Driven Healthcare (DDH), has been attracting widespread interests and attentions. Recent progress in the design and applications of deep learning methods has shown promising results and is forcing massive changes in healthcare academia and industry, but most of these methods rely on massive labeled data. In this work, we propose a general deep learning framework which is able to boost risk prediction performance with limited EHR data. Our model takes a modified generative adversarial network namely ehrGAN, which can provide plausible labeled EHR data by mimicking real patient records, to augment the training dataset in a semi-supervised learning manner. We use this generative model together with a convolutional neural network (CNN) based prediction model to improve the onset prediction performance. Experiments on two real healthcare datasets demonstrate that our proposed framework produces realistic data samples and achieves significant improvements on classification tasks with the generated data over several stat-of-the-art baselines.",
"We propose a new recurrent generative adversarial architecture named RNN-GAN to mitigate imbalance data problem in medical image semantic segmentation where the number of pixels belongs to the desired object are significantly lower than those belonging to the background. A model trained with imbalanced data tends to bias towards healthy data which is not desired in clinical applications and predicted outputs by these networks have high precision and low recall. To mitigate imbalanced training data impact, we train RNN-GAN with proposed complementary segmentation mask, in addition, ordinary segmentation masks. The RNN-GAN consists of two components: a generator and a discriminator. The generator is trained on the sequence of medical images to learn corresponding segmentation label map plus proposed complementary label both at a pixel level, while the discriminator is trained to distinguish a segmentation image coming from the ground truth or from the generator network. Both generator and discriminator substituted with bidirectional LSTM units to enhance temporal consistency and get inter and intra-slice representation of the features. We show evidence that the proposed framework is applicable to different types of medical images of varied sizes. In our experiments on ACDC-2017, HVSMR-2016, and LiTS-2017 benchmarks we find consistently improved results, demonstrating the efficacy of our approach."
]
} |
1905.09355 | 2951214791 | Reduced models of large Markov decision processes accelerate planning by considering a subset of outcomes for each state-action pair. This reduction in reachable states leads to replanning when the agent encounters states without a precomputed action during plan execution. However, not all states are suitable for replanning. In the worst case, the agent may not be able to reach the goal from the newly encountered state. Agents should be better prepared to handle such risky situations and avoid replanning in risky states. Hence, we consider replanning in states that are unsafe for deliberation as a negative side effect of planning with reduced models. While the negative side effects can be minimized by always using the full model, this defeats the purpose of using reduced models. The challenge is to plan with reduced models, but somehow account for the possibility of encountering risky situations. An agent should thus only replan in states that the user has approved as safe for replanning. To that end, we propose planning using a portfolio of reduced models, a planning paradigm that minimizes the negative side effects of planning using reduced models by alternating between different outcome selection approaches. We empirically demonstrate the effectiveness of our approach on three domains: an electric vehicle charging domain using real-world data from a university campus and two benchmark planning problems. | Interest in using reduced models to accelerate planning increased after the success of FF-Replan @cite_25 , which won the 2004 IPPC using the Fast Forward (FF) technique to generate deterministic plans @cite_14 . Following the success of FF-Replan, researchers have proposed various methods to improve determinization @cite_18 @cite_24 @cite_19 . Robust FF (RFF) generates a plan for an envelope of states such that the probability of reaching a state outside the envelope is below some predefined threshold @cite_3 and replans to the subset of states for which the policy is already computed. Our work differs from that of RFF since we compute reachability as a pre-processing to formulate and not the policy. | {
"cite_N": [
"@cite_18",
"@cite_14",
"@cite_3",
"@cite_24",
"@cite_19",
"@cite_25"
],
"mid": [
"2103158016",
"2097083865",
"1603989336",
"2212912172",
"2401856889",
"127341816"
],
"abstract": [
"Recently, 'determinization in hindsight' has enjoyed surprising success in on-line probabilistic planning. This technique evaluates the actions available in the current state by using non-probabilistic planning in deterministic approximations of the original domain. Although the approach has proven itself effective in many challenging domains, it is computationally very expensive. In this paper, we present three significant improvements to help mitigate this expense. First, we use a method for detecting potentially useful actions, allowing us to avoid estimating the values of unnecessary ones. Second, we exploit determinism in the domain by reusing relevant plans rather than computing new ones. Third, we improve action evaluation by increasing the chance that at least one deterministic plan reaches a goal. Taken together, these improvements allow determinization in hindsight to scale significantly better on large or mostly-deterministic problems.",
"Fast-forward (FF) was the most successful automatic planner in the Fifth International Conference on Artificial Intelligence Planning and Scheduling (AIPS '00) planning systems competition. Like the well-known hsp system, FF relies on forward search in the state space, guided by a heuristic that estimates goal distances by ignoring delete lists. It differs from HSP in a number of important details. This article describes the algorithmic techniques used in FF in comparison to hsp and evaluates their benefits in terms of run-time and solution-length behavior.",
"Despite the recent advances in planning with MDPs, the problem of generating good policies is still hard. This paper describes a way to generate policies in MDPs by (1) determinizing the given MDP model into a classical planning problem; (2) building partial policies off-line by producing solution plans to the classical planning problem and incrementally aggregating them into a policy, and (3) using sequential Monte-Carlo (MC) simulations of the partial policies before execution, in order to assess the probability of replanning for a policy during execution. The objective of this approach is to quickly generate policies whose probability of replanning is low and below a given threshold. We describe our planner RFF, which incorporates the above ideas. We present theorems showing the termination, soundness and completeness properties of RFF. RFF was the winner of the fully-observable probabilistic track in the 2008 International Planning Competition (IPC-08). In addition to our analyses of the IPC-08 results, we analyzed RFF's performance with different plan aggregation and determinization strategies, with varying amount of MC sampling, and with varying threshold values for probability of replanning. The results of these experiments revealed how they impact the time performance of RFF to generate solution policies and the quality of those solution policies (i.e., the average accumulated reward gathered from the execution of the policies).",
"Most predominant approaches in probabilistic planning utilize techniques from the more thoroughly investigated field of classical planning by determinizing the problem at hand. In this paper, we present a method to map probabilistic operators to an equivalent set of probabilistic operators in a novel normal form, requiring polynomial time and space. From this, we directly derive a determinization which can be used for, e. g., replanning strategies incorporating a classical planning system. Unlike previously described all outcome determinizations, the number of deterministic operators is not exponentially but polynomially bounded in the number of parallel probabilistic effects, enabling the use of more sophisticated determinization-based techniques in the future.",
"Inspired by the success of the satisfiability approach for deterministic planning, we propose a novel framework for on-line stochastic planning, by embedding the idea of hindsight optimization into a reduction to integer linear programming. In contrast to the previous work using reductions or hindsight optimization, our formulation is general purpose by working with domain specifications over factored state and action spaces, and by doing so is also scalable in principle to exponentially large action spaces. Our approach is competitive with state-of-the-art stochastic planners on challenging benchmark problems, and sometimes exceeds their performance especially in large action spaces.",
"FF-Replan was the winner of the 2004 International Probabilistic Planning Competition (IPPC-04) (Younes & Littman 2004a) and was also the top performer on IPPC-06 domains, though it was not an official entry. This success was quite surprising, due to the simplicity of the approach. In particular, FF-Replan calls FF on a carefully constructed deterministic variant of the planning problem and selects actions according to the plan until observing an unexpected effect, upon which it replans. Despite the obvious shortcomings of the approach and its strawman nature, it is the state-of-the-art in probabilistic planning as measured on recent competition benchmarks. This paper gives the first technical description of FF-Replan and provides an analysis of its results on all of the recent IPPC-04 and IPPC-06 domains. We hope that this will inspire extensions and insight into the approach and planning domains themselves that will soon lead to the dethroning of FF-Replan."
]
} |
1905.09173 | 2945683540 | The one-class kernel spectral regression (OC-KSR), the regression-based formulation of the kernel null-space approach has been found to be an effective Fisher criterion-based methodology for one-class classification (OCC), achieving state-of-the-art performance in one-class classification while providing relatively high robustness against data corruption. This work extends the OC-KSR methodology to a multi-task setting where multiple one-class problems share information for improved performance. By viewing the multi-task structure learning problem as one of compositional function learning, first, the OC-KSR method is extended to learn multiple tasks' structure by posing it as an instantiation of the separable kernel learning problem in a vector-valued reproducing kernel Hilbert space where an output kernel encodes tasks' structure while another kernel captures input similarities. Next, a non-linear structure learning mechanism is proposed which captures multiple tasks' relationships via an output kernel. The non-linear structure learning method is then extended to a sparse setting where different tasks compete in an output composition mechanism, leading to a sparse non-linear structure among multiple problems. Through extensive experiments on different data sets, the merits of the proposed multi-task kernel null-space techniques are verified against the baseline as well as other existing multi-task one-class learning techniques. | In this section, a brief overview of the existing multi-task one-class learning approaches is presented. A number of non-OCC multi-target regression methods, related to the present work, shall be briefly reviewed too. For a detailed review on the general concept of multi-task learning the reader is referred to @cite_4 . | {
"cite_N": [
"@cite_4"
],
"mid": [
"2742079690"
],
"abstract": [
"Multi-Task Learning (MTL) is a learning paradigm in machine learning and its aim is to leverage useful information contained in multiple related tasks to help improve the generalization performance of all the tasks. In this paper, we give a survey for MTL. First, we classify different MTL algorithms into several categories, including feature learning approach, low-rank approach, task clustering approach, task relation learning approach, and decomposition approach, and then discuss the characteristics of each approach. In order to improve the performance of learning tasks further, MTL can be combined with other learning paradigms including semi-supervised learning, active learning, unsupervised learning, reinforcement learning, multi-view learning and graphical models. When the number of tasks is large or the data dimensionality is high, batch MTL models are difficult to handle this situation and online, parallel and distributed MTL models as well as dimensionality reduction and feature hashing are reviewed to reveal their computational and storage advantages. Many real-world applications use MTL to boost their performance and we review representative works. Finally, we present theoretical analyses and discuss several future directions for MTL."
]
} |
1905.09347 | 2946390729 | We study how to maximize the broker's (expected) profit in a two-sided market, where she buys items from a set of sellers and resells them to a set of buyers. Each seller has a single item to sell and holds a private value on her item, and each buyer has a valuation function over the bundles of the sellers' items. We consider the Bayesian setting where the agents' values are independently drawn from prior distributions, and aim at designing dominant-strategy incentive-compatible (DSIC) mechanisms that are approximately optimal. Production-cost markets, where each item has a publicly-known cost to be produced, provide a platform for us to study two-sided markets. Briefly, we show how to covert a mechanism for production-cost markets into a mechanism for the broker, whenever the former satisfies cost-monotonicity. This reduction holds even when buyers have general combinatorial valuation functions. When the buyers' valuations are additive, we generalize an existing mechanism to production-cost markets in an approximation-preserving way. We then show that the resulting mechanism is cost-monotone and thus can be converted into an 8-approximation mechanism for two-sided markets. | Two-sided markets are also called double auctions @cite_0 , bilateral trading @cite_5 or market intermediation @cite_6 in the literature. Maximizing the broker's profit is an important objective for two-sided market. The seminal paper @cite_5 characterized the optimal mechanism for one seller and one buyer, which is further generalized by @cite_22 to multiple single-parameter sellers and buyers. Unlike our work, @cite_22 studies the Bayesian Incentive Compatible (BIC) mechanisms. DSIC mechanisms are also studied in the literature, but only for some special cases: @cite_6 studies the case of a single buyer and multiple sellers, @cite_3 studies the case of a single seller and multiple buyers, and @cite_11 studies the optimal mechanism when the numbers of sellers and buyers are both constants. Although @cite_20 studies two-sided markets with multiple buyers and multiple sellers, the dealer there has a fixed budget and their mechanism guarantees that the payment to sellers is within the budget. Before our work, it remained unknown how to design a (simple) DSIC mechanism that approximates the optimal profit in multi-parameter settings with a general number of sellers and buyers. | {
"cite_N": [
"@cite_22",
"@cite_3",
"@cite_6",
"@cite_0",
"@cite_5",
"@cite_20",
"@cite_11"
],
"mid": [
"2028289369",
"2898469734",
"1635784337",
"2138848780",
"2087530244",
"2406185007",
"2516752364"
],
"abstract": [
"We study double auction market design where the market maker wants to maximize its total revenue by buying low from the sell- ers and selling high to the buyers. We consider a Bayesian setting where buyers and sellers have independent probability distributions on the val- ues of products on the market. For the simplest setting where each seller has one kind of item that can be sold in whole to a buyer, and each buyer's value can be represented by a single parameter, i.e., single-parameter setting, we develop a maximum mechanism for the market maker to maximize its own revenue. For the more general case where the product may be different, we consider various models in terms of supplies and demands constraints. For each of them, we develop a polynomial time computable truthful mechanism for the market maker to achieve a revenue at least a constant α times the revenue of any other truthful mechanism. We consider a double auction market maker who collects valuations from buy- ers and sellers about a certain product to decide on the prices each seller gets and each buyer pays. The buyers may want to buy many units and the sellers may have many units to part with. The buyers and sellers may have different valuations of the product, and there is public knowledge of the probability dis- tributions of the valuations (but each valuation, sampled from its distribution, is known only to its own buyer or seller). For simplicity, we assume that the probability distributions are independent. For the sellers and buyers, they know their own private values exactly. The market maker purchases the products from the sellers and sell them to the buyers. Our goal is to design a market mechanism that maximizes the revenue of the market maker. In other words, the market maker is to buy the same amount of products from the sellers as the amount sold to the buyers with the objective of maximizing the difference of its collected payment from the buyers and the total amount paid to the sellers. When in addi- tion we assume public knowledge of distributions of buyers' private values from the previous sales, we call it a revenue maximization Bayesian double auction market maker. � Supported by EPSRC grant EP G069239 1 \"Efficient Decentralised Approaches in",
"",
"We study the optimal mechanism design problem faced by a market intermediary who makes revenue by connecting buyers and sellers. We first show that the optimal intermediation protocol has substantial structure: it is the solution to an algorithmic pricing problem in which seller’s costs are replaced with virtual costs, and the sellers’ payments need only depend on the buyer’s behavior and not the buyer’s actual valuation function. Since the underlying algorithmic pricing problem may be difficult to solve o we study specific models of buyer behavior and give mechanisms with provable approximation guarantees. We show that offering only the single most profitable item for sale guarantees an ( 1 log n ) fraction of the optimal revenue when item value distributions are independent and have monotone hazard rates. We also give constant factor approximations when the buyer considers all items at once, k items at once, or items in sequence.",
"A double auction mechanism that provides dominant strategies for both buyers and sellers is analyzed. This mechanism satisfies the 1n convergence to efficiency of the buyer's bid double auction. In addition, the mechanism always produces full information first best prices; the inefficiency arises because the least valuable profitable trade may be prohibited by the mechanism. The mechanism has an oral implementation utilizing bid and asked prices.",
"We consider bargaining problems between one buyer and one seller for a single object. The seller’s valuation and the buyer’s valuation for the object are assumed to be independent random variables, and each individual’s valuation is unknown to the other. We characterize the set of allocation mechanisms that are Bayesian incentive compatible and individually rational, and show the general impossibility of ex post efficient mechanisms without outside subsidies. For a wide class of problems we show how to compute mechanisms that maximize expected total gains from trade, and mechanisms that can maximize a broker’s expected profit. Journal of Economic Literature Classification Number: 026.",
"We consider the problem of designing budget feasible mechanisms for a dealer, who aims to maximize revenue by buying items from a seller market and selling them to a buyer market that consists of unit-demand buyers. Different from the related literature, the dealer's value'' for a set of items that he purchased from the seller market is not directly given as a number but it is defined to be the maximum revenue the dealer can obtain from selling the items to the buyers. We aim to design mechanisms that are dominant-strategy truthful for the sellers to report their costs and envy-free for the buyers to purchase their most preferred items (given their prices) in the final outcome, such that the total payment to the sellers does not exceed the dealer's budget and the dealer's revenue is (approximately) maximized. First, to understand the structure of the optimal mechanisms, we show that the maximum (envy-free) revenue obtainable by the dealer as a function of the set of purchased items is monotone and subadditive. Thus, existing results on subadditive optimization problems are potentially applicable in solving the mechanism design problem for the dealer. However, a crucial assumption adopted by all previous studies on subadditive functions is that the mechanism or algorithm has access to the value oracle and or the demand oracle. In the dealer's problem, instead, we show that (1) the demand oracle can be efficiently simulated by the value oracle and (2) both have efficient O(log n)-approximation algorithms, where n is the number of buyers. This is particularly interesting given the literature, since, in general, the demand oracle can always efficiently simulate the value oracle, and there are cases where the demand oracle is strictly more powerful. Our results show that, for the dealer's problem, the two oracles are as powerful as each other. Finally, using the results above, we provide a polynomial-time budget feasible mechanism for the dealer that doesn't use any oracle and provides an O((log2n)(log2m))-approximation of the optimal revenue, where m is the number of sellers.",
"We study the computational challenge faced by a intermediary who attempts to profit from trade with a small number of buyers and sellers of some item. In the version of the problem that we study, the number of buyers and sellers is constant, but their joint distribution of the item’s value may be complicated. We consider discretized distributions, where the complexity parameter is the support size, or number of different prices that may occur. We show that maximizing the expected revenue is computationally tractable (via an LP) if we are allowed to use randomized mechanisms. For the deterministic case, we show how an optimal mechanism can be efficiently computed for the one-seller one-buyer case, but give a contrasting NP-completeness result for the one-seller two-buyer case."
]
} |
1905.09347 | 2946390729 | We study how to maximize the broker's (expected) profit in a two-sided market, where she buys items from a set of sellers and resells them to a set of buyers. Each seller has a single item to sell and holds a private value on her item, and each buyer has a valuation function over the bundles of the sellers' items. We consider the Bayesian setting where the agents' values are independently drawn from prior distributions, and aim at designing dominant-strategy incentive-compatible (DSIC) mechanisms that are approximately optimal. Production-cost markets, where each item has a publicly-known cost to be produced, provide a platform for us to study two-sided markets. Briefly, we show how to covert a mechanism for production-cost markets into a mechanism for the broker, whenever the former satisfies cost-monotonicity. This reduction holds even when buyers have general combinatorial valuation functions. When the buyers' valuations are additive, we generalize an existing mechanism to production-cost markets in an approximation-preserving way. We then show that the resulting mechanism is cost-monotone and thus can be converted into an 8-approximation mechanism for two-sided markets. | Finally, we briefly discuss the efficiency of two-sided markets, which is measured by gain-from-trade (GFT), i.e., the total value gained by the buyers minus the value contributed by the sellers. @cite_0 gave the first approximation mechanism for the one seller and one buyer case, and @cite_15 gives approximation mechanisms for multiple buyers with unit demand valuations. Recently, @cite_17 and @cite_18 study the asymptotically efficient mechanisms instead of constant approximations. For maximizing social welfare , @cite_4 @cite_7 provide constant-approximation mechanisms. | {
"cite_N": [
"@cite_18",
"@cite_4",
"@cite_7",
"@cite_0",
"@cite_15",
"@cite_17"
],
"mid": [
"2962680143",
"2262059501",
"2553461381",
"2138848780",
"2627057954",
"2962728805"
],
"abstract": [
"",
"Mechanism design for one-sided markets is an area of extensive research in economics and, since more than a decade, in computer science as well. Two-sided markets, on the other hand, have not received the same attention despite the numerous applications to web advertisement, stock exchange, and frequency spectrum allocation. This work studies double auctions, in which unit-demand buyers and unit-supply sellers act strategically. An ideal goal in double auction design is to maximize the social welfare of buyers and sellers with individually rational (IR), incentive compatible (IC) and strongly budget-balanced (SBB) mechanisms. The first two properties are standard. SBB requires that the payments charged to the buyers are entirely handed to the sellers. This property is crucial in all the contexts that do not allow the auctioneer retaining a share of buyers' payments or subsidizing the market. Unfortunately, this goal is known to be unachievable even for the special case of bilateral trade, where there is only one buyer and one seller. Therefore, in subsequent papers, meaningful trade-offs between these requirements have been investigated. Our main contribution is the first IR, IC and SBB mechanism that provides an O(1)-approximation to the optimal social welfare. This result holds for any number of buyers and sellers with arbitrary, independent distributions. Moreover, our result continues to hold when there is an additional matroid constraint on the sets of buyers who may get allocated an item. To prove our main result, we devise an extension of sequential posted price mechanisms to two-sided markets. In addition to this, we improve the best-known approximation bounds for the bilateral trade problem.",
"We develop and extend a line of recent work on the design of mechanisms for two-sided markets. The markets we consider consist of buyers and sellers of a number of items, and the aim of a mechanism is to improve the social welfare by arranging purchases and sales of the items. A mechanism is given prior distributions on the agents' valuations of the items, but not the actual valuations; thus the aim is to maximise the expected social welfare over these distributions. As in previous work, we are interested in the worst-case ratio between the social welfare achieved by a truthful mechanism, and the best social welfare possible. Our main result is an incentive compatible and budget balanced constant-factor approximation mechanism in a setting where buyers have XOS valuations and sellers' valuations are additive. This is the first such approximation mechanism for a two-sided market setting where the agents have combinatorial valuation functions. To achieve this result, we introduce a more general kind of demand query that seems to be needed in this situation. In the simpler case that sellers have unit supply (each having just one item to sell), we give a new mechanism whose welfare guarantee improves on a recent one in the literature. We also introduce a more demanding version of the strong budget balance (SBB) criterion, aimed at ruling out certain \"unnatural\" transactions satisfied by SBB. We show that the stronger version is satisfied by our mechanisms.",
"A double auction mechanism that provides dominant strategies for both buyers and sellers is analyzed. This mechanism satisfies the 1n convergence to efficiency of the buyer's bid double auction. In addition, the mechanism always produces full information first best prices; the inefficiency arises because the least valuable profitable trade may be prohibited by the mechanism. The mechanism has an oral implementation utilizing bid and asked prices.",
"We design simple mechanisms to approximate the Gains from Trade (GFT) in two-sided markets with multiple unit-supply sellers and multiple unit-demand buyers. A classical impossibility result by Myerson and Satterthwaite showed that even with only one seller and one buyer, no Bayesian Incentive Compatible (BIC), Individually Rational (IR), and Budget-Balanced (BB) mechanism can achieve full GFT (trade whenever buyer's value is higher than the seller's cost). The same paper also proposed the second-best'' mechanism that maximizes the GFT subject to BIC, IR, and BB constraints, which is unfortunately rather complex for even the single-seller single-buyer case. Our mechanism is simple, BIC, IR, and BB and achieves 1 2 of the optimal GFT among all BIC, IR, and BB mechanisms. The result holds for arbitrary distributions of the buyers' and sellers' values and can accommodate any downward-closed feasibility constraints over the allocations. The analysis of our mechanism is facilitated by extending the Cai-Weinberg-Devanur duality framework to two-sided markets.",
""
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | A straightforward way to simulate graphs is to generate them using well-established network models @cite_12 : (1) Barab ' a si-Albert model for the scale-free network @cite_8 , (2) Watts-Strogatz small-world model for the small-world network @cite_34 , (3) Erd o s-R ' e nyi model for the random graph network @cite_38 @cite_3 @cite_41 @cite_33 , (4) Forest-fire Model model @cite_13 @cite_7 . | {
"cite_N": [
"@cite_38",
"@cite_33",
"@cite_7",
"@cite_8",
"@cite_41",
"@cite_3",
"@cite_34",
"@cite_13",
"@cite_12"
],
"mid": [
"1993744672",
"2146162554",
"",
"2008620264",
"2908457301",
"",
"2112090702",
"2111708605",
"2964139317"
],
"abstract": [
"The weak connectivity γ of a random net is defined and computed by an approximation method as a function ofa, the axone density. It is shown that γ rises rapidly witha, attaining 0.8 of its asymptotic value (unity) fora=2, where the number of neurons in the net is arbitrarily large. The significance of this parameter is interpreted also in terms of the maximum expected spread of an epidemic under certain conditions.",
"If G is an arbitrary non-complete graph, let cZ1(G) denote the least number k such that by deleting k appropriately chosen vertices from G (i. e. deleting the k points in question and all edges starting from these points) the resulting graph is not connected. If G is a complete graph of order n, we put c,(G) = n-l. Let cY(G) denote the least number 1 such that by deleting 1 appropriately chosen edges from G the resulting graph is not connected. We may measure the strength of connectedness of G by any of the numbers c],(G), c$,(G) and in a certain sense if G is known to be connected) also by c(G). Evidently one has",
"",
"Systems as diverse as genetic networks or the World Wide Web are best described as networks with complex topology. A common property of many large networks is that the vertex connectivities follow a scale-free power-law distribution. This feature was found to be a consequence of two generic mechanisms: (i) networks expand continuously by the addition of new vertices, and (ii) new vertices attach preferentially to sites that are already well connected. A model based on these two ingredients reproduces the observed stationary scale-free distributions, which indicates that the development of large networks is governed by robust self-organizing phenomena that go beyond the particulars of the individual systems.",
"",
"",
"Networks of coupled dynamical systems have been used to model biological oscillators1,2,3,4, Josephson junction arrays5,6, excitable media7, neural networks8,9,10, spatial games11, genetic control networks12 and many other self-organizing systems. Ordinarily, the connection topology is assumed to be either completely regular or completely random. But many biological, technological and social networks lie somewhere between these two extremes. Here we explore simple models of networks that can be tuned through this middle ground: regular networks ‘rewired’ to introduce increasing amounts of disorder. We find that these systems can be highly clustered, like regular lattices, yet have small characteristic path lengths, like random graphs. We call them ‘small-world’ networks, by analogy with the small-world phenomenon13,14 (popularly known as six degrees of separation15). The neural network of the worm Caenorhabditis elegans, the power grid of the western United States, and the collaboration graph of film actors are shown to be small-world networks. Models of dynamical systems with small-world coupling display enhanced signal-propagation speed, computational power, and synchronizability. In particular, infectious diseases spread more easily in small-world networks than in regular lattices.",
"How do real graphs evolve over time? What are \"normal\" growth patterns in social, technological, and information networks? Many studies have discovered patterns in static graphs, identifying properties in a single snapshot of a large network, or in a very small number of snapshots; these include heavy tails for in- and out-degree distributions, communities, small-world phenomena, and others. However, given the lack of information about network evolution over long periods, it has been hard to convert these findings into statements about trends over time.Here we study a wide range of real graphs, and we observe some surprising phenomena. First, most of these graphs densify over time, with the number of edges growing super-linearly in the number of nodes. Second, the average distance between nodes often shrinks over time, in contrast to the conventional wisdom that such distance parameters should increase slowly as a function of the number of nodes (like O(log n) or O(log(log n)).Existing graph generation models do not exhibit these types of behavior, even at a qualitative level. We provide a new graph generator, based on a \"forest fire\" spreading process, that has a simple, intuitive justification, requires very few parameters (like the \"flammability\" of nodes), and produces graphs exhibiting the full range of properties observed both in prior work and in the present study.",
"Abstract Networks are everywhere and their many types, including social networks, the Internet, food webs etc., have been studied for the last few decades. However, in real-world networks, it’s hard to find examples that can be easily comparable, i.e. have the same density or even number of nodes and edges. We propose a flexible and extensible NetSim framework to understand how properties in different types of networks change with varying number of edges and vertices. Our approach enables to simulate three classical network models (random, small-world and scale-free) with easily adjustable model parameters and network size. To be able to compare different networks, for a single experimental setup we kept the number of edges and vertices fixed across the models. To understand how they change depending on the number of nodes and edges we ran over 30,000 simulations and analysed different network characteristics that cannot be derived analytically. Two of the main findings from the analysis are that the average shortest path does not change with the density of the scale-free network but changes for small-world and random networks; the apparent difference in mean betweenness centrality of the scale-free network compared with random and small-world networks."
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | The scale-free model shows power law node degree distribution @math (where @math is the node degree and typically @math ) for a social network. This kind of distribution was first discussed by . Price, in turn, was inspired by Herbert Simon, who discusses power law in a variety of non-network economic data @cite_16 . | {
"cite_N": [
"@cite_16"
],
"mid": [
"2036671379"
],
"abstract": [
"It is the purpose of this paper to analyse a class of distribution functions that appears in a wide range of empirical data-particularly data describing sociological, biological and economic phenomena. Its appearance is so frequent, and the phenomena in which it appears so diverse, that one is led to the conjecture that if these phenomena have any property in common it can only be a similarity in the structure of the underlying probability mechanisms. The empirical distributions to which we shall refer specifically are: (A) distributions of words in prose samples by their frequency of occurrence, (B) distributions of scientists by number of papers published, (C) distributions of cities by population, (D) distributions of incomes by size, and (E) distributions of biological genera by number of species. No one supposes that there is any connexion between horse-kicks suffered by soldiers in the German army and blood cells on a microscope slide other than that the same urn scheme provides a satisfactory abstract model of both phenomena. It is in the same direction that we shall look for an explanation of the observed close similarities among the five classes of distributions listed above. The observed distributions have the following characteristics in common: (a) They are J-shaped, or at least highly skewed, with very long upper tails. The tails can generally be approximated closely by a function of the form"
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | Transitivity measured by the network clustering coefficient despite being extensively studied, is still one of the least understood properties in network analysis according to . Another important property we observe in real networks is the small-world effect -- all nodes are connected with each other by relatively short paths. To model these two properties Watts and Strogatz introduced a small-world network model @cite_34 . | {
"cite_N": [
"@cite_34"
],
"mid": [
"2112090702"
],
"abstract": [
"Networks of coupled dynamical systems have been used to model biological oscillators1,2,3,4, Josephson junction arrays5,6, excitable media7, neural networks8,9,10, spatial games11, genetic control networks12 and many other self-organizing systems. Ordinarily, the connection topology is assumed to be either completely regular or completely random. But many biological, technological and social networks lie somewhere between these two extremes. Here we explore simple models of networks that can be tuned through this middle ground: regular networks ‘rewired’ to introduce increasing amounts of disorder. We find that these systems can be highly clustered, like regular lattices, yet have small characteristic path lengths, like random graphs. We call them ‘small-world’ networks, by analogy with the small-world phenomenon13,14 (popularly known as six degrees of separation15). The neural network of the worm Caenorhabditis elegans, the power grid of the western United States, and the collaboration graph of film actors are shown to be small-world networks. Models of dynamical systems with small-world coupling display enhanced signal-propagation speed, computational power, and synchronizability. In particular, infectious diseases spread more easily in small-world networks than in regular lattices."
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | These quintessential network models are one of the most important contribution towards understanding and modelling complex networks. However, these mathematical models are solely driven by the topology of a network. For example, the Scale-free model considers the degree of a node and the Small-world model considers mutual friends. Neither features nor labels of nodes and or connections are mimicked by those models. However, one can generate synthetic social networks with features is to find similarities correlations between randomly assigned @math number of features and let those similarities define connections @cite_4 @cite_29 @cite_20 . For obvious reasons, this na "ive approach is not ideal due to several limitations. Firstly, correlations between feature vectors do not consider the network topology. Secondly, a common correlation metric would assume every person in a social network views and prefers a potential friend's features equally in a linear fashion. Finally, it is often not obvious what the node labels are, which is an issue we discuss in detail in . | {
"cite_N": [
"@cite_29",
"@cite_4",
"@cite_20"
],
"mid": [
"2112397748",
"1875112053",
"2166639727"
],
"abstract": [
"Online social networks (OSNs) like Facebook, and Myspace recommend new friends to registered users based on local features of the graph (i.e. based on the number of common friends that two users share). However, OSNs do not exploit the whole structure of the network. Instead, they consider only pathways of maximum length 2 between a user and his candidate friends. On the other hand, there are global approaches, which detect the overall path structure in a network, being computationally prohibitive for huge-size social networks. In this paper, we define a basic node similarity measure that captures effectively local graph features. We also exploit global graph features introducing transitive node similarity. Moreover, we derive variants of our method that apply in signed networks. We perform extensive experimental comparison of the proposed method against existing recommendation algorithms using synthetic and real data sets (Facebook, Hi5 and Epinions). Our experimental results show that our FriendTNS algorithm outperforms other approaches in terms of accuracy and it is also time efficient. We show that a significant accuracy improvement can be gained by using information about both positive and negative edges.",
"The scientific study of networks, including computer networks, social networks, and biological networks, has received an enormous amount of interest in the last few years. The rise of the Internet and the wide availability of inexpensive computers have made it possible to gather and analyze network data on a large scale, and the development of a variety of new theoretical tools has allowed us to extract new knowledge from many different kinds of networks.The study of networks is broadly interdisciplinary and important developments have occurred in many fields, including mathematics, physics, computer and information sciences, biology, and the social sciences. This book brings together for the first time the most important breakthroughs in each of these fields and presents them in a coherent fashion, highlighting the strong interconnections between work in different areas. Subjects covered include the measurement and structure of networks in many branches of science, methods for analyzing network data, including methods developed in physics, statistics, and sociology, the fundamentals of graph theory, computer algorithms, and spectral methods, mathematical models of networks, including random graph models and generative models, and theories of dynamical processes taking place on networks.",
"Online social networks (OSNs) recommend new friends to registered users based on local-based features of the graph (i.e. based on the number of common friends that two users share). However, OSNs do not exploit all different length paths of the network. Instead, they consider only pathways of maximum length 2 between a user and his candidate friends. On the other hand, there are global-based approaches, which detect the overall path structure in a network, being computationally prohibitive for huge-sized social networks. In this paper we provide friend recommendations, also known as the link prediction problem, by traversing all paths of a limited length, based on the ''algorithmic small world hypothesis''. As a result, we are able to provide more accurate and faster friend recommendations. We also derive variants of our method that apply to different types of networks (directed undirected and signed unsigned). We perform an extensive experimental comparison of the proposed method against existing link prediction algorithms, using synthetic and three real data sets (Epinions, Facebook and Hi5). We also show that a significant accuracy improvement can be gained by using information about both positive and negative edges. Finally, we discuss extensively various experimental considerations, such as a possible MapReduce implementation of FriendLink algorithm to achieve scalability."
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | provide a guideline on the agent-based modelling of social networks. In the paper by , it is argued that the interplay between the micro and macro level characteristics is complex, and the macro level characteristics are not emergent solely from the simple aggregation of micro level characteristics or low level entities such as social network users @cite_14 . Instead, micro and macro level behaviour or characteristics form a feedback loop, resulting in a nonlinear interaction. From a social network point of view, the graph level and node level characteristics could be thought of as macro and micro level characteristics of the network respectively. | {
"cite_N": [
"@cite_14"
],
"mid": [
"2041157860"
],
"abstract": [
"Models of collective behavior are developed for situations where actors have two alternatives and the costs and or benefits of each depend on how many other actors choose which alternative. The key concept is that of \"threshold\": the number or proportion of others who must make one decision before a given actor does so; this is the point where net benefits begin to exceed net costs for that particular actor. Beginning with a frequency distribution of thresholds, the models allow calculation of the ultimate or \"equilibrium\" number making each decision. The stability of equilibrium results against various possible changes in threshold distributions is considered. Stress is placed on the importance of exact distributions distributions for outcomes. Groups with similar average preferences may generate very different results; hence it is hazardous to infer individual dispositions from aggregate outcomes or to assume that behavior was directed by ultimately agreed-upon norms. Suggested applications are to riot ..."
]
} |
1905.09087 | 2946464731 | A limitation of the Graph Convolutional Networks (GCNs) is that it assumes at a particular @math layer of the neural network model only the @math order neighbourhood nodes of a social network are influential. Furthermore, the GCN has been evaluated on citation and knowledge graphs, but not extensively on friendship-based social graphs. The drawback associated with the dependencies between layers and the order of node neighbourhood for the GCN can be more prevalent for friendship-based graphs. The evaluation of the full potential of the GCN on friendship-based social network requires openly available datasets in larger quantities. However, most available social network datasets are not complete. Also, the majority of the available social network datasets do not contain both the features and ground truth labels. In this work, firstly, we provide a guideline on simulating dynamic social networks, with ground truth labels and features, both coupled with the topology. Secondly, we introduce an open-source Python-based simulation library. We argue that the topology of the network is driven by a set of latent variables, termed as the social DNA (sDNA). We consider the sDNA as labels for the nodes. Finally, by evaluating on our simulated datasets, we propose four new variants of the GCN, mainly to overcome the limitation of dependency between the order of node-neighbourhood and a particular layer of the model. We then evaluate the performance of all the models and our results show that on 27 out of the 30 simulated datasets our proposed GCN variants outperform the original model. | One recent interesting development in modelling dynamic event-based graph is the Cognition-driven Social Network (CogSNet) model @cite_35 . The CogSNet models social network-based on the human memory model. Authors argue that, similar to the human memory, a social event is strengthened by repeated exposure to a similar event and weakens by deprivation of that event. Although CogSNet proposes a new paradigm in social network modelling, it does not provide an explicit explanation modelling features within the dynamic event based graph. Providing open source social network datasets with labels, features, and graph or topological characteristics is the primary goal of this study. | {
"cite_N": [
"@cite_35"
],
"mid": [
"2807762952"
],
"abstract": [
"Human relations are driven by social events - people interact, exchange information, share knowledge and emotions, or gather news from mass media. These events leave traces in human memory. The initial strength of a trace depends on cognitive factors such as emotions or attention span. Each trace continuously weakens over time unless another related event activity strengthens it. Here, we introduce a novel Cognition-driven Social Network (CogSNet) model that accounts for cognitive aspects of social perception and explicitly represents human memory dynamics. For validation, we apply our model to NetSense data on social interactions among university students. The results show that CogSNet significantly improves quality of modeling of human interactions in social networks."
]
} |
1905.09381 | 2951122437 | Humans prove theorems by relying on substantial high-level reasoning and problem-specific insights. Proof assistants offer a formalism that resembles human mathematical reasoning, representing theorems in higher-order logic and proofs as high-level tactics. However, human experts have to construct proofs manually by entering tactics into the proof assistant. In this paper, we study the problem of using machine learning to automate the interaction with proof assistants. We construct CoqGym, a large-scale dataset and learning environment containing 71K human-written proofs from 123 projects developed with the Coq proof assistant. We develop ASTactic, a deep learning-based model that generates tactics as programs in the form of abstract syntax trees (ASTs). Experiments show that ASTactic trained on CoqGym can generate effective tactics and can be used to prove new theorems not previously provable by automated methods. Code is available at this https URL. | FastSMT generates tactics to interact with the Z3 SMT solver @cite_0 , which determines the satisfiability of a certain class of logical formulas. Compared to our model, FastSMT uses substantially simpler tactics---all of them have only boolean and integer arguments, whereas tactics in Coq can have compound expressions as arguments. As a result, the approach of FastSMT does not output ASTs and is not directly applicable to our setting. | {
"cite_N": [
"@cite_0"
],
"mid": [
"1480909796"
],
"abstract": [
"Satisfiability Modulo Theories (SMT) problem is a decision problem for logical first order formulas with respect to combinations of background theories such as: arithmetic, bit-vectors, arrays, and uninterpreted functions. Z3 is a new and efficient SMT Solver freely available from Microsoft Research. It is used in various software verification and analysis applications."
]
} |
1905.09352 | 2946530307 | In this paper, we perform a comprehensive study of 2,470 patched Android vulnerabilities that we collect from different data sources such as Android security bulletins, CVEDetails, Qualcomm Code Aurora, AOSP Git repository, and Linux Patchwork. In our data analysis, we focus on determining the affected layers, OS versions, severity levels, and common weakness enumerations (CWE) associated with the patched vulnerabilities. Further, we assess the timeline of each vulnerability, including discovery and patch dates. We find that (i) even though the number of patched vulnerabilities changes considerably from month to month, the relative number of patched vulnerabilities for each severity level remains stable over time, (ii) there is a significant delay in patching vulnerabilities that originate from the Linux community or concern Qualcomm components, even though Linux and Qualcomm provide and release their own patches earlier, (iii) different AOSP versions receive security updates for different periods of time, (iv) for 94 of patched Android vulnerabilities, the date of disclosure in public datasets is not before the patch release date, (v) there exist some inconsistencies among public vulnerability data sources, e.g., some CVE IDs are listed in Android Security bulletins with detailed information, but in CVEDetails they are listed as unknown, (vi) many patched vulnerabilities for newer Android versions likely also affect older versions that do not receive security patches due to end-of-life. | Most closely related to our work, Linares-V 'a @cite_91 conducted an empirical study based on the collection of 660 Android-related vulnerabilities mined from Android security bulletins, CVEDetails and XML feeds provided by NVD. They investigated three issues. First, they studied the CWE hierarchies and the types of vulnerabilities affecting Android. Second, they studied the Android layers affected by vulnerabilities. Third, they investigated the time intervals between the introduction date of the vulnerability and its fix date. Note that they mined the patched vulnerabilities up to November 2016. On the contrary, we collect 2,470 patched vulnerabilities in Android from August 2015 up to January 2019, which is more comprehensive. Moreover, our study not only covers all of their analysis with a more comprehensive dataset, but also studies new issues. These new investigations include, but are not limited to, the duration of security support for different versions and the delay in patching vulnerabilities originating from Linux and Qualcomm. | {
"cite_N": [
"@cite_91"
],
"mid": [
"1550000763"
],
"abstract": [
"Web users are shown an invalid certificate warning when their browser cannot validate the identity of the websites they are visiting. While these warnings often appear in benign situations, they can also signal a man-in-the-middle attack. We conducted a survey of over 400 Internet users to examine their reactions to and understanding of current SSL warnings. We then designed two new warnings using warnings science principles and lessons learned from the survey. We evaluated warnings used in three popular web browsers and our two warnings in a 100- participant, between-subjects laboratory study. Our warnings performed significantly better than existing warnings, but far too many participants exhibited dangerous behavior in all warning conditions. Our results suggest that, while warnings can be improved, a better approach may be to minimize the use of SSL warnings altogether by blocking users from making unsafe connections and eliminating warnings in benign situations."
]
} |
1905.09352 | 2946530307 | In this paper, we perform a comprehensive study of 2,470 patched Android vulnerabilities that we collect from different data sources such as Android security bulletins, CVEDetails, Qualcomm Code Aurora, AOSP Git repository, and Linux Patchwork. In our data analysis, we focus on determining the affected layers, OS versions, severity levels, and common weakness enumerations (CWE) associated with the patched vulnerabilities. Further, we assess the timeline of each vulnerability, including discovery and patch dates. We find that (i) even though the number of patched vulnerabilities changes considerably from month to month, the relative number of patched vulnerabilities for each severity level remains stable over time, (ii) there is a significant delay in patching vulnerabilities that originate from the Linux community or concern Qualcomm components, even though Linux and Qualcomm provide and release their own patches earlier, (iii) different AOSP versions receive security updates for different periods of time, (iv) for 94 of patched Android vulnerabilities, the date of disclosure in public datasets is not before the patch release date, (v) there exist some inconsistencies among public vulnerability data sources, e.g., some CVE IDs are listed in Android Security bulletins with detailed information, but in CVEDetails they are listed as unknown, (vi) many patched vulnerabilities for newer Android versions likely also affect older versions that do not receive security patches due to end-of-life. | From the user perspective, there are several works that study the role of humans regarding software updates and upgrades. Vaniea and Rashidi @cite_49 found six stages that users go through during their software updates. These six stages are awareness, deciding to update, preparation, installation, troubleshooting, post-state of the update. @cite_36 conducted a survey to better understand the relevant factors for upgrade decisions in Microsoft operating systems by recruiting 239 participants. They studied how users perceive privacy issues associated with OS upgrade decisions, and whether security constitutes a significant decision-making factor. @cite_24 also investigated the life cycle of Microsoft and Apple vulnerabilities by defining a new metric called . They defined it as the number of vulnerability patches that a vendor releases when the vulnerabilities are publicly disclosed. Using this metric, they compared Apple with Microsoft in terms of their security performance and concluded that while Apple shows an ascending trend, Microsoft is more stable than Apple with regard to the average number of unpatched vulnerabilities. Further, Li and Paxson @cite_89 scraped 3,000 vulnerabilities that belong to 682 different open-source software projects and studied the patch development life cycle. | {
"cite_N": [
"@cite_36",
"@cite_24",
"@cite_49",
"@cite_89"
],
"mid": [
"2902464930",
"",
"2338846068",
"2766411424"
],
"abstract": [
"Software upgrades play a pivotal role in enhancing software performance, and are a critical component of resolving software bugs and patching security issues. However, consumers' eagerness to upgrade to the newest operating system is often tempered after release. In this paper, we focus on the upgrade perceptions and practices of users utilizing Microsoft Windows, with particular consideration given to the current upgrade cycle to Windows 10, which was, for a time, offered at no monetary cost to many users. To better understand the relevant factors for upgrade decisions, we deployed a structured survey, including several open-ended questions to add additional depth. We collected data from 239 Microsoft Windows users and utilized qualitative and quantitative methods to analyze user upgrade practices. Important themes include how to best notify users of upcoming upgrade opportunities, how users perceive privacy issues associated with OS upgrade decisions, and whether security constitutes a significant decision-making factor. We also explore how end-of-life dates, indicating the end of support by the vendor, are perceived by users.",
"",
"Updates alter the way software functions by fixing bugs, changing features, and modifying the user interface. Sometimes changes are welcome, even anticipated, and sometimes they are unwanted leading to users avoiding potentially unwanted updates. If users delay or do not install updates it can have serious security implications for their computer. Updates are one of the primary mechanisms for correcting discovered vulnerabilities, when a user does not update they remain vulnerable to an increasing number of attacks. In this work we detail the process users go through when updating their software, including both the positive and negative issues they experience. We asked 307 survey respondents to provide two contrasting software update stories. Using content analysis we analysed the stories and found that users go through six stages while updating: awareness, deciding to update, preparation, installation, troubleshooting, and post state. We further detail the issues respondents experienced during each stage and the impact on their willingness to update.",
"Given how the \"patching treadmill\" plays a central role for enabling sites to counter emergent security concerns, it behooves the security community to understand the patch development process and characteristics of the resulting fixes. Illumination of the nature of security patch development can inform us of shortcomings in existing remediation processes and provide insights for improving current practices. In this work we conduct a large-scale empirical study of security patches, investigating more than 4,000 bug fixes for over 3,000 vulnerabilities that affected a diverse set of 682 open-source software projects. For our analysis we draw upon the National Vulnerability Database, information scraped from relevant external references, affected software repositories, and their associated security fixes. Leveraging this diverse set of information, we conduct an analysis of various aspects of the patch development life cycle, including investigation into the duration of impact a vulnerability has on a code base, the timeliness of patch development, and the degree to which developers produce safe and reliable fixes. We then characterize the nature of security fixes in comparison to other non-security bug fixes, exploring the complexity of different types of patches and their impact on code bases. Among our findings we identify that: security patches have a lower footprint in code bases than non-security bug patches; a third of all security issues were introduced more than 3 years prior to remediation; attackers who monitor open-source repositories can often get a jump of weeks to months on targeting not-yet-patched systems prior to any public disclosure and patch distribution; nearly 5 of security fixes negatively impacted the associated software; and 7 failed to completely remedy the security hole they targeted."
]
} |
1905.09352 | 2946530307 | In this paper, we perform a comprehensive study of 2,470 patched Android vulnerabilities that we collect from different data sources such as Android security bulletins, CVEDetails, Qualcomm Code Aurora, AOSP Git repository, and Linux Patchwork. In our data analysis, we focus on determining the affected layers, OS versions, severity levels, and common weakness enumerations (CWE) associated with the patched vulnerabilities. Further, we assess the timeline of each vulnerability, including discovery and patch dates. We find that (i) even though the number of patched vulnerabilities changes considerably from month to month, the relative number of patched vulnerabilities for each severity level remains stable over time, (ii) there is a significant delay in patching vulnerabilities that originate from the Linux community or concern Qualcomm components, even though Linux and Qualcomm provide and release their own patches earlier, (iii) different AOSP versions receive security updates for different periods of time, (iv) for 94 of patched Android vulnerabilities, the date of disclosure in public datasets is not before the patch release date, (v) there exist some inconsistencies among public vulnerability data sources, e.g., some CVE IDs are listed in Android Security bulletins with detailed information, but in CVEDetails they are listed as unknown, (vi) many patched vulnerabilities for newer Android versions likely also affect older versions that do not receive security patches due to end-of-life. | @cite_10 conducted an exploratory measurement study of 46,310 vulnerabilities disclosed from 1988 to 2011 concerning different software vendors. Their main goals were to analyze disclosure trends, as well as the evolution of CVSS scores and so-called vector metrics such as confidentiality impact, access complexity, and availability impact. One of their findings was that the annual number of disclosed vulnerabilities has stopped increasing since 2008, which is opposite to what we are observing now in the Android context. | {
"cite_N": [
"@cite_10"
],
"mid": [
"2168234580"
],
"abstract": [
"Software systems inherently contain vulnerabilities that have been exploited in the past resulting in significant revenue losses. The study of vulnerability life cycles can help in the development, deployment, and maintenance of software systems. It can also help in designing future security policies and conducting audits of past incidents. Furthermore, such an analysis can help customers to assess the security risks associated with software products of different vendors. In this paper, we conduct an exploratory measurement study of a large software vulnerability data set containing 46310 vulnerabilities disclosed since 1988 till 2011. We investigate vulnerabilities along following seven dimensions: (1) phases in the life cycle of vulnerabilities, (2) evolution of vulnerabilities over the years, (3) functionality of vulnerabilities, (4) access requirement for exploitation of vulnerabilities, (5) risk level of vulnerabilities, (6) software vendors, and (7) software products. Our exploratory analysis uncovers several statistically significant findings that have important implications for software development and deployment."
]
} |
1905.09352 | 2946530307 | In this paper, we perform a comprehensive study of 2,470 patched Android vulnerabilities that we collect from different data sources such as Android security bulletins, CVEDetails, Qualcomm Code Aurora, AOSP Git repository, and Linux Patchwork. In our data analysis, we focus on determining the affected layers, OS versions, severity levels, and common weakness enumerations (CWE) associated with the patched vulnerabilities. Further, we assess the timeline of each vulnerability, including discovery and patch dates. We find that (i) even though the number of patched vulnerabilities changes considerably from month to month, the relative number of patched vulnerabilities for each severity level remains stable over time, (ii) there is a significant delay in patching vulnerabilities that originate from the Linux community or concern Qualcomm components, even though Linux and Qualcomm provide and release their own patches earlier, (iii) different AOSP versions receive security updates for different periods of time, (iv) for 94 of patched Android vulnerabilities, the date of disclosure in public datasets is not before the patch release date, (v) there exist some inconsistencies among public vulnerability data sources, e.g., some CVE IDs are listed in Android Security bulletins with detailed information, but in CVEDetails they are listed as unknown, (vi) many patched vulnerabilities for newer Android versions likely also affect older versions that do not receive security patches due to end-of-life. | @cite_92 investigated the relation between software vulnerability disclosure and patch release time. They found that disclosure accelerates patch release, and open source vendors are quicker in releasing the patches. Alhazmi and Malaiya @cite_88 focused on different vulnerability discovery models in operating systems and compared the results of the models with each other by performing statistical tests. One of the limitations of their model was that it did not differentiate among vulnerabilities with different severity levels. Jo @cite_78 proposed a model for examining competition intensity in the context of patching security vulnerabilities for free-of-charge software products. She found that an increase in market concentration improves vendor response in patching vulnerabilities. To test the model, she focused on the web browser market. @cite_46 studied the issue of competition in a different domain, i.e., the competition in the Android ecosystem and customization of different vendors' offerings which may result in security issues. To solve that issue, they proposed a regulatory fine model, which incentivizes a certain level of investment in security while decreasing the price of products. | {
"cite_N": [
"@cite_88",
"@cite_92",
"@cite_46",
"@cite_78"
],
"mid": [
"1580060001",
"2120197657",
"2778324999",
""
],
"abstract": [
"Security vulnerabilities in servers and operating systems are software defects that represent great risks. Both software developers and users are struggling to contain the risk posed by these vulnerabilities. The vulnerabilities are discovered by both developers and external testers throughout the life-span of a software system. A few models for the vulnerability discovery process have just been published recently. Such models will allow effective resource allocation for patch development and are also needed for evaluating the risk of vulnerability exploitation. Here we examine these models for the vulnerability discovery process. The models are examined both analytically and using actual data on vulnerabilities discovered in three widely-used systems. The applicability of the proposed models and significance of the parameters involved are discussed. The limitations of the proposed models are examined and major research challenges are identified",
"A key aspect of better and more secure software is timely patch release by software vendors for the vulnerabilities in their products. Software vulnerability disclosure, which refers to the publication of vulnerability information, has generated intense debate. An important consideration in this debate is the behavior of software vendors. How quickly do vendors patch vulnerabilities and how does disclosure affect patch release time? This paper compiles a unique data set from the Computer Emergency Response Team Coordination Center (CERT) and SecurityFocus to answer this question. Our results suggest that disclosure accelerates patch release. The instantaneous probability of releasing the patch rises by nearly two and a half times because of disclosure. Open source vendors release patches more quickly than closed source vendors. Vendors are more responsive to more severe vulnerabilities. We also find that vendors respond more slowly to vulnerabilities not disclosed by CERT. We verify our results by using another publicly available data set and find that results are consistent. We also show how our estimates can aid policy makers in their decision making.",
"Vendors in the Android ecosystem typically customize their devices by modifying Android Open Source Project (AOSP) code, adding in-house developed proprietary software, and pre-installing third-party applications. However, research has documented how various security problems are associated with this customization process. We develop a model of the Android ecosystem utilizing the concepts of game theory and product differentiation to capture the competition involving two vendors customizing the AOSP platform. We show how the vendors are incentivized to differentiate their products from AOSP and from each other, and how prices are shaped through this differentiation process. We also consider two types of consumers: security-conscious consumers who understand and care about security, and na \"ive consumers who lack the ability to correctly evaluate security properties of vendor-supplied Android products or simply ignore security. It is evident that vendors shirk on security investments in the latter case. Regulators such as the U.S. Federal Trade Commission have sanctioned Android vendors for underinvestment in security, but the exact effects of these sanctions are difficult to disentangle with empirical data. Here, we model the impact of a regulator-imposed fine that incentivizes vendors to match a minimum security standard. Interestingly, we show how product prices will decrease for the same cost of customization in the presence of a fine, or a higher level of regulator-imposed minimum security.",
""
]
} |
1905.09352 | 2946530307 | In this paper, we perform a comprehensive study of 2,470 patched Android vulnerabilities that we collect from different data sources such as Android security bulletins, CVEDetails, Qualcomm Code Aurora, AOSP Git repository, and Linux Patchwork. In our data analysis, we focus on determining the affected layers, OS versions, severity levels, and common weakness enumerations (CWE) associated with the patched vulnerabilities. Further, we assess the timeline of each vulnerability, including discovery and patch dates. We find that (i) even though the number of patched vulnerabilities changes considerably from month to month, the relative number of patched vulnerabilities for each severity level remains stable over time, (ii) there is a significant delay in patching vulnerabilities that originate from the Linux community or concern Qualcomm components, even though Linux and Qualcomm provide and release their own patches earlier, (iii) different AOSP versions receive security updates for different periods of time, (iv) for 94 of patched Android vulnerabilities, the date of disclosure in public datasets is not before the patch release date, (v) there exist some inconsistencies among public vulnerability data sources, e.g., some CVE IDs are listed in Android Security bulletins with detailed information, but in CVEDetails they are listed as unknown, (vi) many patched vulnerabilities for newer Android versions likely also affect older versions that do not receive security patches due to end-of-life. | A set of studies also investigates the ecosystem around so-called bug bounty platforms, which host vulnerability discovery programs for different companies and offer (monetary) incentives for white hat hackers to participate @cite_98 @cite_75 @cite_64 . While a number of companies offering IoT and mobile services participate in bug bounty programs, Android is not a primary focus of these platforms. | {
"cite_N": [
"@cite_98",
"@cite_64",
"@cite_75"
],
"mid": [
"2520816865",
"2021348304",
"2513442265"
],
"abstract": [
"Bug-bounty programs have the potential to harvest the efforts and diverse knowledge of thousands of white hat hackers. As a consequence, they are becoming increasingly popular as a key part of the security culture of organizations. However, bug-bounty programs can be riddled with myriads of invalid vulnerability-report submissions, which are partially the result of misaligned incentives between white hats and organizations. To further improve the effectiveness of bug-bounty programs, we introduce a theoretical model for evaluating approaches for reducing the number of invalid reports. We develop an economic framework and investigate the strengths and weaknesses of existing canonical approaches for effectively incentivizing higher validation efforts by white hats. Finally, we introduce a novel approach, which may improve efficiency by enabling different white hats to exert validation effort at their individually optimal levels.",
"In recent years, many organizations have established bounty programs that attract white hat hackers who contribute vulnerability reports of web systems. In this paper, we collect publicly available data of two representative web vulnerability discovery ecosystems (Wooyun and HackerOne) and study their characteristics, trajectory, and impact. We find that both ecosystems include large and continuously growing white hat communities which have provided significant contributions to organizations from a wide range of business sectors. We also analyze vulnerability trends, response and resolve behaviors, and reward structures of participating organizations. Our analysis based on the HackerOne dataset reveals that a considerable number of organizations exhibit decreasing trends for reported web vulnerabilities. We further conduct a regression study which shows that monetary incentives have a significantly positive correlation with the number of vulnerabilities reported. Finally, we make recommendations aimed at increasing participation by white hats and organizations in such ecosystems.",
"Author(s): Maillart, Thomas; Zhao, Mingyi; Grossklags, Jens; Chuang, John | Abstract: Bug bounty programs offer a modern way for organizations to crowdsource their software security, and for security researchers to be fairly rewarded for the vulnerabilities they find. However, little is known on the incentives set by bug bounty programs – how they drive engagement and new bug discoveries. This article provides an empirical investigation of the strategic interactions among the managers and participants of bug bounty programs, as well as the intermediation by bug bounty platforms. We find that for a given bug bounty program, each security researcher can only expect to discover a bounded number of bugs. This result offers a validation step to a theory brought forth early on by This theory proposes that each security researcher inspecting a piece of software offers a unique environment of skills and mindset, which is amenable to the discovery of bugs that others may not be able to uncover. Bug bounty programs indeed benefit from the engagement of large crowds of researchers. Conversely, security researchers benefit greatly from searching for bugs in multiple bug bounty programs. However, we find that following a strong front-loading effect, newly launched programs attract researchers at the expense of older programs: the probability of finding bugs decays as ∼1 t0.4∼1 t0.4 after the launch of a program, even though bugs found later yield on average higher rewards. Our results lead us to formulate three recommendations for organizing bug bounty programs and platforms: (i) organize enrollment, mobility, and renewal of security researchers across bounty programs, (ii) highlight and organize programs for front-loading, and (iii) organize fluid market transactions to reduce uncertainty and thus reduce incentives for security researchers to sell on the black market."
]
} |
1905.09272 | 2944828972 | Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10 Top-5 accuracy and supervised methods by 20 . Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. | The promise of unsupervised representation learning has existed since the earliest days of learning in computer vision @cite_43 , inspired by the fact that humans and animals need little supervision to perform a range of complex visual tasks. The intuition that a good representation should faithfully reconstruct the input data was soon exploited to achieve representations that resemble simple cells in the brain @cite_16 . The idea was later formalized in the context of probabilistic generative models @cite_66 , which have become a staple of modern unsupervised learning @cite_17 @cite_28 @cite_12 @cite_47 . | {
"cite_N": [
"@cite_28",
"@cite_43",
"@cite_47",
"@cite_16",
"@cite_66",
"@cite_12",
"@cite_17"
],
"mid": [
"2099471712",
"197865394",
"2962897886",
"2145889472",
"1993845689",
"",
"2136922672"
],
"abstract": [
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.",
"A neural network model, called a “neocognitron”, is proposed for a mechanism of visual pattern recognition. It is demonstrated by computer simulation that the neocognitron has characteristics similar to those of visual systems of vertebrates.",
"We marry ideas from deep neural networks and approximate Bayesian inference to derive a generalised class of deep, directed generative models, endowed with a new algorithm for scalable inference and learning. Our algorithm introduces a recognition model to represent an approximate posterior distribution and uses this for optimisation of a variational lower bound. We develop stochastic backpropagation - rules for gradient backpropagation through stochastic variables - and derive an algorithm that allows for joint optimisation of the parameters of both the generative and recognition models. We demonstrate on several real-world data sets that by using stochastic backpropagation and variational inference, we obtain models that are able to generate realistic samples of data, allow for accurate imputations of missing data, and provide a useful tool for high-dimensional data visualisation.",
"THE receptive fields of simple cells in mammalian primary visual cortex can be characterized as being spatially localized, oriented1–4 and bandpass (selective to structure at different spatial scales), comparable to the basis functions of wavelet transforms5,6. One approach to understanding such response properties of visual neurons has been to consider their relationship to the statistical structure of natural images in terms of efficient coding7–12. Along these lines, a number of studies have attempted to train unsupervised learning algorithms on natural images in the hope of developing receptive fields with similar properties13–18, but none has succeeded in producing a full set that spans the image space and contains all three of the above properties. Here we investigate the proposal8,12 that a coding strategy that maximizes sparseness is sufficient to account for these properties. We show that a learning algorithm that attempts to find sparse linear codes for natural scenes will develop a complete family of localized, oriented, bandpass receptive fields, similar to those found in the primary visual cortex. The resulting sparse image code provides a more efficient representation for later stages of processing because it possesses a higher degree of statistical independence among its outputs.",
"An unsupervised learning algorithm for a multilayer network of stochastic neurons is described. Bottom-up \"recognition\" connections convert the input into representations in successive hidden layers, and top-down \"generative\" connections reconstruct the representation in one layer from the representation in the layer above. In the \"wake\" phase, neurons are driven by recognition connections, and generative connections are adapted to increase the probability that they would reconstruct the correct activity vector in the layer below. In the \"sleep\" phase, neurons are driven by generative connections, and recognition connections are adapted to increase the probability that they would produce the correct activity vector in the layer above.",
"",
"We show how to use \"complementary priors\" to eliminate the explaining-away effects that make inference difficult in densely connected belief nets that have many hidden layers. Using complementary priors, we derive a fast, greedy algorithm that can learn deep, directed belief networks one layer at a time, provided the top two layers form an undirected associative memory. The fast, greedy algorithm is used to initialize a slower learning procedure that fine-tunes the weights using a contrastive version of the wake-sleep algorithm. After fine-tuning, a network with three hidden layers forms a very good generative model of the joint distribution of handwritten digit images and their labels. This generative model gives better digit classification than the best discriminative learning algorithms. The low-dimensional manifolds on which the digits lie are modeled by long ravines in the free-energy landscape of the top-level associative memory, and it is easy to explore these ravines by using the directed connections to display what the associative memory has in mind."
]
} |
1905.09272 | 2944828972 | Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10 Top-5 accuracy and supervised methods by 20 . Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. | A key problem that generative models must overcome, however, is that not all low-level details are equally important in visual perception: for instance, lighting, compression artifacts, and the exact position of texture elements are irrelevant to most downstream tasks. This resulted in some early attempts to structure representations with losses that did not rely on reconstruction: notably, slow feature analysis @cite_5 @cite_71 used the persistence of objects in videos to learn representations that generalized across views. | {
"cite_N": [
"@cite_5",
"@cite_71"
],
"mid": [
"2096388912",
"2146444479"
],
"abstract": [
"The visual system can reliably identify objects even when the retinal image is transformed considerably by commonly occurring changes in the environment. A local learning rule is proposed, which allows a network to learn to generalize across such transformations. During the learning phase, the network is exposed to temporal sequences of patterns undergoing the transformation. An application of the algorithm is presented in which the network learns invariance to shift in retinal position. Such a principle may be involved in the development of the characteristic shift invariance property of complex cells in the primary visual cortex, and also in the development of more complicated invariance properties of neurons in higher visual areas.",
"Invariant features of temporally varying signals are useful for analysis and classification. Slow feature analysis (SFA) is a new method for learning invariant or slowly varying features from a vectorial input signal. It is based on a nonlinear expansion of the input signal and application of principal component analysis to this expanded signal and its time derivative. It is guaranteed to find the optimal solution within a family of functions directly and can learn to extract a large number of decorrelated features, which are ordered by their degree of invariance. SFA can be applied hierarchically to process high-dimensional input signals and extract complex features. SFA is applied first to complex cell tuning properties based on simple cell output, including disparity and motion. Then more complicated input-output functions are learned by repeated application of SFA. Finally, a hierarchical network of SFA modules is presented as a simple model of the visual system. The same unstructured network can learn translation, size, rotation, contrast, or, to a lesser degree, illumination invariance for one-dimensional objects, depending on only the training stimulus. Surprisingly, only a few training objects suffice to achieve good generalization to new objects. The generated representation is suitable for object recognition. Performance degrades if the network is trained to learn multiple invariances simultaneously."
]
} |
1905.09272 | 2944828972 | Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10 Top-5 accuracy and supervised methods by 20 . Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. | Recent years have witnessed a proliferation of such unsupervised objectives. For example, simply asking a network to recognize the spatial layout of an image, without reconstructing it, led to representations that transferred to popular vision tasks such as classification and detection @cite_8 @cite_3 @cite_73 . Other works showed that color @cite_40 @cite_29 , image orientation @cite_62 , and data augmentation @cite_76 can support useful self-supervised tasks from single images. | {
"cite_N": [
"@cite_62",
"@cite_8",
"@cite_29",
"@cite_3",
"@cite_40",
"@cite_76",
"@cite_73"
],
"mid": [
"2785325870",
"343636949",
"2599837529",
"2321533354",
"2326925005",
"2148349024",
"2842511635"
],
"abstract": [
"Over the last years, deep convolutional neural networks (ConvNets) have transformed the field of computer vision thanks to their unparalleled capacity to learn high level semantic image features. However, in order to successfully learn those features, they usually require massive amounts of manually labeled data, which is both expensive and impractical to scale. Therefore, unsupervised semantic feature learning, i.e., learning without requiring manual annotation effort, is of crucial importance in order to successfully harvest the vast amount of visual data that are available today. In our work we propose to learn image features by training ConvNets to recognize the 2d rotation that is applied to the image that it gets as input. We demonstrate both qualitatively and quantitatively that this apparently simple task actually provides a very powerful supervisory signal for semantic feature learning. We exhaustively evaluate our method in various unsupervised feature learning benchmarks and we exhibit in all of them state-of-the-art performance. Specifically, our results on those benchmarks demonstrate dramatic improvements w.r.t. prior state-of-the-art approaches in unsupervised representation learning and thus significantly close the gap with supervised feature learning. For instance, in PASCAL VOC 2007 detection task our unsupervised pre-trained AlexNet model achieves the state-of-the-art (among unsupervised methods) mAP of 54.4 that is only 2.4 points lower from the supervised case. We get similarly striking results when we transfer our unsupervised learned features on various other tasks, such as ImageNet classification, PASCAL classification, PASCAL segmentation, and CIFAR-10 classification. The code and models of our paper will be published on: this https URL .",
"This work explores the use of spatial context as a source of free and plentiful supervisory signal for training a rich visual representation. Given only a large, unlabeled image collection, we extract random pairs of patches from each image and train a convolutional neural net to predict the position of the second patch relative to the first. We argue that doing well on this task requires the model to learn to recognize objects and their parts. We demonstrate that the feature representation learned using this within-image context indeed captures visual similarity across images. For example, this representation allows us to perform unsupervised visual discovery of objects like cats, people, and even birds from the Pascal VOC 2011 detection dataset. Furthermore, we show that the learned ConvNet can be used in the R-CNN framework [19] and provides a significant boost over a randomly-initialized ConvNet, resulting in state-of-the-art performance among algorithms which use only Pascal-provided training set annotations.",
"We investigate and improve self-supervision as a drop-in replacement for ImageNet pretraining, focusing on automatic colorization as the proxy task. Self-supervised training has been shown to be more promising for utilizing unlabeled data than other, traditional unsupervised learning methods. We build on this success and evaluate the ability of our self-supervised network in several contexts. On VOC segmentation and classification tasks, we present results that are state-of-the-art among methods not using ImageNet labels for pretraining representations. Moreover, we present the first in-depth analysis of self-supervision via colorization, concluding that formulation of the loss, training details and network architecture play important roles in its effectiveness. This investigation is further expanded by revisiting the ImageNet pretraining paradigm, asking questions such as: How much training data is needed? How many labels are needed? How much do features change when fine-tuned? We relate these questions back to self-supervision by showing that colorization provides a similarly powerful supervisory signal as various flavors of ImageNet pretraining.",
"We propose a novel unsupervised learning approach to build features suitable for object detection and classification. The features are pre-trained on a large dataset without human annotation and later transferred via fine-tuning on a different, smaller and labeled dataset. The pre-training consists of solving jigsaw puzzles of natural images. To facilitate the transfer of features to other tasks, we introduce the context-free network (CFN), a siamese-ennead convolutional neural network. The features correspond to the columns of the CFN and they process image tiles independently (i.e., free of context). The later layers of the CFN then use the features to identify their geometric arrangement. Our experimental evaluations show that the learned features capture semantically relevant content. We pre-train the CFN on the training set of the ILSVRC2012 dataset and transfer the features on the combined training and validation set of Pascal VOC 2007 for object detection (via fast RCNN) and classification. These features outperform all current unsupervised features with (51.8 , ) for detection and (68.6 , ) for classification, and reduce the gap with supervised learning ( (56.5 , ) and (78.2 , ) respectively).",
"Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result. The system is implemented as a feed-forward pass in a CNN at test time and is trained on over a million color images. We evaluate our algorithm using a “colorization Turing test,” asking human participants to choose between a generated and ground truth color image. Our method successfully fools humans on 32 of the trials, significantly higher than previous methods. Moreover, we show that colorization can be a powerful pretext task for self-supervised feature learning, acting as a cross-channel encoder. This approach results in state-of-the-art performance on several feature learning benchmarks.",
"Current methods for training convolutional neural networks depend on large amounts of labeled samples for supervised training. In this paper we present an approach for training a convolutional neural network using only unlabeled data. We train the network to discriminate between a set of surrogate classes. Each surrogate class is formed by applying a variety of transformations to a randomly sampled 'seed' image patch. We find that this simple feature learning algorithm is surprisingly successful when applied to visual object recognition. The feature representation learned by our algorithm achieves classification results matching or outperforming the current state-of-the-art for unsupervised learning on several popular datasets (STL-10, CIFAR-10, Caltech-101).",
"While supervised learning has enabled great progress in many applications, unsupervised learning has not seen such widespread adoption, and remains an important and challenging endeavor for artificial intelligence. In this work, we propose a universal unsupervised learning approach to extract useful representations from high-dimensional data, which we call Contrastive Predictive Coding. The key insight of our model is to learn such representations by predicting the future in latent space by using powerful autoregressive models. We use a probabilistic contrastive loss which induces the latent space to capture information that is maximally useful to predict future samples. It also makes the model tractable by using negative sampling. While most prior work has focused on evaluating representations for a particular modality, we demonstrate that our approach is able to learn useful representations achieving strong performance on four distinct domains: speech, images, text and reinforcement learning in 3D environments."
]
} |
1905.09272 | 2944828972 | Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10 Top-5 accuracy and supervised methods by 20 . Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. | Extra information that occurs with images can allow for even richer objectives to train representations. Video is a particularly strong source of cues, where notable works have leveraged object tracking @cite_9 , frame ordering @cite_2 , and object boundary cues @cite_67 @cite_64 . Beyond video, information about camera motion @cite_4 @cite_55 , scene geometry @cite_57 , or sound @cite_30 @cite_48 can all serve as natural sources of supervision. One particularly important domain for self-supervised learning is robotics, where representations can be learned by interacting with the environment @cite_78 @cite_24 @cite_32 @cite_54 , and where success greatly depends on making effective use of unlabeled data. Tasks can often be combined to form even more powerful learning signals, allowing larger networks to be trained successfully @cite_21 @cite_63 @cite_34 . | {
"cite_N": [
"@cite_30",
"@cite_67",
"@cite_64",
"@cite_4",
"@cite_78",
"@cite_48",
"@cite_9",
"@cite_55",
"@cite_54",
"@cite_32",
"@cite_21",
"@cite_57",
"@cite_24",
"@cite_63",
"@cite_2",
"@cite_34"
],
"mid": [
"2619697695",
"2962958090",
"2950341389",
"1520997877",
"2473208550",
"2963115079",
"219040644",
"2198618282",
"2949536664",
"2528973876",
"2962824366",
"2520377600",
"2201912979",
"2798512429",
"2487442924",
""
],
"abstract": [
"We consider the question: what can be learnt by looking at and listening to a large number of unlabelled videos? There is a valuable, but so far untapped, source of information contained in the video itself – the correspondence between the visual and the audio streams, and we introduce a novel “Audio-Visual Correspondence” learning task that makes use of this. Training visual and audio networks from scratch, without any additional supervision other than the raw unconstrained videos themselves, is shown to successfully solve this task, and, more interestingly, result in good visual and audio representations. These features set the new state-of-the-art on two sound classification benchmarks, and perform on par with the state-of-the-art selfsupervised approaches on ImageNet classification. We also demonstrate that the network is able to localize objects in both modalities, as well as perform fine-grained recognition tasks.",
"Data-driven approaches for edge detection have proven effective and achieve top results on modern benchmarks. However, all current data-driven edge detectors require manual supervision for training in the form of hand-labeled region segments or object boundaries. Specifically, human annotators mark semantically meaningful edges which are subsequently used for training. Is this form of strong, highlevel supervision actually necessary to learn to accurately detect edges? In this work we present a simple yet effective approach for training edge detectors without human supervision. To this end we utilize motion, and more specifically, the only input to our method is noisy semi-dense matches between frames. We begin with only a rudimentary knowledge of edges (in the form of image gradients), and alternate between improving motion estimation and edge detection in turn. Using a large corpus of video data, we show that edge detectors trained using our unsupervised scheme approach the performance of the same methods trained with full supervision (within 3-5 ). Finally, we show that when using a deep network for the edge detector, our approach provides a novel pre-training scheme for object detection.",
"This paper presents a novel yet intuitive approach to unsupervised feature learning. Inspired by the human visual system, we explore whether low-level motion-based grouping cues can be used to learn an effective visual representation. Specifically, we use unsupervised motion-based segmentation on videos to obtain segments, which we use as 'pseudo ground truth' to train a convolutional network to segment objects from a single frame. Given the extensive evidence that motion plays a key role in the development of the human visual system, we hope that this straightforward approach to unsupervised learning will be more effective than cleverly designed 'pretext' tasks studied in the literature. Indeed, our extensive experiments show that this is the case. When used for transfer learning on object detection, our representation significantly outperforms previous unsupervised approaches across multiple settings, especially when training data for the target task is scarce.",
"The current dominant paradigm for feature learning in computer vision relies on training neural networks for the task of object recognition using millions of hand labelled images. Is it also possible to learn features for a diverse set of visual tasks using any other form of supervision? In biology, living organisms developed the ability of visual perception for the purpose of moving and acting in the world. Drawing inspiration from this observation, in this work we investigated if the awareness of egomotion(i.e. self motion) can be used as a supervisory signal for feature learning. As opposed to the knowledge of class labels, information about egomotion is freely available to mobile agents. We found that using the same number of training images, features learnt using egomotion as supervision compare favourably to features learnt using class-label as supervision on the tasks of scene recognition, object recognition, visual odometry and keypoint matching.",
"We investigate an experiential learning paradigm for acquiring an internal model of intuitive physics. Our model is evaluated on a real-world robotic manipulation task that requires displacing objects to target locations by poking. The robot gathered over 400 hours of experience by executing more than 100K pokes on different objects. We propose a novel approach based on deep neural networks for modeling the dynamics of robot's interactions directly from images, by jointly estimating forward and inverse models of dynamics. The inverse model objective provides supervision to construct informative visual features, which the forward model can then predict and in turn regularize the feature space for the inverse model. The interplay between these two objectives creates useful, accurate models that can then be used for multi-step decision making. This formulation has the additional benefit that it is possible to learn forward models in an abstract feature space and thus alleviate the need of predicting pixels. Our experiments show that this joint modeling approach outperforms alternative methods.",
"In this paper our objectives are, first, networks that can embed audio and visual inputs into a common space that is suitable for cross-modal retrieval; and second, a network that can localize the object that sounds in an image, given the audio signal. We achieve both these objectives by training from unlabelled video using only audio-visual correspondence (AVC) as the objective function. This is a form of cross-modal self-supervision from video.",
"Is strong supervision necessary for learning a good visual representation? Do we really need millions of semantically-labeled images to train a Convolutional Neural Network (CNN)? In this paper, we present a simple yet surprisingly powerful approach for unsupervised learning of CNN. Specifically, we use hundreds of thousands of unlabeled videos from the web to learn visual representations. Our key idea is that visual tracking provides the supervision. That is, two patches connected by a track should have similar visual representation in deep feature space since they probably belong to same object or object part. We design a Siamese-triplet network with a ranking loss function to train this CNN representation. Without using a single image from ImageNet, just using 100K unlabeled videos and the VOC 2012 dataset, we train an ensemble of unsupervised networks that achieves 52 mAP (no bounding box regression). This performance comes tantalizingly close to its ImageNet-supervised counterpart, an ensemble which achieves a mAP of 54.4 . We also show that our unsupervised network can perform competitively in other tasks such as surface-normal estimation.",
"Understanding how images of objects and scenes behave in response to specific ego-motions is a crucial aspect of proper visual development, yet existing visual learning methods are conspicuously disconnected from the physical source of their images. We propose to exploit proprioceptive motor signals to provide unsupervised regularization in convolutional neural networks to learn visual representations from egocentric video. Specifically, we enforce that our learned features exhibit equivariance, i.e, they respond predictably to transformations associated with distinct ego-motions. With three datasets, we show that our unsupervised feature learning approach significantly outperforms previous approaches on visual recognition and next-best-view prediction tasks. In the most challenging test, we show that features learned from video captured on an autonomous driving platform improve large-scale scene recognition in static images from a disjoint domain.",
"We propose a self-supervised approach for learning representations and robotic behaviors entirely from unlabeled videos recorded from multiple viewpoints, and study how this representation can be used in two robotic imitation settings: imitating object interactions from videos of humans, and imitating human poses. Imitation of human behavior requires a viewpoint-invariant representation that captures the relationships between end-effectors (hands or robot grippers) and the environment, object attributes, and body pose. We train our representations using a metric learning loss, where multiple simultaneous viewpoints of the same observation are attracted in the embedding space, while being repelled from temporal neighbors which are often visually similar but functionally different. In other words, the model simultaneously learns to recognize what is common between different-looking images, and what is different between similar-looking images. This signal causes our model to discover attributes that do not change across viewpoint, but do change across time, while ignoring nuisance variables such as occlusions, motion blur, lighting and background. We demonstrate that this representation can be used by a robot to directly mimic human poses without an explicit correspondence, and that it can be used as a reward function within a reinforcement learning algorithm. While representations are learned from an unlabeled collection of task-related videos, robot behaviors such as pouring are learned by watching a single 3rd-person demonstration by a human. Reward functions obtained by following the human demonstrations under the learned representation enable efficient reinforcement learning that is practical for real-world robotic systems. Video results, open-source code and dataset are available at this https URL",
"There has been a recent paradigm shift in robotics to data-driven learning for planning and control. Due to large number of experiences required for training, most of these approaches use a self-supervised paradigm: using sensors to measure success failure. However, in most cases, these sensors provide weak supervision at best. In this work, we propose an adversarial learning framework that pits an adversary against the robot learning the task. In an effort to defeat the adversary, the original robot learns to perform the task with more robustness leading to overall improved performance. We show that this adversarial framework forces the the robot to learn a better grasping model in order to overcome the adversary. By grasping 82 of presented novel objects compared to 68 without an adversary, we demonstrate the utility of creating adversaries. We also demonstrate via experiments that having robots in adversarial setting might be a better learning strategy as compared to having collaborative multiple robots.",
"We investigate methods for combining multiple selfsupervised tasks—i.e., supervised tasks where data can be collected without manual labeling—in order to train a single visual representation. First, we provide an apples-toapples comparison of four different self-supervised tasks using the very deep ResNet-101 architecture. We then combine tasks to jointly train a network. We also explore lasso regularization to encourage the network to factorize the information in its representation, and methods for “harmonizing” network inputs in order to learn a more unified representation. We evaluate all methods on ImageNet classification, PASCAL VOC detection, and NYU depth prediction. Our results show that deeper networks work better, and that combining tasks—even via a na¨ýve multihead architecture—always improves performance. Our best joint network nearly matches the PASCAL performance of a model pre-trained on ImageNet classification, and matches the ImageNet network on NYU depth prediction.",
"Though a large body of computer vision research has investigated developing generic semantic representations, efforts towards developing a similar representation for 3D has been limited. In this paper, we learn a generic 3D representation through solving a set of foundational proxy 3D tasks: object-centric camera pose estimation and wide baseline feature matching. Our method is based upon the premise that by providing supervision over a set of carefully selected foundational tasks, generalization to novel tasks and abstraction capabilities can be achieved. We empirically show that the internal representation of a multi-task ConvNet trained to solve the above core problems generalizes to novel 3D tasks (e.g., scene layout estimation, object pose estimation, surface normal estimation) without the need for fine-tuning and shows traits of abstraction abilities (e.g., cross modality pose estimation).",
"Current model free learning-based robot grasping approaches exploit human-labeled datasets for training the models. However, there are two problems with such a methodology: (a) since each object can be grasped in multiple ways, manually labeling grasp locations is not a trivial task; (b) human labeling is biased by semantics. While there have been attempts to train robots using trial-and-error experiments, the amount of data used in such experiments remains substantially low and hence makes the learner prone to over-fitting. In this paper, we take the leap of increasing the available training data to 40 times more than prior work, leading to a dataset size of 50K data points collected over 700 hours of robot grasping attempts. This allows us to train a Convolutional Neural Network (CNN) for the task of predicting grasp locations without severe overfitting. In our formulation, we recast the regression problem to an 18-way binary classification over image patches. We also present a multi-stage learning approach where a CNN trained in one stage is used to collect hard negatives in subsequent stages. Our experiments clearly show the benefit of using large-scale datasets (and multi-stage training) for the task of grasping. We also compare to several baselines and show state-of-the-art performance on generalization to unseen objects for grasping.",
"Do visual tasks have a relationship, or are they unrelated? For instance, could having surface normals simplify estimating the depth of an image? Intuition answers these questions positively, implying existence of a structure among visual tasks. Knowing this structure has notable values; it is the concept underlying transfer learning and provides a principled way for identifying redundancies across tasks, e.g., to seamlessly reuse supervision among related tasks or solve many tasks in one system without piling up the complexity. We proposes a fully computational approach for modeling the structure of space of visual tasks. This is done via finding (first and higher-order) transfer learning dependencies across a dictionary of twenty six 2D, 2.5D, 3D, and semantic tasks in a latent space. The product is a computational taxonomic map for task transfer learning. We study the consequences of this structure, e.g. nontrivial emerged relationships, and exploit them to reduce the demand for labeled data. For example, we show that the total number of labeled datapoints needed for solving a set of 10 tasks can be reduced by roughly 2 3 (compared to training independently) while keeping the performance nearly the same. We provide a set of tools for computing and probing this taxonomical structure including a solver that users can employ to devise efficient supervision policies for their use cases.",
"In this paper, we present an approach for learning a visual representation from the raw spatiotemporal signals in videos. Our representation is learned without supervision from semantic labels. We formulate our method as an unsupervised sequential verification task, i.e., we determine whether a sequence of frames from a video is in the correct temporal order. With this simple task and no semantic labels, we learn a powerful visual representation using a Convolutional Neural Network (CNN). The representation contains complementary information to that learned from supervised image datasets like ImageNet. Qualitative results show that our method captures information that is temporally varying, such as human pose. When used as pre-training for action recognition, our method gives significant gains over learning without external data on benchmark datasets like UCF101 and HMDB51. To demonstrate its sensitivity to human pose, we show results for pose estimation on the FLIC and MPII datasets that are competitive, or better than approaches using significantly more supervision. Our method can be combined with supervised representations to provide an additional boost in accuracy.",
""
]
} |
1905.09272 | 2944828972 | Large scale deep learning excels when labeled images are abundant, yet data-efficient learning remains a longstanding challenge. While biological vision is thought to leverage vast amounts of unlabeled data to solve classification problems with limited supervision, computer vision has so far not succeeded in this semi-supervised' regime. Our work tackles this challenge with Contrastive Predictive Coding, an unsupervised objective which extracts stable structure from still images. The result is a representation which, equipped with a simple linear classifier, separates ImageNet categories better than all competing methods, and surpasses the performance of a fully-supervised AlexNet model. When given a small number of labeled images (as few as 13 per class), this representation retains a strong classification performance, outperforming state-of-the-art semi-supervised methods by 10 Top-5 accuracy and supervised methods by 20 . Finally, we find our unsupervised representation to serve as a useful substrate for image detection on the PASCAL-VOC 2007 dataset, approaching the performance of representations trained with a fully annotated ImageNet dataset. We expect these results to open the door to pipelines that use scalable unsupervised representations as a drop-in replacement for supervised ones for real-world vision tasks where labels are scarce. | Other approaches to semi-supervised learning train simultaneously on labeled and unlabeled data. Many such works have roots in early work on bootstrapping @cite_10 and co-training @cite_75 from text (see @cite_15 for a survey). An increasingly popular approach in deep learning consists in propagating information learned from labeled images to unlabeled ones @cite_50 @cite_77 while enforcing invariance to perturbations @cite_25 @cite_20 @cite_60 . In relying on labeled data to provide a learning signal for unlabeled data, these methods could be limited in settings with very low labeled data. | {
"cite_N": [
"@cite_75",
"@cite_60",
"@cite_77",
"@cite_50",
"@cite_15",
"@cite_10",
"@cite_25",
"@cite_20"
],
"mid": [
"2048679005",
"",
"",
"2145494108",
"2136504847",
"1930023685",
"2964159205",
"2963435192"
],
"abstract": [
"We consider the problem of using a large unlabeled sample to boost performance of a learning algorit,hrn when only a small set of labeled examples is available. In particular, we consider a problem setting motivated by the task of learning to classify web pages, in which the description of each example can be partitioned into two distinct views. For example, the description of a web page can be partitioned into the words occurring on that page, and the words occurring in hyperlinks t,hat point to that page. We assume that either view of the example would be sufficient for learning if we had enough labeled data, but our goal is to use both views together to allow inexpensive unlabeled data to augment, a much smaller set of labeled examples. Specifically, the presence of two distinct views of each example suggests strategies in which two learning algorithms are trained separately on each view, and then each algorithm’s predictions on new unlabeled examples are used to enlarge the training set of the other. Our goal in this paper is to provide a PAC-style analysis for this setting, and, more broadly, a PAC-style framework for the general problem of learning from both labeled and unlabeled data. We also provide empirical results on real web-page data indicating that this use of unlabeled examples can lead to significant improvement of hypotheses in practice. *This research was supported in part by the DARPA HPKB program under contract F30602-97-1-0215 and by NSF National Young investigator grant CCR-9357793. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. TO copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and or a fee. COLT 98 Madison WI USA Copyright ACM 1998 l-58113-057--0 98 7... 5.00 92 Tom Mitchell School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213-3891 mitchell+@cs.cmu.edu",
"",
"",
"We consider the semi-supervised learning problem, where a decision rule is to be learned from labeled and unlabeled data. In this framework, we motivate minimum entropy regularization, which enables to incorporate unlabeled data in the standard supervised learning. Our approach includes other approaches to the semi-supervised problem as particular or limiting cases. A series of experiments illustrates that the proposed solution benefits from unlabeled data. The method challenges mixture models when the data are sampled from the distribution class spanned by the generative model. The performances are definitely in favor of minimum entropy regularization when generative models are misspecified, and the weighting of unlabeled data provides robustness to the violation of the \"cluster assumption\". Finally, we also illustrate that the method can also be far superior to manifold learning in high dimension spaces.",
"Door lock apparatus in which a door latch mechanism is operated by inner and outer door handles coupled to a latch shaft extending through the latch mechanism. Handles are coupled to ends of latch shaft by coupling devices enabling door to be locked from the inside to prevent entry from the outside but can still be opened from the inside by normal operation of outside handle. Inside coupling device has limited lost-motion which is used to operate cam device to unlock the door on actuation of inner handles.",
"",
"We propose a new regularization method based on virtual adversarial loss: a new measure of local smoothness of the conditional label distribution given input. Virtual adversarial loss is defined as the robustness of the conditional label distribution around each input data point against local perturbation. Unlike adversarial training, our method defines the adversarial direction without label information and is hence applicable to semi-supervised learning. Because the directions in which we smooth the model are only “virtually” adversarial, we call our method virtual adversarial training (VAT). The computational cost of VAT is relatively low. For neural networks, the approximated gradient of virtual adversarial loss can be computed with no more than two pairs of forward- and back-propagations. In our experiments, we applied VAT to supervised and semi-supervised learning tasks on multiple benchmark datasets. With a simple enhancement of the algorithm based on the entropy minimization principle, our VAT achieves state-of-the-art performance for semi-supervised learning tasks on SVHN and CIFAR-10.",
"Effective convolutional neural networks are trained on large sets of labeled data. However, creating large labeled datasets is a very costly and time-consuming task. Semi-supervised learning uses unlabeled data to train a model with higher accuracy when there is a limited set of labeled data available. In this paper, we consider the problem of semi-supervised learning with convolutional neural networks. Techniques such as randomized data augmentation, dropout and random max-pooling provide better generalization and stability for classifiers that are trained using gradient descent. Multiple passes of an individual sample through the network might lead to different predictions due to the non-deterministic behavior of these techniques. We propose an unsupervised loss function that takes advantage of the stochastic nature of these methods and minimizes the difference between the predictions of multiple passes of a training sample through the network. We evaluate the proposed method on several benchmark datasets."
]
} |
1905.08789 | 2946689852 | Can we improve detection in the thermal domain by borrowing features from rich domains like visual RGB? In this paper, we propose a pseudo-multimodal object detector trained on natural image domain data to help improve the performance of object detection in thermal images. We assume access to a large-scale dataset in the visual RGB domain and relatively smaller dataset (in terms of instances) in the thermal domain, as is common today. We propose the use of well-known image-to-image translation frameworks to generate pseudo-RGB equivalents of a given thermal image and then use a multi-modal architecture for object detection in the thermal image. We show that our framework outperforms existing benchmarks without the explicit need for paired training examples from the two domains. We also show that our framework has the ability to learn with less data from thermal domain when using our approach. | Among efforts that use a multimodal approach, Wagner @cite_9 applied Aggregated Channel Features (ACF) and Boosted Decision trees (BDT) for proposal generation and classified these proposals with a CNN, which fuses Visual and IR information. Choi @cite_28 uses two separate region proposal networks for both Visual and IR images and evaluates the proposals generated by both the networks with Support Vector Regression on fused deep features. The efforts closest to our work are that of Konig @cite_0 and Liu @cite_29 , both of which propose a multi-modal framework that combines RGB and thermal information in a Faster-RCNN architecture by posing it as a convolutional network fusion problem. However, all of these multimodal efforts assume the availability of a dataset with paired training examples from the visual and thermal domain. On the other hand, our work assumes only the presence of thermal imagery and seeks to leverage the use of publicly available RGB datasets (which may not be paired with the thermal dataset) to obtain significant improvement in thermal object detection performance. | {
"cite_N": [
"@cite_28",
"@cite_9",
"@cite_29",
"@cite_0"
],
"mid": [
"2608096492",
"2902314041",
"2549063375",
"2741620214"
],
"abstract": [
"This paper presents a method for detecting a pedestrian by leveraging multi-spectral image pairs. Our approach is based on the observation that a multi-spectral image, especially far-infrared (FIR) image, enables us to overcome inherent limitations for pedestrian detection under challenging circumstances, such as even dark environments. For that task, multi-spectral color-FIR image pairs are used in a synergistic manner for pedestrian detection through deep convolutional neural networks (CNNs) learning and support vector regression (SVR). For inferring the confidence of a pedestrian, we first learn CNNs between color images (or FIR images) and bounding box annotations of pedestrians, respectively. Furthermore, for each object proposal, we extract intermediate activation features from network, and learn the probability of pedestrian using SVR. To improve the detection performance, the learned probability of pedestrian for each proposal is accumulated on the image domain. Based on the pedestrian confidence estimated from each network and accumulated pedestrian probabilities, the most probable pedestrian is finally localized among object proposal candidates. Thanks to its high robustness of multi-spectral imaging in dark environments and its high discriminative power of deep CNNs, our framework is shown to surpass state-of-the-art pedestrian detection methods on multi-spectral pedestrian benchmark.",
"",
"Multispectral pedestrian detection is essential for around-the-clock applications, e.g., surveillance and autonomous driving. We deeply analyze Faster R-CNN for multispectral pedestrian detection task and then model it into a convolutional network (ConvNet) fusion problem. Further, we discover that ConvNet-based pedestrian detectors trained by color or thermal images separately provide complementary information in discriminating human instances. Thus there is a large potential to improve pedestrian detection by using color and thermal images in DNNs simultaneously. We carefully design four ConvNet fusion architectures that integrate two-branch ConvNets on different DNNs stages, all of which yield better performance compared with the baseline detector. Our experimental results on KAIST pedestrian benchmark show that the Halfway Fusion model that performs fusion on the middle-level convolutional features outperforms the baseline method by 11 and yields a missing rate 3.5 lower than the other proposed architectures.",
"Multispectral images that combine visual-optical (VIS) and infrared (IR) image information are a promising source of data for automatic person detection. Especially in automotive or surveillance applications, challenging conditions such as insufficient illumination or large distances between camera and object occur regularly and can affect image quality. This leads to weak image contrast or low object resolution. In order to detect persons under such conditions, we apply deep learning for effectively fusing the VIS and IR information in multispectral images. We present a novel multispectral Region Proposal Network (RPN) that is built up on the pre-trained very deep convolutional network VGG-16. The proposals of this network are further evaluated using a Boosted Decision Trees classifier in order to reduce potential false positive detections. With a log-average miss rate of 29:83 on the reasonable test set of the KAIST Multispectral Pedestrian Detection Benchmark, we improve the current state-of-the-art by about 18 ."
]
} |
1905.08838 | 2945236379 | Models for predicting the time of a future event are crucial for risk assessment, across a diverse range of applications. Existing time-to-event (survival) models have focused primarily on preserving pairwise ordering of estimated event times, or relative risk. Model calibration is relatively under explored, despite its critical importance in time-to-event applications. We present a survival function estimator for probabilistic predictions in time-to-event models, based on a neural network model for draws from the distribution of event times, without explicit assumptions on the form of the distribution. This is done like in adversarial learning, but we achieve learning without a discriminator or adversarial objective. The proposed estimator can be used in practice as a means of estimating and comparing conditional survival distributions, while accounting for the predictive uncertainty of probabilistic models. Extensive experiments show that the proposed model outperforms existing approaches, trained both with and without adversarial learning, in terms of both calibration and concentration of time-to-event distributions. | Existing calibration literature in predictive models has primarily focused on recalibration techniques for predictions from classification or regression models . For classification tasks, the Brier score is a commonly used proper score metric, quantifying the accuracy of probabilistic predictions, and thus it is often used to assess calibration. The Brier score has also been used to asses calibration in time-to-event models @cite_0 @cite_14 @cite_2 @cite_11 , however, this score has to be evaluated at pre-specified (thresholded) time horizons. Alternatively, S-CRPS @cite_12 considers the integral of the Brier score evaluated at all possible thresholds @cite_13 , which is a more principled and comprehensive approach than calibration at pre-specifying time horizon thresholds. | {
"cite_N": [
"@cite_14",
"@cite_0",
"@cite_2",
"@cite_13",
"@cite_12",
"@cite_11"
],
"mid": [
"2914771173",
"2726304173",
"2891794453",
"2025720061",
"2809313038",
"2921311865"
],
"abstract": [
"",
"Censoring is a common phenomenon that arises in many longitudinal studies where an event of interest could not be recorded within the given time frame. Censoring causes missing time-to-event labels, and this effect is compounded when dealing with datasets which have high amounts of censored instances. In addition, dependent censoring in the data, where censoring is dependent on the covariates in the data leads to bias in standard survival estimators. This motivates us to develop an approach for pre-processing censored data which calibrates the right censored (RC) times in an attempt to reduce the bias in the survival estimators. This calibration is done using an imputation method which estimates the sparse inverse covariance matrix over the dataset in an iterative convergence framework. During estimation, we apply row and column-based regularization to account for both row and column-wise correlations between different instances while imputing them. This is followed by comparing these imputed censored times with the original RC times to obtain the final calibrated RC times. These calibrated RC times can now be used in the survival dataset in place of the original RC times for more effective prediction. One of the major benefits of our calibration approach is that it is a pre-processing method for censored data which can be used in conjunction with any survival prediction algorithm and improve its performance. We evaluate the goodness of our approach using a wide array of survival prediction algorithms which are applied over crowdfunding data, electronic health records (EHRs), and synthetic censored datasets. Experimental results indicate that our calibration method improves the AUC values of survival prediction algorithms, compared to applying them directly on the original survival data.",
"We propose Lomax delegate racing (LDR) to explicitly model the mechanism of survival under competing risks and to interpret how the covariates accelerate or decelerate the time to event. LDR explains non-monotonic covariate effects by racing a potentially infinite number of sub-risks, and consequently relaxes the ubiquitous proportional-hazards assumption which may be too restrictive. Moreover, LDR is naturally able to model censoring and missing event times or types. For inference, we develop a Gibbs sampler under data augmentation for moderately sized data, along with a stochastic gradient descent maximum a posteriori inference algorithm for big data applications. Illustrative experiments are provided on both synthetic and real datasets, and comparison with various benchmark algorithms for survival analysis with competing risks demonstrates distinguished performance of LDR.",
"Scoring rules assess the quality of probabilistic forecasts, by assigning a numerical score based on the predictive distribution and on the event or value that materializes. A scoring rule is proper if the forecaster maximizes the expected score for an observation drawn from the distributionF if he or she issues the probabilistic forecast F, rather than G ≠ F. It is strictly proper if the maximum is unique. In prediction problems, proper scoring rules encourage the forecaster to make careful assessments and to be honest. In estimation problems, strictly proper scoring rules provide attractive loss and utility functions that can be tailored to the problem at hand. This article reviews and develops the theory of proper scoring rules on general probability spaces, and proposes and discusses examples thereof. Proper scoring rules derive from convex functions and relate to information measures, entropy functions, and Bregman divergences. In the case of categorical variables, we prove a rigorous version of the ...",
"Personalized probabilistic forecasts of time to event (such as mortality) can be crucial in decision making, especially in the clinical setting. Inspired by ideas from the meteorology literature, we approach this problem through the paradigm of maximizing sharpness of prediction distributions, subject to calibration. In regression problems, it has been shown that optimizing the continuous ranked probability score (CRPS) instead of maximum likelihood leads to sharper prediction distributions while maintaining calibration. We introduce the Survival-CRPS, a generalization of the CRPS to the time to event setting, and present right-censored and interval-censored variants. To holistically evaluate the quality of predicted distributions over time to event, we present the Survival-AUPRC evaluation metric, an analog to area under the precision-recall curve. We apply these ideas by building a recurrent neural network for mortality prediction, using an Electronic Health Record dataset covering millions of patients. We demonstrate significant benefits in models trained by the Survival-CRPS objective instead of maximum likelihood.",
"The importance of survival analysis in many disciplines (especially in medicine) has led to the development of a variety of approaches to modeling the survival function. Models constructed via various approaches o er di erent strengths and weaknesses in terms of discriminative performance and calibration, but no one model is best across all datasets or even across all time horizons within a single dataset. Because we require both good calibration and good discriminative performance over di erent time horizons, conventional model selection and ensemble approaches are not applicable. This paper develops a novel approach that combines the collective intelligence of different underlying survival models to produce a valid survival function that is well-calibrated and o ers superior discriminative performance at di erent time horizons. Empirical results show that our approach provides signi cant gains over the benchmarks on a variety of real-world datasets."
]
} |
1905.08838 | 2945236379 | Models for predicting the time of a future event are crucial for risk assessment, across a diverse range of applications. Existing time-to-event (survival) models have focused primarily on preserving pairwise ordering of estimated event times, or relative risk. Model calibration is relatively under explored, despite its critical importance in time-to-event applications. We present a survival function estimator for probabilistic predictions in time-to-event models, based on a neural network model for draws from the distribution of event times, without explicit assumptions on the form of the distribution. This is done like in adversarial learning, but we achieve learning without a discriminator or adversarial objective. The proposed estimator can be used in practice as a means of estimating and comparing conditional survival distributions, while accounting for the predictive uncertainty of probabilistic models. Extensive experiments show that the proposed model outperforms existing approaches, trained both with and without adversarial learning, in terms of both calibration and concentration of time-to-event distributions. | The approach presented here is inspired by @cite_12 . They considered calibration slope as a metric for evaluating performance in time-to-event models. However, our formulation is very different from that of @cite_12 , in the sense that they encourage calibration by optimizing a proper score rule, the Continuous Ranked Probability Score (CRPS), whereas we tackle it directly as a survival-distribution-matching problem. In the experiments in , we show empirically that our more direct approach to calibration consistently outperforms CRPS. | {
"cite_N": [
"@cite_12"
],
"mid": [
"2809313038"
],
"abstract": [
"Personalized probabilistic forecasts of time to event (such as mortality) can be crucial in decision making, especially in the clinical setting. Inspired by ideas from the meteorology literature, we approach this problem through the paradigm of maximizing sharpness of prediction distributions, subject to calibration. In regression problems, it has been shown that optimizing the continuous ranked probability score (CRPS) instead of maximum likelihood leads to sharper prediction distributions while maintaining calibration. We introduce the Survival-CRPS, a generalization of the CRPS to the time to event setting, and present right-censored and interval-censored variants. To holistically evaluate the quality of predicted distributions over time to event, we present the Survival-AUPRC evaluation metric, an analog to area under the precision-recall curve. We apply these ideas by building a recurrent neural network for mortality prediction, using an Electronic Health Record dataset covering millions of patients. We demonstrate significant benefits in models trained by the Survival-CRPS objective instead of maximum likelihood."
]
} |
1905.08838 | 2945236379 | Models for predicting the time of a future event are crucial for risk assessment, across a diverse range of applications. Existing time-to-event (survival) models have focused primarily on preserving pairwise ordering of estimated event times, or relative risk. Model calibration is relatively under explored, despite its critical importance in time-to-event applications. We present a survival function estimator for probabilistic predictions in time-to-event models, based on a neural network model for draws from the distribution of event times, without explicit assumptions on the form of the distribution. This is done like in adversarial learning, but we achieve learning without a discriminator or adversarial objective. The proposed estimator can be used in practice as a means of estimating and comparing conditional survival distributions, while accounting for the predictive uncertainty of probabilistic models. Extensive experiments show that the proposed model outperforms existing approaches, trained both with and without adversarial learning, in terms of both calibration and concentration of time-to-event distributions. | Our work is related to other recently proposed time-to-event approaches, including Survival CRPS (S-CRPS) , that uses an AFT (log-normal distribution) specification; the approach [DATE,][] chapfuwa2018adversarial ; the AFT-based DRAFT model ; the Weibull-based Deep Survival Analysis (DSA) ; nonparametric DSA based on ; and Gaussian-process-based models . Interestingly, excluding approaches that address calibration with Brier-scores @cite_0 @cite_14 @cite_2 @cite_11 , only S-CRPS considers calibration as a performance metric. All the others focus on accuracy-centric performance estimates, , C-Index and relative absolute error. | {
"cite_N": [
"@cite_0",
"@cite_14",
"@cite_11",
"@cite_2"
],
"mid": [
"2726304173",
"2914771173",
"2921311865",
"2891794453"
],
"abstract": [
"Censoring is a common phenomenon that arises in many longitudinal studies where an event of interest could not be recorded within the given time frame. Censoring causes missing time-to-event labels, and this effect is compounded when dealing with datasets which have high amounts of censored instances. In addition, dependent censoring in the data, where censoring is dependent on the covariates in the data leads to bias in standard survival estimators. This motivates us to develop an approach for pre-processing censored data which calibrates the right censored (RC) times in an attempt to reduce the bias in the survival estimators. This calibration is done using an imputation method which estimates the sparse inverse covariance matrix over the dataset in an iterative convergence framework. During estimation, we apply row and column-based regularization to account for both row and column-wise correlations between different instances while imputing them. This is followed by comparing these imputed censored times with the original RC times to obtain the final calibrated RC times. These calibrated RC times can now be used in the survival dataset in place of the original RC times for more effective prediction. One of the major benefits of our calibration approach is that it is a pre-processing method for censored data which can be used in conjunction with any survival prediction algorithm and improve its performance. We evaluate the goodness of our approach using a wide array of survival prediction algorithms which are applied over crowdfunding data, electronic health records (EHRs), and synthetic censored datasets. Experimental results indicate that our calibration method improves the AUC values of survival prediction algorithms, compared to applying them directly on the original survival data.",
"",
"The importance of survival analysis in many disciplines (especially in medicine) has led to the development of a variety of approaches to modeling the survival function. Models constructed via various approaches o er di erent strengths and weaknesses in terms of discriminative performance and calibration, but no one model is best across all datasets or even across all time horizons within a single dataset. Because we require both good calibration and good discriminative performance over di erent time horizons, conventional model selection and ensemble approaches are not applicable. This paper develops a novel approach that combines the collective intelligence of different underlying survival models to produce a valid survival function that is well-calibrated and o ers superior discriminative performance at di erent time horizons. Empirical results show that our approach provides signi cant gains over the benchmarks on a variety of real-world datasets.",
"We propose Lomax delegate racing (LDR) to explicitly model the mechanism of survival under competing risks and to interpret how the covariates accelerate or decelerate the time to event. LDR explains non-monotonic covariate effects by racing a potentially infinite number of sub-risks, and consequently relaxes the ubiquitous proportional-hazards assumption which may be too restrictive. Moreover, LDR is naturally able to model censoring and missing event times or types. For inference, we develop a Gibbs sampler under data augmentation for moderately sized data, along with a stochastic gradient descent maximum a posteriori inference algorithm for big data applications. Illustrative experiments are provided on both synthetic and real datasets, and comparison with various benchmark algorithms for survival analysis with competing risks demonstrates distinguished performance of LDR."
]
} |
1905.08910 | 2946339137 | We follow the idea of formulating vision as inverse graphics and propose a new type of element for this task, a neural-symbolic capsule. It is capable of de-rendering a scene into semantic information feed-forward, as well as rendering it feed-backward. An initial set of capsules for graphical primitives is obtained from a generative grammar and connected into a full capsule network. Lifelong meta-learning continuously improves this network's detection capabilities by adding capsules for new and more complex objects it detects in a scene using few-shot learning. Preliminary results demonstrate the potential of our novel approach. | In @cite_29 Hinton al introduced capsules, extending the idea of classical neurons by allowing them to output vectors instead of scalars. These vectors can be interpreted as attributes of an object and aim to reduce information-loss between layers of convolutional neural networks (CNN) @cite_18 @cite_22 . Capsules require specialized routing protocols, such as routing-by-agreement @cite_25 , where the activation probability of a capsule is dependent on the agreement of its real inputs with their expected values. Further extensions for capsules have been proposed, such as using matrices internally @cite_21 or using 3D input @cite_5 . | {
"cite_N": [
"@cite_18",
"@cite_22",
"@cite_29",
"@cite_21",
"@cite_5",
"@cite_25"
],
"mid": [
"2163605009",
"2310919327",
"1989789052",
"2785994986",
"2964228567",
"2963703618"
],
"abstract": [
"We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5 and 17.0 which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called \"dropout\" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3 , compared to 26.2 achieved by the second-best entry.",
"",
"Abstract Hybrid learning methods use theoretical knowledge of a domain and a set of classified examples to develop a method for accurately classifying examples not seen during training. The challenge of hybrid learning systems is to use the information provided by one source of information to offset information missing from the other source. By so doing, a hybrid learning system should learn more effectively than systems that use only one of the information sources. KBANN ( Knowledge-Based Artificial Neural Networks ) is a hybrid learning system built on top of connectionist learning techniques. It maps problem-specific “domain theories”, represented in propositional logic, into neural networks and then refines this reformulated knowledge using backpropagation. KBANN is evaluated by extensive empirical tests on two problems from molecular biology. Among other results, these tests show that the networks created by KBANN generalize better than a wide variety of learning systems, as well as several techniques proposed by biologists.",
"A capsule is a group of neurons whose outputs represent different properties of the same entity. Each layer in a capsule network contains many capsules [a group of capsules forms a capsule layer and can be used in place of a traditional layer in a neural net]. We describe a version of capsules in which each capsule has a logistic unit to represent the presence of an entity and a 4x4 matrix which could learn to represent the relationship between that entity and the viewer (the pose). A capsule in one layer votes for the pose matrix of many different capsules in the layer above by multiplying its own pose matrix by trainable viewpoint-invariant transformation matrices that could learn to represent part-whole relationships. Each of these votes is weighted by an assignment coefficient. These coefficients are iteratively updated for each image using the Expectation-Maximization algorithm such that the output of each capsule is routed to a capsule in the layer above that receives a cluster of similar votes. The transformation matrices are trained discriminatively by backpropagating through the unrolled iterations of EM between each pair of adjacent capsule layers. On the smallNORB benchmark, capsules reduce the number of test errors by 45 compared to the state-of-the-art. Capsules also show far more resistance to white box adversarial attack than our baseline convolutional neural network.",
"",
"A capsule is a group of neurons whose activity vector represents the instantiation parameters of a specific type of entity such as an object or object part. We use the length of the activity vector to represent the probability that the entity exists and its orientation to represent the instantiation parameters. Active capsules at one level make predictions, via transformation matrices, for the instantiation parameters of higher-level capsules. When multiple predictions agree, a higher level capsule becomes active. We show that a discrimininatively trained, multi-layer capsule system achieves state-of-the-art performance on MNIST and is considerably better than a convolutional net at recognizing highly overlapping digits. To achieve these results we use an iterative routing-by-agreement mechanism: A lower-level capsule prefers to send its output to higher level capsules whose activity vectors have a big scalar product with the prediction coming from the lower-level capsule."
]
} |
1905.08910 | 2946339137 | We follow the idea of formulating vision as inverse graphics and propose a new type of element for this task, a neural-symbolic capsule. It is capable of de-rendering a scene into semantic information feed-forward, as well as rendering it feed-backward. An initial set of capsules for graphical primitives is obtained from a generative grammar and connected into a full capsule network. Lifelong meta-learning continuously improves this network's detection capabilities by adding capsules for new and more complex objects it detects in a scene using few-shot learning. Preliminary results demonstrate the potential of our novel approach. | There has been a strong effort to make a purely neural approach to vision more interpretable @cite_27 @cite_28 @cite_9 @cite_10 @cite_13 @cite_24 . An alternative approach to interpretability has been to deeply intertwine symbolic methods into connectionist methods. For computer vision, this is proving fruitful for de-rendering and scene decomposition @cite_2 @cite_32 @cite_8 @cite_6 @cite_23 @cite_14 @cite_0 . Many of these approaches use shape programs, a decomposition of the scene into a set of rendering instructions. The scene-graphs we construct can be viewed as such shape programs, but in a different representation. This symbolic information is well suited for more complex tasks, such as visual question answering (VQA) @cite_20 @cite_3 and scene manipulation @cite_30 . We follow many of the presented ideas in our work. | {
"cite_N": [
"@cite_13",
"@cite_30",
"@cite_14",
"@cite_8",
"@cite_28",
"@cite_9",
"@cite_32",
"@cite_6",
"@cite_3",
"@cite_24",
"@cite_0",
"@cite_27",
"@cite_23",
"@cite_2",
"@cite_10",
"@cite_20"
],
"mid": [
"2962851944",
"2963706553",
"",
"",
"1915485278",
"2657631929",
"2912464212",
"2963375929",
"2963738360",
"2764024122",
"2963369474",
"2439568532",
"2557269700",
"1691728462",
"2282821441",
"2908791737"
],
"abstract": [
"This paper addresses the visualisation of image classification models, learnt using deep Convolutional Networks (ConvNets). We consider two visualisation techniques, based on computing the gradient of the class score with respect to the input image. The first one generates an image, which maximises the class score [5], thus visualising the notion of the class, captured by a ConvNet. The second technique computes a class saliency map, specific to a given image and class. We show that such maps can be employed for weakly supervised object segmentation using classification ConvNets. Finally, we establish the connection between the gradient-based ConvNet visualisation methods and deconvolutional networks [13].",
"We aim to obtain an interpretable, expressive, and disentangled scene representation that contains comprehensive structural and textural information for each object. Previous representations learned by neural networks are often uninterpretable, limited to a single object, or lacking 3D knowledge. In this work, we propose 3D scene de-rendering networks (3D-SDN) to address the above issues by integrating disentangled representations for object semantics, appearance, and geometry into a deep generative model. Our scene encoder performs inverse graphics, translating a scene into a structured object representation. Our decoder has two components: a differentiable shape renderer and a neural texture generator. The disentanglement of semantics, appearance, and geometry supports various 3D-aware scene manipulation applications, , rotating and moving objects freely while maintaining consistent shape and texture, and changing the object appearance without affecting its shape. We systematically evaluate our model and demonstrate that our editing scheme based on 3D-SDN is superior to its 2D counterpart.",
"",
"",
"Image representations, from SIFT and Bag of Visual Words to Convolutional Neural Networks (CNNs), are a crucial component of almost any image understanding system. Nevertheless, our understanding of them remains limited. In this paper we conduct a direct analysis of the visual information contained in representations by asking the following question: given an encoding of an image, to which extent is it possible to reconstruct the image itself? To answer this question we contribute a general framework to invert representations. We show that this method can invert representations such as HOG more accurately than recent alternatives while being applicable to CNNs too. We then use this technique to study the inverse of recent state-of-the-art CNN image representations for the first time. Among our findings, we show that several layers in CNNs retain photographically accurate information about the image, with different degrees of geometric and photometric invariance.",
"Abstract This paper provides an entry point to the problem of interpreting a deep neural network model and explaining its predictions. It is based on a tutorial given at ICASSP 2017. As a tutorial paper, the set of methods covered here is not exhaustive, but sufficiently representative to discuss a number of questions in interpretability, technical challenges, and possible applications. The second part of the tutorial focuses on the recently proposed layer-wise relevance propagation (LRP) technique, for which we provide theory, recommendations, and tricks, to make most efficient use of it on real data.",
"",
"",
"We marry two powerful ideas: deep representation learning for visual recognition and language understanding, and symbolic program execution for reasoning. Our visual question answering (VQA) system first recovers a structural scene representation from the image and a program trace from the question. It then executes the program on the scene representation to obtain an answer. Incorporating symbolic structure as prior knowledge offers three advantages. First, executing programs on a symbolic space is more robust to long program traces. Our model can solve complex reasoning tasks better, achieving an accuracy of 99.8 on the CLEVR dataset. Second, the model is more data- and memory-efficient: it learns to perform well on a small number of training data; it can also encode an image into a compact representation and answer questions offline, using only 1 of the storage needed by the best competing methods. Third, symbolic program execution offers full transparency to the reasoning process; we are thus able to interpret and diagnose each execution step. Our model recovers the ground truth programs precisely.",
"This paper proposes a method to modify a traditional convolutional neural network (CNN) into an interpretable CNN, in order to clarify knowledge representations in high conv-layers of the CNN. In an interpretable CNN, each filter in a high conv-layer represents a specific object part. Our interpretable CNNs use the same training data as ordinary CNNs without a need for any annotations of object parts or textures for supervision. The interpretable CNN automatically assigns each filter in a high conv-layer with an object part during the learning process. We can apply our method to different types of CNNs with various structures. The explicit knowledge representation in an interpretable CNN can help people understand the logic inside a CNN, i.e. what patterns are memorized by the CNN for prediction. Experiments have shown that filters in an interpretable CNN are more semantically meaningful than those in a traditional CNN. The code is available at https: github.com zqs1022 interpretableCNN.",
"The success of various applications including robotics, digital content creation, and visualization demand a structured and abstract representation of the 3D world from limited sensor data. Inspired by the nature of human perception of 3D shapes as a collection of simple parts, we explore such an abstract shape representation based on primitives. Given a single depth image of an object, we present 3DPRNN, a generative recurrent neural network that synthesizes multiple plausible shapes composed of a set of primitives. Our generative model encodes symmetry characteristics of common man-made objects, preserves long-range structural coherence, and describes objects of varying complexity with a compact representation. We also propose a method based on Gaussian Fields to generate a large scale dataset of primitive-based shape representations to train our network. We evaluate our approach on a wide range of examples and show that it outperforms nearest-neighbor based shape retrieval methods and is on-par with voxelbased generative models while using a significantly reduced parameter space.",
"Supervised machine learning models boast remarkable predictive capabilities. But can you trust your model? Will it work in deployment? What else can it tell you about the world? We want models to be not only good, but interpretable. And yet the task of interpretation appears underspecified. Papers provide diverse and sometimes non-overlapping motivations for interpretability, and offer myriad notions of what attributes render models interpretable. Despite this ambiguity, many papers proclaim interpretability axiomatically, absent further explanation. In this paper, we seek to refine the discourse on interpretability. First, we examine the motivations underlying interest in interpretability, finding them to be diverse and occasionally discordant. Then, we address model properties and techniques thought to confer interpretability, identifying transparency to humans and post-hoc explanations as competing notions. Throughout, we discuss the feasibility and desirability of different notions, and question the oft-made assertions that linear models are interpretable and that deep neural networks are not.",
"We present a learning framework for abstracting complex shapes by learning to assemble objects using 3D volumetric primitives. In addition to generating simple and geometrically interpretable explanations of 3D objects, our framework also allows us to automatically discover and exploit consistent structure in the data. We demonstrate that using our method allows predicting shape representations which can be leveraged for obtaining a consistent parsing across the instances of a shape collection and constructing an interpretable shape similarity measure. We also examine applications for image-based prediction as well as shape manipulation.",
"This paper presents the Deep Convolution Inverse Graphics Network (DC-IGN), a model that aims to learn an interpretable representation of images, disentangled with respect to three-dimensional scene structure and viewing transformations such as depth rotations and lighting variations. The DC-IGN model is composed of multiple layers of convolution and de-convolution operators and is trained using the Stochastic Gradient Variational Bayes (SGVB) algorithm [10]. We propose a training procedure to encourage neurons in the graphics code layer to represent a specific transformation (e.g. pose or light). Given a single input image, our model can generate new images of the same object with variations in pose and lighting. We present qualitative and quantitative tests of the model's efficacy at learning a 3D rendering engine for varied object classes including faces and chairs.",
"Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally varound the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted.",
""
]
} |
1905.08932 | 2945630033 | Edge and fog computing have grown popular as IoT deployments become wide-spread. While application composition and scheduling on such resources are being explored, there exists a gap in a distributed data storage service on the edge and fog layer, instead depending solely on the cloud for data persistence. Such a service should reliably store and manage data on fog and edge devices, even in the presence of failures, and offer transparent discovery and access to data for use by edge computing applications. Here, we present Elfstore, a first-of-its-kind edge-local federated store for streams of data blocks. It uses reliable fog devices as a super-peer overlay to monitor the edge resources, offers federated metadata indexing using Bloom filters, locates data within 2-hops, and maintains approximate global statistics about the reliability and storage capacity of edges. Edges host the actual data blocks, and we use a unique differential replication scheme to select edges on which to replicate blocks, to guarantee a minimum reliability and to balance storage utilization. Our experiments on two IoT virtual deployments with 20 and 272 devices show that ElfStore has low overheads, is bound only by the network bandwidth, has scalable performance, and offers tunable resilience. | @cite_14 proposes a distributed key-value store on fog resources with replication and differential consistency. Our focus is on reliably storing a stream of blocks of a much larger size, where resilience and capacity constraints are met. Others @cite_25 propose repositories hosted on stable fogs (referred to as edges'') that are populated by data from transient edges ( mobile devices''), and act as a reverse-Content Distribution Network (CDN) to serve requests from the cloud too. Reliability is a non-goal in their design and no experiments are presented. @cite_17 supports context-aware placement of data on fog and cloud resources, with mobile devices generating and consuming these data. It uses a rules engine to place and locate data based on its context metadata, but ignores reliability as edge devices do not store data. | {
"cite_N": [
"@cite_14",
"@cite_25",
"@cite_17"
],
"mid": [
"2964228733",
"2885461866",
"2785267797"
],
"abstract": [
"Stateful applications and virtualized network functions (VNFs) can benefit from state externalization to increase their reliability, scalability, and inter-operability. To keep and share the externalized state, distributed data stores (DDSs) are a powerful tool allowing for the management of classical trade-offs in consistency, availability and partitioning tolerance. With the advent of Fog and Edge Computing, stateful applications and VNFs are pushed from the data centers toward the network edge. This poses new challenges on DDSs that are tailored to a deployment in Cloud data centers. In this paper, we propose two novel design goals for DDSs that are tailored to Fog Computing: (1) Fog-aware replica placement, and (2) context-sensitive differential consistency. To realize those design goals on top of existing DDSs, we propose the FogStore system. FogStore manages the needed adaptations in replica placement and consistency management transparently, so that existing DDSs can be plugged into the system. To show the benefits of FogStore, we perform a set of evaluations using the Yahoo Cloud Serving Benchmark.",
"In a future IoT-dominated environment the majority of data will be produced at the edge, which may be moved to the network core. We argue that this reverses today’s “core-to-edge” data flow to an “edge-to-core” model and puts severe stress on edge access cellular links. In this paper, we propose a data-centric communication approach which treats storage and wire the same as far as their ability to supply the requested data is concerned. Given that storage is cheaper to provide and scales better than wires, we argue for enhancing network connectivity with local storage services (e.g., in WiFi Access Points, or similar) at the edge of the network. Such local storage services can be used to buffer IoT and user-generated data at the edge, prior to data-cloud synchronization.",
"The way mobile users store and share their data today is completely decoupled from their current usage context and actual intentions. Furthermore, the paradigm of cloud computing, where all data is placed in distant cloud data centers is seldom questioned. As a result, we are faced with congested networks and high latencies when retrieving data stored at distant locations. The emergence of edge computing provides an opportunity to overcome this issue. In this paper, we present vStore, a framework that provides the capabilities for context-aware micro-storage. The framework is targeted at mobile users and leverages small-scale, decentralized storage nodes at the extreme edge of the network. The decision where to store data is made based on rules that can either be pushed globally to the framework or created individually by users. We motivate our approach with different use cases, one of which is the sharing of data at events where cellular networks tend to be congested. To demonstrate the feasibility of our approach, we implement a demo application on the Android platform, leveraging storage nodes placed at different locations in a major city. By conducting a field trial, we demonstrate the key functionalities of vStore and report on first usage insights."
]
} |
1905.08932 | 2945630033 | Edge and fog computing have grown popular as IoT deployments become wide-spread. While application composition and scheduling on such resources are being explored, there exists a gap in a distributed data storage service on the edge and fog layer, instead depending solely on the cloud for data persistence. Such a service should reliably store and manage data on fog and edge devices, even in the presence of failures, and offer transparent discovery and access to data for use by edge computing applications. Here, we present Elfstore, a first-of-its-kind edge-local federated store for streams of data blocks. It uses reliable fog devices as a super-peer overlay to monitor the edge resources, offers federated metadata indexing using Bloom filters, locates data within 2-hops, and maintains approximate global statistics about the reliability and storage capacity of edges. Edges host the actual data blocks, and we use a unique differential replication scheme to select edges on which to replicate blocks, to guarantee a minimum reliability and to balance storage utilization. Our experiments on two IoT virtual deployments with 20 and 272 devices show that ElfStore has low overheads, is bound only by the network bandwidth, has scalable performance, and offers tunable resilience. | @cite_0 is a distributed file system hosted on the cloud but optimized for mobile clients (edges) with transient network connectivity. While the cloud holds the encrypted master data, clients selectively pre-fetch, decrypt and cache parts of the file based on their access patterns. Clients have exclusive access to their encrypted home directory, and common access to shared directories. The master data in the cloud is reliable. | {
"cite_N": [
"@cite_0"
],
"mid": [
"2071051316"
],
"abstract": [
"Due to the increasing number of applications (and their data) being placed on mobile devices, access to dependable storage is becoming a key issue in mobile system design -- and cloud storage is becoming an attractive solution. However, this introduces a number of new issues related to unpredictable wireless network connectivity and data privacy over the network. In this article we present RFS, a wireless-friendly network file system for mobile devices and the cloud. RFS provides deviceaware cache management and client-driven data security and privacy protection. We implement the RFS client in the Linux kernel and the RFS server with Amazon S3 cloud storage, and we employ two new optimizations: server prepush (a server-side data pre-fetching mechanism) and client reintegration (synchronizing a mobile device's cache with the cloud). The empirical results over wired, WiFi and 3G networks show that RFS achieves good performance compared to Coda and FScache, and it reduces network activity visibly. Further, the privacy overhead is acceptable when RFS is run over wireless networks. We present a case study of booting Android over RFS, thereby demonstrating the ability for RFS to host a full mobile system. Overall, RFS can deliver a good user experience under undependable network conditions, allowing mobile users to seamlessly, and safely, use the cloud for data storage."
]
} |
1905.08932 | 2945630033 | Edge and fog computing have grown popular as IoT deployments become wide-spread. While application composition and scheduling on such resources are being explored, there exists a gap in a distributed data storage service on the edge and fog layer, instead depending solely on the cloud for data persistence. Such a service should reliably store and manage data on fog and edge devices, even in the presence of failures, and offer transparent discovery and access to data for use by edge computing applications. Here, we present Elfstore, a first-of-its-kind edge-local federated store for streams of data blocks. It uses reliable fog devices as a super-peer overlay to monitor the edge resources, offers federated metadata indexing using Bloom filters, locates data within 2-hops, and maintains approximate global statistics about the reliability and storage capacity of edges. Edges host the actual data blocks, and we use a unique differential replication scheme to select edges on which to replicate blocks, to guarantee a minimum reliability and to balance storage utilization. Our experiments on two IoT virtual deployments with 20 and 272 devices show that ElfStore has low overheads, is bound only by the network bandwidth, has scalable performance, and offers tunable resilience. | like Chord, Pastry and BitTorrent have proposed distributed file, block and key-value storage on unreliable peers on wide-area networks @cite_7 . We adopt several of these concepts such as super-peers @cite_24 , but simplify and enhance their performance for edge and fog deployments with less device flux, guarantee a minimum durability for stored blocks, and balance the storage capacity across peers. We also use an efficient federated indexing using Bloom filters @cite_12 . | {
"cite_N": [
"@cite_24",
"@cite_12",
"@cite_7"
],
"mid": [
"2104023883",
"",
"2118428193"
],
"abstract": [
"A super-peer is a node in a peer-to-peer network that operates both as a server to a set of clients, and as an equal in a network of super-peers. Super-peer networks strike a balance between the efficiency of centralized search, and the autonomy, load balancing and robustness to attacks provided by distributed search. Furthermore, they take advantage of the heterogeneity of capabilities (e.g., bandwidth, processing power) across peers, which recent studies have shown to be enormous. Hence, new and old P2P systems like KaZaA and Gnutella are adopting super-peers in their design. Despite their growing popularity, the behavior of super-peer networks is not well understood. For example, what are the potential drawbacks of super-peer networks? How can super-peers be made more reliable? How many clients should a super-peer take on to maximize efficiency? we examine super-peer networks in detail, gaming an understanding of their fundamental characteristics and performance tradeoffs. We also present practical guidelines and a general procedure for the design of an efficient super-peer network.",
"",
"A fundamental problem that confronts peer-to-peer applications is the efficient location of the node that stores a desired data item. This paper presents Chord, a distributed lookup protocol that addresses this problem. Chord provides support for just one operation: given a key, it maps the key onto a node. Data location can be easily implemented on top of Chord by associating a key with each data item, and storing the key data pair at the node to which the key maps. Chord adapts efficiently as nodes join and leave the system, and can answer queries even if the system is continuously changing. Results from theoretical analysis and simulations show that Chord is scalable: Communication cost and the state maintained by each node scale logarithmically with the number of Chord nodes."
]
} |
1905.09043 | 2945106964 | In this paper we address a classification problem that has not been considered before, namely motion segmentation given pairwise matches only. Our contribution to this unexplored task is a novel formulation of motion segmentation as a two-step process. First, motion segmentation is performed on image pairs independently. Secondly, we combine independent pairwise segmentation results in a robust way into the final globally consistent segmentation. Our approach is inspired by the success of averaging methods. We demonstrate in simulated as well as in real experiments that our method is very effective in reducing the errors in the pairwise motion segmentation and can cope with large number of mismatches. | The goal of is to cluster high-dimensional data drawn from multiple low-dimensional subspaces. Existing solutions include Generalized Principal Component Analysis (GPCA) @cite_37 , Local Subspace Affinity (LSA) @cite_49 , Power Factorization (PF) @cite_3 , Agglomerative Lossy Compression (ALC) @cite_27 , Low-Rank Representation (LRR) @cite_45 , Sparse Subspace Clustering (SSC) @cite_22 , Structured Sparse Subspace Clustering (S @math C) @cite_15 , and Robust Shape Interaction Matrix (RSIM) @cite_35 . Motion segmentation can be cast as subspace separation since -- under the affine camera model -- the point trajectories lie in the union of @math subspaces in @math of dimension at most 4, where @math denotes the number of motions and @math denotes the number of images. Subspace separation techniques can also be used to solve motion segmentation in two images under the perspective camera model, since corresponding points undergoing the same motion -- after a proper rearrangement of coordinates -- belong to a subspace of @math of dimension at most 8, as observed in @cite_24 . | {
"cite_N": [
"@cite_35",
"@cite_37",
"@cite_22",
"@cite_3",
"@cite_24",
"@cite_27",
"@cite_45",
"@cite_49",
"@cite_15"
],
"mid": [
"2114554887",
"2177347332",
"1993962865",
"2056639756",
"2082712442",
"2150414161",
"1997201895",
"1950520880",
"1931947167"
],
"abstract": [
"The Shape Interaction Matrix (SIM) is one of the earliest approaches to performing subspace clustering (i.e., separating points drawn from a union of subspaces). In this paper, we revisit the SIM and reveal its connections to several recent subspace clustering methods. Our analysis lets us derive a simple, yet effective algorithm to robustify the SIM and make it applicable to realistic scenarios where the data is corrupted by noise. We justify our method by intuitive examples and the matrix perturbation theory. We then show how this approach can be extended to handle missing data, thus yielding an efficient and general subspace clustering algorithm. We demonstrate the benefits of our approach over state-of-the-art subspace clustering methods on several challenging motion segmentation and face clustering problems, where the data includes corruptions and missing measurements.",
"This paper presents an algebro-geometric solution to the problem of segmenting an unknown number of subspaces of unknown and varying dimensions from sample data points. We represent the subspaces with a set of homogeneous polynomials whose degree is the number of subspaces and whose derivatives at a data point give normal vectors to the subspace passing through the point. When the number of subspaces is known, we show that these polynomials can be estimated linearly from data; hence, subspace segmentation is reduced to classifying one point per subspace. We select these points optimally from the data set by minimizing certain distance function, thus dealing automatically with moderate noise in the data. A basis for the complement of each subspace is then recovered by applying standard PCA to the collection of derivatives (normal vectors). Extensions of GPCA that deal with data in a high-dimensional space and with an unknown number of subspaces are also presented. Our experiments on low-dimensional data show that GPCA outperforms existing algebraic algorithms based on polynomial factorization and provides a good initialization to iterative techniques such as k-subspaces and expectation maximization. We also present applications of GPCA to computer vision problems such as face clustering, temporal video segmentation, and 3D motion segmentation from point correspondences in multiple affine views.",
"Many real-world problems deal with collections of high-dimensional data, such as images, videos, text, and web documents, DNA microarray data, and more. Often, such high-dimensional data lie close to low-dimensional structures corresponding to several classes or categories to which the data belong. In this paper, we propose and study an algorithm, called sparse subspace clustering, to cluster data points that lie in a union of low-dimensional subspaces. The key idea is that, among the infinitely many possible representations of a data point in terms of other points, a sparse representation corresponds to selecting a few points from the same subspace. This motivates solving a sparse optimization program whose solution is used in a spectral clustering framework to infer the clustering of the data into subspaces. Since solving the sparse optimization program is in general NP-hard, we consider a convex relaxation and show that, under appropriate conditions on the arrangement of the subspaces and the distribution of the data, the proposed minimization program succeeds in recovering the desired sparse representations. The proposed algorithm is efficient and can handle data points near the intersections of subspaces. Another key advantage of the proposed algorithm with respect to the state of the art is that it can deal directly with data nuisances, such as noise, sparse outlying entries, and missing entries, by incorporating the model of the data into the sparse optimization program. We demonstrate the effectiveness of the proposed algorithm through experiments on synthetic data as well as the two real-world problems of motion segmentation and face clustering.",
"We consider the problem of segmenting multiple rigid-body motions from point correspondences in multiple affine views. We cast this problem as a subspace clustering problem in which point trajectories associated with each motion live in a linear subspace of dimension two, three or four. Our algorithm involves projecting all point trajectories onto a 5-dimensional subspace using the SVD, the PowerFactorization method, or RANSAC, and fitting multiple linear subspaces representing different rigid-body motions to the points in ?5 using GPCA. Unlike previous work, our approach does not restrict the motion subspaces to be four-dimensional and independent. Instead, it deals gracefully with all the spectrum of possible affine motions: from two-dimensional and partially dependent to four-dimensional and fully independent. Our algorithm can handle the case of missing data, meaning that point tracks do not have to be visible in all images, by using the PowerFactorization method to project the data. In addition, our method can handle outlying trajectories by using RANSAC to perform the projection. We compare our approach to other methods on a database of 167 motion sequences with full motions, independent motions, degenerate motions, partially dependent motions, missing data, outliers, etc. On motion sequences with complete data our method achieves a misclassification error of less that 5 for two motions and 29 for three motions.",
"This paper addresses real-world challenges in the motion segmentation problem, including perspective effects, missing data, and unknown number of motions. It first formulates the 3-D motion segmentation from two perspective views as a subspace clustering problem, utilizing the epipolar constraint of an image pair. It then combines the point correspondence information across multiple image frames via a collaborative clustering step, in which tight integration is achieved via a mixed norm optimization scheme. For model selection, we propose an over-segment and merge approach, where the merging step is based on the property of the ell_1-norm of the mutual sparse representation of two over-segmented groups. The resulting algorithm can deal with incomplete trajectories and perspective effects substantially better than state-of-the-art two-frame and multi-frame methods. Experiments on a 62-clip dataset show the significant superiority of the proposed idea in both segmentation accuracy and model selection.",
"In this paper, we study the problem of segmenting tracked feature point trajectories of multiple moving objects in an image sequence. Using the affine camera model, this problem can be cast as the problem of segmenting samples drawn from multiple linear subspaces. In practice, due to limitations of the tracker, occlusions, and the presence of nonrigid objects in the scene, the obtained motion trajectories may contain grossly mistracked features, missing entries, or corrupted entries. In this paper, we develop a robust subspace separation scheme that deals with these practical issues in a unified mathematical framework. Our methods draw strong connections between lossy compression, rank minimization, and sparse representation. We test our methods extensively on the Hopkins155 motion segmentation database and other motion sequences with outliers and missing data. We compare the performance of our methods to state-of-the-art motion segmentation methods based on expectation-maximization and spectral clustering. For data without outliers or missing information, the results of our methods are on par with the state-of-the-art results and, in many cases, exceed them. In addition, our methods give surprisingly good performance in the presence of the three types of pathological trajectories mentioned above. All code and results are publicly available at http: perception.csl.uiuc.edu coding motion .",
"In this paper, we address the subspace clustering problem. Given a set of data samples (vectors) approximately drawn from a union of multiple subspaces, our goal is to cluster the samples into their respective subspaces and remove possible outliers as well. To this end, we propose a novel objective function named Low-Rank Representation (LRR), which seeks the lowest rank representation among all the candidates that can represent the data samples as linear combinations of the bases in a given dictionary. It is shown that the convex program associated with LRR solves the subspace clustering problem in the following sense: When the data is clean, we prove that LRR exactly recovers the true subspace structures; when the data are contaminated by outliers, we prove that under certain conditions LRR can exactly recover the row space of the original data and detect the outlier as well; for data corrupted by arbitrary sparse errors, LRR can also approximately recover the row space with theoretical guarantees. Since the subspace membership is provably determined by the row space, these further imply that LRR can perform robust subspace clustering and error correction in an efficient and effective way.",
"We cast the problem of motion segmentation of feature trajectories as linear manifold finding problems and propose a general framework for motion segmentation under affine projections which utilizes two properties of trajectory data: geometric constraint and locality. The geometric constraint states that the trajectories of the same motion lie in a low dimensional linear manifold and different motions result in different linear manifolds; locality, by which we mean in a transformed space a data and its neighbors tend to lie in the same linear manifold, provides a cue for efficient estimation of these manifolds. Our algorithm estimates a number of linear manifolds, whose dimensions are unknown beforehand, and segment the trajectories accordingly. It first transforms and normalizes the trajectories; secondly, for each trajectory it estimates a local linear manifold through local sampling; then it derives the affinity matrix based on principal subspace angles between these estimated linear manifolds; at last, spectral clustering is applied to the matrix and gives the segmentation result. Our algorithm is general without restriction on the number of linear manifolds and without prior knowledge of the dimensions of the linear manifolds. We demonstrate in our experiments that it can segment a wide range of motions including independent, articulated, rigid, non-rigid, degenerate, non-degenerate or any combination of them. In some highly challenging cases where other state-of-the-art motion segmentation algorithms may fail, our algorithm gives expected results.",
"Subspace clustering refers to the problem of segmenting data drawn from a union of subspaces. State of the art approaches for solving this problem follow a two-stage approach. In the first step, an affinity matrix is learned from the data using sparse or low-rank minimization techniques. In the second step, the segmentation is found by applying spectral clustering to this affinity. While this approach has led to state of the art results in many applications, it is sub-optimal because it does not exploit the fact that the affinity and the segmentation depend on each other. In this paper, we propose a unified optimization framework for learning both the affinity and the segmentation. Our framework is based on expressing each data point as a structured sparse linear combination of all other data points, where the structure is induced by a norm that depends on the unknown segmentation. We show that both the segmentation and the structured sparse representation can be found via a combination of an alternating direction method of multipliers with spectral clustering. Experiments on a synthetic data set, the Hopkins 155 motion segmentation database, and the Extended Yale B data set demonstrate the effectiveness of our approach."
]
} |
1905.09043 | 2945106964 | In this paper we address a classification problem that has not been considered before, namely motion segmentation given pairwise matches only. Our contribution to this unexplored task is a novel formulation of motion segmentation as a two-step process. First, motion segmentation is performed on image pairs independently. Secondly, we combine independent pairwise segmentation results in a robust way into the final globally consistent segmentation. Our approach is inspired by the success of averaging methods. We demonstrate in simulated as well as in real experiments that our method is very effective in reducing the errors in the pairwise motion segmentation and can cope with large number of mismatches. | The goal of is to simultaneously estimate the motion between each object and the camera as well as the 3D structure of each object, given a set of images of a dynamic scene. This problem can be seen as the generalization of structure from motion @cite_1 to the dynamic case, where motion segmentation has to be solved in addition to 3D reconstruction. Geometric solutions are available for two images @cite_60 and three images @cite_51 . Other techniques follow a statistical approach @cite_32 @cite_59 @cite_53 @cite_23 @cite_47 @cite_20 , whereas in @cite_39 @cite_5 @cite_8 @cite_9 motion segmentation and structure from motion are combined. More details can be found in survey @cite_6 . | {
"cite_N": [
"@cite_60",
"@cite_8",
"@cite_53",
"@cite_9",
"@cite_1",
"@cite_32",
"@cite_6",
"@cite_39",
"@cite_59",
"@cite_23",
"@cite_5",
"@cite_47",
"@cite_51",
"@cite_20"
],
"mid": [
"2115659783",
"2165926549",
"2004333984",
"2089695649",
"2963221299",
"2002875082",
"2788608285",
"2143920871",
"2117070904",
"2142501512",
"2118154608",
"2149884854",
"",
"1985172090"
],
"abstract": [
"We propose a geometric approach to 3D motion segmentation from point correspondences in three perspective views. We demonstrate that after applying a polynomial embedding to the point correspondences, they become related by the so-called multibody trilinear constraint and its associated multibody trifocal tensor, which are natural generalizations of the trilinear constraint and the trifocal tensor to multiple motions. We derive a rank constraint on the embedded correspondences from which one can estimate the number of independent motions, as well as linearly solve for the multibody trifocal tensor. We then show how to compute the epipolar lines associated with each image point from the common root of a set of univariate polynomials and the epipoles by solving a pair of plane clustering problems using Generalized Principal Component Analysis (GPCA). The individual trifocal tensors are then obtained from the second-order derivatives of the multibody trilinear constraint. Given epipolar lines and epipoles or trifocal tensors, one can immediately obtain an initial clustering of the correspondences. We use this clustering to initialize an iterative algorithm that alternates between the computation of the trifocal tensors and the segmentation of the correspondences. We test our algorithm on various synthetic and real scenes and compare it with other algebraic and iterative algorithms.",
"Given point correspondences in multiple perspective views of a scene containing multiple rigid-body motions, we present an algorithm for segmenting the correspondences according to the multiple motions. We exploit the fact that when the depths of the points are known, the point trajectories associated with a single motion live in a subspace of dimension at most four. Thus motion segmentation with known depths can be achieved by methods of subspace separation, such as GPCA or LSA. When the depths are unknown, we proceed iteratively. Given the segmentation, we compute the depths using standard techniques. Given the depths, we use GPCA or LSA to segment the scene into multiple motions. Experiments on the Hopkins 155 motion segmentation database show that our method compares favorably against existing affine motion segmentation methods in terms of segmentation error and execution time.",
"Given an image sequence of a scene consisting of multiple rigidly moving objects, multi-body structure-and-motion (MSaM) is the task to segment the image feature tracks into the different rigid objects and compute the multiple-view geometry of each object. We present a framework for multibody structure-and-motion based on model selection. In a recover-and-select procedure, a redundant set of hypothetical scene motions is generated. Each subset of this pool of motion candidates is regarded as a possible explanation of the image feature tracks, and the most likely explanation is selected with model selection. The framework is generic and can be used with any parametric camera model, or with a combination of different models. It can deal with sets of correspondences, which change over time, and it is robust to realistic amounts of outliers. The framework is demonstrated for different camera and scene models.",
"We present a novel optimisation framework for the estimation of the multi-body motion segmentation and 3D reconstruction of a set of point trajectories in the presence of missing data. The proposed solution not only assigns the trajectories to the correct motion but it also solves for the 3D location of multi-body shape and it fills the missing entries in the measurement matrix. Such a solution is based on two fundamental principles: each of the multi-body motions is controlled by a set of metric constraints that are given by the specific camera model, and the shape matrix that describes the multi-body 3D shape is generally sparse. We jointly include such constraints in a unique optimisation framework which, starting from an initial segmentation, iteratively enforces these set of constraints in three stages. First, metric constraints are used to estimate the 3D metric shape and to fill the missing entries according to an orthographic camera model. Then, wrongly segmented trajectories are detected by using sparse optimisation of the shape matrix. A final reclassification strategy assigns the detected points to the right motion or discards them as outliers. We provide experiments that show consistent improvements to previous approaches both on synthetic and real data.",
"",
"Motion segmentation involves clustering features together that belong to independently moving objects. The image features on each of these objects conform to one of several putative motion models, but the number and type of motion is unknown a priori. In order to cluster these features, the problems of model selection, robust estimation and clustering must all be addressed simultaneously. Within this paper we place the three problems into a common statistical framework; investigating the use of information criteria and robust mixture models as a principled way for motion segmentation of images. The final result is a general fully automatic algorithm for clustering that works in the presence of noise and outliers.",
"In the last few decades, Structure from Motion (SfM) and visual Simultaneous Localization and Mapping (visual SLAM) techniques have gained significant interest from both the computer vision and robotic communities. Many variants of these techniques have started to make an impact in a wide range of applications, including robot navigation and augmented reality. However, despite some remarkable results in these areas, most SfM and visual SLAM techniques operate based on the assumption that the observed environment is static. However, when faced with moving objects, overall system accuracy can be jeopardized. In this article, we present for the first time a survey of visual SLAM and SfM techniques that are targeted toward operation in dynamic environments. We identify three main problems: how to perform reconstruction (robust visual SLAM), how to segment and track dynamic objects, and how to achieve joint motion segmentation and reconstruction. Based on this categorization, we provide a comprehensive taxonomy of existing approaches. Finally, the advantages and disadvantages of each solution class are critically discussed from the perspective of practicality and robustness.",
"We want to deduce, from a sequence of noisy two-dimensional images of a scene of several rigid bodies moving independently in three dimensions, the number of bodies and the grouping of given feature points in the images to the bodies. Prior processing is assumed to have identified features or points common to all frames and the images are assumed to be created by orthographic projection (i.e., perspective effects are minimal). We describe a computationally inexpensive algorithm that can determine which points or features belong to which rigid body using the fact that, with exact observations in orthographic projection, points on a single body lie in a three or less dimensional linear manifold of frame space. If there are enough observations and independent motions, these manifolds can be viewed as a set linearly independent, four or less dimensional subspaces. We show that the row echelon canonical form provides direct information on the grouping of points to these subspaces. Treatment of the noise is the most difficult part of the problem. This paper uses a statistical approach to estimate the grouping of points to subspaces in the presence of noise by computing which partition has the maximum likelihood. The input data is assumed to be contaminated with independent Gaussian noise. The algorithm can base its estimates on a user-supplied standard deviation of the noise, or it can estimate the noise from the data. The algorithm can also be used to estimate the probability of a user-specified partition so that the hypothesis can be combined with others using Bayesian statistics.",
"The problem of simultaneous structure from motion estimation for multiple independently moving objects from a monocular image sequence is addressed. Two Bayesian algorithms are presented for solving this problem using the sequential importance sampling (SIS) technique. The empirical posterior distribution of object motion and feature separation parameters is approximated by weighted samples. The first algorithm addresses the problem when only two moving objects are present. A singular value decomposition (SVD)-based sample clustering algorithm is shown to be capable of separating samples related to different objects. A pair of SIS procedures is used to track the posterior distribution of the motion parameters. In the second algorithm, a balancing step is added into the SIS procedure to preserve samples of low weights so that all objects have enough samples to propagate empirical motion distributions. By using the proposed algorithms, the relative motions of all the moving objects with respect to the camera can be simultaneously estimated. Both algorithms have been tested on synthetic and real- image sequences. Improved results have been achieved.",
"An efficient and robust framework is proposed for two-view multiple structure-and-motion segmentation of unknown number of rigid objects. The segmentation problem has three unknowns, namely the object memberships, the corresponding fundamental matrices, and the number of objects. To handle this otherwise recursive problem, hypotheses for fundamental matrices are generated through local sampling. Once the hypotheses are available, a combinatorial selection problem is formulated to optimize a model selection cost which takes into account the hypotheses likelihoods and the model complexity. An explicit model for outliers is also added for robust segmentation. The model selection cost is minimized through the branch-and-bound technique of combinatorial optimization. The proposed branch-and-bound approach efficiently searches the solution space and guaranties optimality over the current set of hypotheses. The efficiency and the guarantee of optimality of the method is due to its ability to reject solutions without explicitly evaluating them. The proposed approach was validated with synthetic data, and segmentation results are presented for real images.",
"The structure-from-motion problem has been extensively studied in the field of computer vision. Yet, the bulk of the existing work assumes that the scene contains only a single moving object. The more realistic case where an unknown number of objects move in the scene has received little attention, especially for its theoretical treatment. In this paper we present a new method for separating and recovering the motion and shape of multiple independently moving objects in a sequence of images. The method does not require prior knowledge of the number of objects, nor is dependent on any grouping of features into an object at the image level. For this purpose, we introduce a mathematical construct of object shapes, called the shape interaction matrix, which is invariant to both the object motions and the selection of coordinate systems. This invariant structure is computable solely from the observed trajectories of image features without grouping them into individual objects. Once the matrix is computed, it allows for segmenting features into objects by the process of transforming it into a canonical form, as well as recovering the shape and motion of each object. The theory works under a broad set of projection models (scaled orthography, paraperspective and affine) but they must be linear, so it excludes projective “cameras”.",
"Multibody structure from motion (SfM) is the extension of classical SfM to dynamic scenes with multiple rigidly moving objects. Recent research has unveiled some of the mathematical foundations of the problem, but a practical algorithm which can handle realistic sequences is still missing. In this paper, we discuss the requirements for such an algorithm, highlight theoretical issues and practical problems, and describe how a static structure-from-motion framework needs to be extended to handle real dynamic scenes. Theoretical issues include different situations in which the number of independently moving scene objects changes: Moving objects can enter or leave the field of view, merge into the static background (e.g., when a car is parked), or split off from the background and start moving independently. Practical issues arise due to small freely moving foreground objects with few and short feature tracks. We argue that all of these difficulties need to be handled online as structure-from-motion estimation progresses, and present an exemplary solution using the framework of probabilistic model-scoring.",
"",
"In this paper, we tackle the problem of mapping multiple 3D rigid structures and estimating their motions from perspective views through a car-mounted camera. The proposed method complements conventional localization and mapping algorithms (such as Visual Odometry and SLAM) to estimate motions of other moving objects in addition to the vehicle's motion. We present a theoretical framework for robust estimation of multiple motions and structures from perspective images. The method is based on the factorization of the projective trajectory matrix without explicit estimation of projective depth values. We exploit the epipolar geometry of calibrated cameras to generate several hypotheses for motion segments. Once the hypotheses are obtained, they are evaluated in an iterative scheme by alternating between estimation of 3D structures and estimation of multiple motions. The proposed framework does not require any knowledge about the number of motions and is robust to noisy image measurements. The method is evaluated on street-level sequences from a car-mounted camera. A benchmark dataset is also used to compare the results with previous works, although most of the related works use synthetic scenes simulating desktop environments."
]
} |
1905.09043 | 2945106964 | In this paper we address a classification problem that has not been considered before, namely motion segmentation given pairwise matches only. Our contribution to this unexplored task is a novel formulation of motion segmentation as a two-step process. First, motion segmentation is performed on image pairs independently. Secondly, we combine independent pairwise segmentation results in a robust way into the final globally consistent segmentation. Our approach is inspired by the success of averaging methods. We demonstrate in simulated as well as in real experiments that our method is very effective in reducing the errors in the pairwise motion segmentation and can cope with large number of mismatches. | Ad-hoc solutions to motion segmentation are also present in the literature @cite_24 @cite_48 @cite_36 , which are not explicitly related to the aforementioned problems. The authors of @cite_24 formulate a joint optimization problem which builds upon the SSC algorithm, where it is required that all image pairs share a common sparsity profile. In @cite_48 an accumulated correlation matrix is built by sampling homographies over consecutive image pairs, and spectral clustering @cite_56 is applied to get the sought segmentation. Such approach is generalized in @cite_36 where multiple models (affine, fundamental and homography) are combined to get an improved segmentation. Different approaches are analyzed to reach such task, namely Kernel Addition (KerAdd) @cite_28 , Co-Regularization (Coreg) @cite_41 and Subset Constrained Clustering (Subset) @cite_30 . Motion segmentation is also addressed in @cite_10 @cite_50 , where existing algorithms are customized for specific scenarios and acquisition platforms. | {
"cite_N": [
"@cite_30",
"@cite_28",
"@cite_48",
"@cite_36",
"@cite_41",
"@cite_56",
"@cite_24",
"@cite_50",
"@cite_10"
],
"mid": [
"2052478943",
"",
"2528957674",
"2963035446",
"2154415691",
"2132914434",
"2082712442",
"2891687320",
"2412403816"
],
"abstract": [
"Constrained clustering has been well-studied for algorithms such as K-means and hierarchical clustering. However, how to satisfy many constraints in these algorithmic settings has been shown to be intractable. One alternative to encode many constraints is to use spectral clustering, which remains a developing area. In this paper, we propose a flexible framework for constrained spectral clustering. In contrast to some previous efforts that implicitly encode Must-Link (ML) and Cannot-Link (CL) constraints by modifying the graph Laplacian or constraining the underlying eigenspace, we present a more natural and principled formulation, which explicitly encodes the constraints as part of a constrained optimization problem. Our method offers several practical advantages: it can encode the degree of belief in ML and CL constraints; it guarantees to lower-bound how well the given constraints are satisfied using a user-specified threshold; it can be solved deterministically in polynomial time through generalized eigendecomposition. Furthermore, by inheriting the objective function from spectral clustering and encoding the constraints explicitly, much of the existing analysis of unconstrained spectral clustering techniques remains valid for our formulation. We validate the effectiveness of our approach by empirical results on both artificial and real datasets. We also demonstrate an innovative use of encoding large number of constraints: transfer learning via constraints.",
"",
"Motion segmentation is an important task for intelligent transportation systems. In this paper, inspired by the fact that a feature point trajectory can be sparsely represented as a combination of several feature point trajectories that share coherent transformations, an efficient and effective motion segmentation method with a sparsity constraint is proposed. Specifically, we first propose an accumulated scheme to efficiently integrate motion information from all the frames of a video sequence to construct a correlation matrix. Then, a sparse affinity matrix is built on the correlation matrix by using information-theoretic principles, where the nonzero elements in the same row of the sparse affinity matrix correspond to the feature point trajectories more likely belonging to the same motion. Thereafter, a segment and merge procedure is proposed to effectively estimate the number of motions via the sparse affinity matrix. Finally, by applying spectral clustering on the sparse affinity matrix, different motions in the video sequence are accurately segmented based on the estimated number of motions. Experimental results on the Hopkins 155 and the 62-clip datasets demonstrate that the proposed method achieves superior performance compared with several state-of-the-art methods.",
"Many real-world sequences cannot be conveniently categorized as general or degenerate; in such cases, imposing a false dichotomy in using the fundamental matrix or homography model for motion segmentation would lead to difficulty. Even when we are confronted with a general scene-motion, the fundamental matrix approach as a model for motion segmentation still suffers from several defects, which we discuss in this paper. The full potential of the fundamental matrix approach could only be realized if we judiciously harness information from the simpler homography model. From these considerations, we propose a multi-view spectral clustering framework that synergistically combines multiple models together. We show that the performance can be substantially improved in this way. We perform extensive testing on existing motion segmentation datasets, achieving state-of-the-art performance on all of them; we also put forth a more realistic and challenging dataset adapted from the KITTI benchmark, containing real-world effects such as strong perspectives and strong forward translations not seen in the traditional datasets.",
"In many clustering problems, we have access to multiple views of the data each of which could be individually used for clustering. Exploiting information from multiple views, one can hope to find a clustering that is more accurate than the ones obtained using the individual views. Often these different views admit same underlying clustering of the data, so we can approach this problem by looking for clusterings that are consistent across the views, i.e., corresponding data points in each view should have same cluster membership. We propose a spectral clustering framework that achieves this goal by co-regularizing the clustering hypotheses, and propose two co-regularization schemes to accomplish this. Experimental comparisons with a number of baselines on two synthetic and three real-world datasets establish the efficacy of our proposed approaches.",
"In recent years, spectral clustering has become one of the most popular modern clustering algorithms. It is simple to implement, can be solved efficiently by standard linear algebra software, and very often outperforms traditional clustering algorithms such as the k-means algorithm. On the first glance spectral clustering appears slightly mysterious, and it is not obvious to see why it works at all and what it really does. The goal of this tutorial is to give some intuition on those questions. We describe different graph Laplacians and their basic properties, present the most common spectral clustering algorithms, and derive those algorithms from scratch by several different approaches. Advantages and disadvantages of the different spectral clustering algorithms are discussed.",
"This paper addresses real-world challenges in the motion segmentation problem, including perspective effects, missing data, and unknown number of motions. It first formulates the 3-D motion segmentation from two perspective views as a subspace clustering problem, utilizing the epipolar constraint of an image pair. It then combines the point correspondence information across multiple image frames via a collaborative clustering step, in which tight integration is achieved via a mixed norm optimization scheme. For model selection, we propose an over-segment and merge approach, where the merging step is based on the property of the ell_1-norm of the mutual sparse representation of two over-segmented groups. The resulting algorithm can deal with incomplete trajectories and perspective effects substantially better than state-of-the-art two-frame and multi-frame methods. Experiments on a 62-clip dataset show the significant superiority of the proposed idea in both segmentation accuracy and model selection.",
"Though a long-studied problem, motion segmentation has yet to migrate into practical applications. We argue that a vital step towards that goal lies in addressing motion segmentation for the specific setting of interest. To this end, this paper presents a new approach for image-based motion segmentation in the case of vehicles navigating inside an urban environment. We exploit two application-specific factors – the restricted camera movement and the known type of moving objects – to deal with the two major limiting factors – missing data and strong perspective effects – that affect most previous “generic” motion segmentation algorithms. By constraining the geometry and exploiting known semantic classes in the scene, we achieve much higher accuracy than previous approaches. In addition to the novel algorithm, we contribute a more realistic motion segmentation benchmark dataset for moving platforms by annotating real video sequences from the KITTI dataset. Experiments on this dataset and other synthetic data confirm the effectiveness of the proposed approach.",
"This paper addresses the problem of simultaneous estimation of a vehicle's ego motion and motions of multiple moving objects in the scene–called eoru motions –through a monocular vehicle-mounted camera. Localization of multiple moving objects and estimation of their motions is crucial for autonomous vehicles. Conventional localization and mapping techniques (e.g., visual odometry and simultaneous localization and mapping) can only estimate the ego motion of the vehicle. The capability of a robot localization pipeline to deal with multiple motions has not been widely investigated in the literature. We present a theoretical framework for robust estimation of multiple relative motions in addition to the camera ego motion. First, the framework for general unconstrained motion is introduced and then it is adapted to exploit the vehicle kinematic constraints to increase efficiency. The method is based on projective factorization of the multiple-trajectory matrix . First, the ego motion is segmented and then several hypotheses are generated for the eoru motions . All the hypotheses are evaluated and the one with the smallest reprojection error is selected. The proposed framework does not need any a priori knowledge of the number of motions and is robust to noisy image measurements. The method with a constrained motion model is evaluated on a popular street-level image dataset collected in urban environments (the KITTI dataset), including several relative ego-motion and eoru-motion scenarios. A benchmark dataset ( Hopkins 155 ) is used to evaluate this method with a general motion model. The results are compared with those of the state-of-the-art methods considering a similar problem, referred to as multibody structure from motion in the computer vision community."
]
} |
1905.08880 | 2911926823 | We present the design and methodology for the large scale hybrid paper recommender system used by Microsoft Academic. The system provides recommendations for approximately 160 million English research papers and patents. Our approach handles incomplete citation information while also alleviating the cold-start problem that often affects other recommender systems. We use the Microsoft Academic Graph (MAG), titles, and available abstracts of research papers to build a recommendation list for all documents, thereby combining co-citation and content based approaches. Tuning system parameters also allows for blending and prioritization of each approach which, in turn, allows us to balance paper novelty versus authority in recommendation results. We evaluate the generated recommendations via a user study of 40 participants, with over 2400 recommendation pairs graded and discuss the quality of the results using P@10 and nDCG scores. We see that there is a strong correlation between participant scores and the similarity rankings produced by our system but that additional focus needs to be put towards improving recommender precision, particularly for content based recommendations. The results of the user survey and associated analysis scripts are made available via GitHub and the recommendations produced by our system are available as part of the MAG on Azure to facilitate further research and light up novel research paper recommendation applications. | Microsoft Academic https: preview.academic.microsoft.com (MA) is a semantic search engine for academic entities @cite_28 . The top level entities of the Microsoft Academic Graph (MAG) include papers and patents, fields of study, authors, affiliations (institutions and organizations), and venues (conferences and journals), as seen in Fig. . As of October, 2018, there are over 200 million total documents in the MA corpus, of which approximately 160 million are English papers or patents. These figures are growing rapidly @cite_19 . | {
"cite_N": [
"@cite_28",
"@cite_19"
],
"mid": [
"1932742904",
"2610002097"
],
"abstract": [
"In this paper we describe a new release of a Web scale entity graph that serves as the backbone of Microsoft Academic Service (MAS), a major production effort with a broadened scope to the namesake vertical search engine that has been publicly available since 2008 as a research prototype. At the core of MAS is a heterogeneous entity graph comprised of six types of entities that model the scholarly activities: field of study, author, institution, paper, venue, and event. In addition to obtaining these entities from the publisher feeds as in the previous effort, we in this version include data mining results from the Web index and an in-house knowledge base from Bing, a major commercial search engine. As a result of the Bing integration, the new MAS graph sees significant increase in size, with fresh information streaming in automatically following their discoveries by the search engine. In addition, the rich entity relations included in the knowledge base provide additional signals to disambiguate and enrich the entities within and beyond the academic domain. The number of papers indexed by MAS, for instance, has grown from low tens of millions to 83 million while maintaining an above 95 accuracy based on test data sets derived from academic activities at Microsoft Research. Based on the data set, we demonstrate two scenarios in this work: a knowledge driven, highly interactive dialog that seamlessly combines reactive search and proactive suggestion experience, and a proactive heterogeneous entity recommendation.",
"Progress in science has advanced the development of human society across history, with dramatic revolutions shaped by information theory, genetic cloning, and artificial intelligence, among the many scientific achievements produced in the 20th century. However, the way that science advances itself is much less well-understood. In this work, we study the evolution of scientific development over the past century by presenting an anatomy of 89 million digitalized papers published between 1900 and 2015. We find that science has benefited from the shift from individual work to collaborative effort, with over 90 of the world-leading innovations generated by collaborations in this century, nearly four times higher than they were in the 1900s. We discover that rather than the frequent myopic- and self-referencing that was common in the early 20th century, modern scientists instead tend to look for literature further back and farther around. Finally, we also observe the globalization of scientific development from 1900 to 2015, including 25-fold and 7-fold increases in international collaborations and citations, respectively, as well as a dramatic decline in the dominant accumulation of citations by the US, the UK, and Germany, from 95 to 50 over the same period. Our discoveries are meant to serve as a starter for exploring the visionary ways in which science has developed throughout the past century, generating insight into and an impact upon the current scientific innovations and funding policies."
]
} |
1905.08880 | 2911926823 | We present the design and methodology for the large scale hybrid paper recommender system used by Microsoft Academic. The system provides recommendations for approximately 160 million English research papers and patents. Our approach handles incomplete citation information while also alleviating the cold-start problem that often affects other recommender systems. We use the Microsoft Academic Graph (MAG), titles, and available abstracts of research papers to build a recommendation list for all documents, thereby combining co-citation and content based approaches. Tuning system parameters also allows for blending and prioritization of each approach which, in turn, allows us to balance paper novelty versus authority in recommendation results. We evaluate the generated recommendations via a user study of 40 participants, with over 2400 recommendation pairs graded and discuss the quality of the results using P@10 and nDCG scores. We see that there is a strong correlation between participant scores and the similarity rankings produced by our system but that additional focus needs to be put towards improving recommender precision, particularly for content based recommendations. The results of the user survey and associated analysis scripts are made available via GitHub and the recommendations produced by our system are available as part of the MAG on Azure to facilitate further research and light up novel research paper recommendation applications. | We present two possible interaction modes with the MA paper recommender platform. The first mode is via the Related Papers'' section on the MA search engine paper details page. Users can browse the MA search site using the novel semantic interpretation engine powering the search experience to view their desired paper @cite_28 . The paper details page, as seen in Fig., contains a tab for browsing related papers (with dynamic filters) that is populated using the techniques mentioned here. The second mode of interaction is via Microsoft Azure Data Lake (ADL) services. The entire MAG is published under the ODC-By open data license https: opendatacommons.org licenses by 1.0 and available for use via Microsoft Azure. Users can use scripting languages such as U-SQL and python not just to access the pre-computed paper recommendations available as part of the MAG but also to generate on-the-fly paper recommendations for arbitrary text input in a performant manner. The means by which this functionality is achieved is described in the following sections. | {
"cite_N": [
"@cite_28"
],
"mid": [
"1932742904"
],
"abstract": [
"In this paper we describe a new release of a Web scale entity graph that serves as the backbone of Microsoft Academic Service (MAS), a major production effort with a broadened scope to the namesake vertical search engine that has been publicly available since 2008 as a research prototype. At the core of MAS is a heterogeneous entity graph comprised of six types of entities that model the scholarly activities: field of study, author, institution, paper, venue, and event. In addition to obtaining these entities from the publisher feeds as in the previous effort, we in this version include data mining results from the Web index and an in-house knowledge base from Bing, a major commercial search engine. As a result of the Bing integration, the new MAS graph sees significant increase in size, with fresh information streaming in automatically following their discoveries by the search engine. In addition, the rich entity relations included in the knowledge base provide additional signals to disambiguate and enrich the entities within and beyond the academic domain. The number of papers indexed by MAS, for instance, has grown from low tens of millions to 83 million while maintaining an above 95 accuracy based on test data sets derived from academic activities at Microsoft Research. Based on the data set, we demonstrate two scenarios in this work: a knowledge driven, highly interactive dialog that seamlessly combines reactive search and proactive suggestion experience, and a proactive heterogeneous entity recommendation."
]
} |
1905.08880 | 2911926823 | We present the design and methodology for the large scale hybrid paper recommender system used by Microsoft Academic. The system provides recommendations for approximately 160 million English research papers and patents. Our approach handles incomplete citation information while also alleviating the cold-start problem that often affects other recommender systems. We use the Microsoft Academic Graph (MAG), titles, and available abstracts of research papers to build a recommendation list for all documents, thereby combining co-citation and content based approaches. Tuning system parameters also allows for blending and prioritization of each approach which, in turn, allows us to balance paper novelty versus authority in recommendation results. We evaluate the generated recommendations via a user study of 40 participants, with over 2400 recommendation pairs graded and discuss the quality of the results using P@10 and nDCG scores. We see that there is a strong correlation between participant scores and the similarity rankings produced by our system but that additional focus needs to be put towards improving recommender precision, particularly for content based recommendations. The results of the user survey and associated analysis scripts are made available via GitHub and the recommendations produced by our system are available as part of the MAG on Azure to facilitate further research and light up novel research paper recommendation applications. | To tackle these requirements we have developed a hybrid recommender system that uses a tunable mapping function to combine both content-based (CB) and co-citation based (CcB) recommendations to produce a final static list of recommendations for every paper in the MAG. As described in @cite_10 , our approach employs a hybridization approach. Our content-based approach is similar to recent work done on content embedding based citation recommendation @cite_7 but differs mainly in the fact that we employ clustering techniques for additional speedup. Using purely pairwise content embedding similarity for nearest neighbor search is not viable as this is an @math problem over the entire paper corpus, which in our case would be over @math similarity computations. | {
"cite_N": [
"@cite_10",
"@cite_7"
],
"mid": [
"281665770",
"2023930240"
],
"abstract": [
"Recommender systems represent user preferences for the purpose of suggesting items to purchase or examine. They have become fundamental applications in electronic commerce and information access, providing suggestions that effectively prune large information spaces so that users are directed toward those items that best meet their needs and preferences. A variety of techniques have been proposed for performing recommendation, including content-based, collaborative, knowledge-based and other techniques. To improve performance, these methods have sometimes been combined in hybrid recommenders. This paper surveys the landscape of actual and possible hybrid recommenders, and introduces a novel hybrid, EntreeC, a system that combines knowledge-based recommendation and collaborative filtering to recommend restaurants. Further, we show that semantic ratings obtained from the knowledge-based part of the system enhance the effectiveness of collaborative filtering.",
"Citations are important in academic dissemination. To help researchers check the completeness of citations while authoring a paper, we introduce a citation recommendation system called RefSeer. Researchers can use it to find related works to cited while authoring papers. It can also be used by reviewers to check the completeness of a paper's references. RefSeer presents both topic based global recommendation and also citation-context based local recommendation. By evaluating the quality of recommendation, we show that such recommendation system can recommend citations with good precision and recall. We also show that our recommendation system is very efficient and scalable."
]
} |
1905.08856 | 2965420493 | Motivated by concerns about diversity in social networks, we consider the following pattern formation problems in rings. Assume n mobile agents are located at the nodes of an n-node ring network. Each agent is assigned a colour from the set ( c_1, c_2, , c_q ). The ring is divided into k contiguous blocks or neighbourhoods of length p. The agents are required to rearrange themselves in a distributed manner to satisfy given diversity requirements: in each block j and for each colour (c_i ), there must be exactly (n_i(j) >0 ) agents of colour (c_i ) in block j. Agents are assumed to be able to see agents in adjacent blocks, and move to any position in adjacent blocks in one time step. | There is a large literature on distributed computing by autonomous mobile agents, see the book @cite_12 that provides a very complete survey of results in this area. Much of this work is done for mobile agents located in the plane or a continuous region, and the main problems that have been considered are gathering of robots @cite_2 @cite_22 , scattering and covering @cite_26 , and pattern formation @cite_5 @cite_0 by autonomous mobile robots (see [section 1.3] flocchini2019distributed ). The solvability, i.e., termination or convergence, of these problems depends on the type of synchronization, visibility, agreement on orientation @cite_12 . @cite_9 , the authors introduce the problem of using a set of pairwise-communicating processes to constructing a given network. | {
"cite_N": [
"@cite_26",
"@cite_22",
"@cite_9",
"@cite_0",
"@cite_2",
"@cite_5",
"@cite_12"
],
"mid": [
"1967871771",
"2010017329",
"2189116479",
"2512738617",
"2041571902",
"2044484214",
"1980710447"
],
"abstract": [
"This paper studies local algorithms for autonomous robot systems, namely, algorithms that use only information of the positions of a bounded number of their nearest neighbors. The paper focuses on the spreading problem. It defines measures for the quality of spreading, presents a local algorithm for the one-dimensional spreading problem, proves its convergence to the equally spaced configuration and discusses its convergence rate in the synchronous and semi-synchronous settings. It then presents a local algorithm achieving the exact equally spaced configuration in finite time in the synchronous setting, and proves it is time optimal for local algorithms. Finally, the paper also proposes a possible algorithm for the two-dimensional case and presents partial simulation results of its effectiveness.",
"In this paper we study the problem of gathering a collection of identical oblivious mobile robots in the same location of the plane. Previous investigations have focused mostly on the unlimited visibility setting, where each robot can always see all the others regardless of their distance.In the more difficult and realistic setting where the robots have limited visibility, the existing algorithmic results are only for convergence (towards a common point, without ever reaching it) and only for semi-synchronous environments, where robots' movements are assumed to be performed instantaneously.In contrast, we study this problem in a totally asynchronous setting, where robots' actions, computations, and movements require a finite but otherwise unpredictable amount of time. We present a protocol that allows anonymous oblivious robots with limited visibility to gather in the same location in finite time, provided they have orientation (i.e., agreement on a coordinate system).Our result indicates that, with respect to gathering, orientation is at least as powerful as instantaneous movements.",
"In this work, we study protocols so that populations of distributed processes can construct networks. In order to highlight the basic principles of distributed network construction, we keep the model minimal in all respects. In particular, we assume finite-state processes that all begin from the same initial state and all execute the same protocol. Moreover, we assume pairwise interactions between the processes that are scheduled by a fair adversary. In order to allow processes to construct networks, we let them activate and deactivate their pairwise connections. When two processes interact, the protocol takes as input the states of the processes and the state of their connection and updates all of them. Initially all connections are inactive and the goal is for the processes, after interacting and activating deactivating connections for a while, to end up with a desired stable network. We give protocols (optimal in some cases) and lower bounds for several basic network construction problems such as spanning line, spanning ring, spanning star, and regular network. The expected time to convergence of our protocols is analyzed under a uniform random scheduler. Finally, we prove several universality results by presenting generic protocols that are capable of simulating a Turing Machine (TM) and exploiting it in order to construct a large class of networks. We additionally show how to partition the population into ksupernodes, each being a line of @math logk nodes, for the largest such k. This amount of local memory is sufficient for the supernodes to obtain unique names and exploit their names and their memory to realize nontrivial constructions.",
"We consider the Embedded Pattern Formation (epf) problem introduced in [, SIAM J. on Comput., 44(3), 2015]. Given a set F of distinct points in the Euclidean plane (called here fixed-points) and a set R of robots such that (|R|=|F| ), the problem asks for a distributed algorithm that moves robots so as to occupy all points in F. Initially, each robot occupies a distinct position.",
"This paper considers the convergence problem in autonomous mobile robot systems. A natural algorithm for the problem requires the robots to move towards their center of gravity. This paper proves the correctness of the gravitational algorithm in the fully asynchronous model. It also analyzes its convergence rate and establishes its convergence in the presence of crash faults.",
"In this note we make a minor correction to a scheme for robots to broadcast their private information. All major results of the paper [I. Suzuki and M. Yamashita, SIAM J. Comput., 28 (1999), pp. 1347-1363] hold with this correction.",
"The study of what can be computed by a team of autonomous mobile robots, originally started in robotics and AI, has become increasingly popular in theoretical computer science (especially in distributed computing), where it is now an integral part of the investigations on computability by mobile entities. The robots are identical computational entities located and able to move in a spatial universe; they operate without explicit communication and are usually unable to remember the past; they are extremely simple, with limited resources, and individually quite weak. However, collectively the robots are capable of performing complex tasks, and form a system with desirable fault-tolerant and self-stabilizing properties. The research has been concerned with the computational aspects of such systems. In particular, the focus has been on the minimal capabilities that the robots should have in order to solve a problem. This book focuses on the recent algorithmic results in the field of distributed computing by oblivious mobile robots (unable to remember the past). After introducing the computational model with its nuances, we focus on basic coordination problems: pattern formation, gathering, scattering, leader election, as well as on dynamic tasks such as flocking. For each of these problems, we provide a snapshot of the state of the art, reviewing the existing algorithmic results. In doing so, we outline solution techniques, and we analyze the impact of the different assumptions on the robots' computability power. Table of Contents: Introduction Computational Models Gathering and Convergence Pattern Formation Scatterings and Coverings Flocking Other Directions"
]
} |
1905.08856 | 2965420493 | Motivated by concerns about diversity in social networks, we consider the following pattern formation problems in rings. Assume n mobile agents are located at the nodes of an n-node ring network. Each agent is assigned a colour from the set ( c_1, c_2, , c_q ). The ring is divided into k contiguous blocks or neighbourhoods of length p. The agents are required to rearrange themselves in a distributed manner to satisfy given diversity requirements: in each block j and for each colour (c_i ), there must be exactly (n_i(j) >0 ) agents of colour (c_i ) in block j. Agents are assumed to be able to see agents in adjacent blocks, and move to any position in adjacent blocks in one time step. | In discrete spaces, typically graphs like grids or rings, similar problems have been studied, see for example: gathering @cite_17 @cite_25 ; exploration @cite_21 ; pattern formation @cite_29 ; and deployment @cite_15 . Unlike in our paper, in all these studies, only a subset of the nodes of the graph is occupied by agents, and typically agents are identical, and not partitioned into colour classes. | {
"cite_N": [
"@cite_29",
"@cite_21",
"@cite_15",
"@cite_25",
"@cite_17"
],
"mid": [
"2899152902",
"1617059252",
"1969099455",
"43751763",
"2144182788"
],
"abstract": [
"The Arbitrary Pattern Formation problem asks to design a distributed algorithm that allows a set of autonomous mobile robots to form any specific but arbitrary geometric pattern given as input. The problem has been extensively studied in literature in continuous domains. This paper investigates a discrete version of the problem where the robots are operating on a two dimensional infinite grid. The robots are assumed to be autonomous, identical, anonymous and oblivious. They operate in Look-Compute-Move cycles under a fully asynchronous scheduler. The robots do not agree on any common global coordinate system or chirality. We have shown that a set of robots can form any arbitrary pattern, if their starting configuration is asymmetric.",
"We consider the problem of exploring an anonymous unoriented ring of size n by k identical, oblivious, asynchronous mobile robots, that are unable to communicate, yet have the ability to sense their environment and take decisions based on their local view. Previous works in this weak scenario prove that k must not divide n for a deterministic solution to exist. Also, it is known that the minimum number of robots (either deterministic or probabilistic) to explore a ring of size n is 4. An upper bound of 17 robots holds in the deterministic case while 4 probabilistic robots are sufficient. In this paper, we close the complexity gap in the deterministic setting, by proving that no deterministic exploration is feasible with less than five robots, and that five robots are sufficient for any n that is coprime with five. Our protocol completes exploration in O(n) robot moves, which is also optimal.",
"We consider two variants of the task of spreading a swarm of agents uniformly on a ring graph. Ant-like oblivious agents having limited capabilities are considered. The agents are assumed to have little memory, they all execute the same algorithm and no direct communication is allowed between them. Furthermore, the agents do not possess any global information. In particular, the size of the ring (n) and the number of agents in the swarm (k) are unknown to them. The agents are assumed to operate on an unweighted ring graph. Every agent can measure the distance to his two neighbors on the ring, up to a limited range of V edges. The first task considered, is dynamical (i.e. in motion) uniform deployment on the ring. We show that if either the ring is unoriented, or the visibility range is less than @?n [email protected]?, this is an impossible mission for the agents. Then, for an oriented ring and V>[email protected]?n [email protected]?, we propose an algorithm which achieves the deployment task in optimal time. The second task discussed, called quiescent spread, requires the agents to spread uniformly over the ring and stop moving. We prove that under our model, in which every agent can measure the distance only to his two neighbors, this task is impossible. Subsequently, we propose an algorithm which achieves quiescent but only almost uniform spread. The algorithms we present are scalable and robust. In case the environment (the size of the ring) or the number of agents changes during the run, the swarm adapts and re-deploys without requiring any outside interference.",
"The paper studies the gathering problem on grid networks. A team of robots placed at different nodes of a grid, have to meet at some node and remain there. Robots operate in Look-Compute-Move cycles; in one cycle, a robot perceives the current configuration in terms of occupied nodes (Look), decides whether to move towards one of its neighbors (Compute), and in the positive case makes the computed move instantaneously (Move). Cycles are performed asynchronously for each robot. The problem has been deeply studied for the case of ring networks. However, the known techniques used on rings cannot be directly extended to grids. Moreover, on rings, another assumption concerning the so-called multiplicity detection capability was required in order to accomplish the gathering task. That is, a robot is able to detect during its Look operation whether a node is empty, or occupied by one robot, or occupied by an undefined number of robots greater than one. In this paper, we provide a full characterization about gatherable configurations for grids. In particular, we show that in this case, the multiplicity detection is not required. Very interestingly, sometimes the problem appears trivial, as it is for the case of grids with both odd sides, while sometimes the involved techniques require new insights with respect to the well-studied ring case. Moreover, our results reveal the importance of a structure like the grid that allows to overcome the multiplicity detection with respect to the ring case.",
"We consider the problem of gathering identical, memoryless, mobile robots in one node of an anonymous unoriented ring. Robots start from different nodes of the ring. They operate in Look-Compute-Move cycles and have to end up in the same node. In one cycle, a robot takes a snapshot of the current configuration (Look), makes a decision to stay idle or to move to one of its adjacent nodes (Compute), and in the latter case makes an instantaneous move to this neighbor (Move). Cycles are performed asynchronously for each robot. For an odd number of robots we prove that gathering is feasible if and only if the initial configuration is not periodic, and we provide a gathering algorithm for any such configuration. For an even number of robots we decide the feasibility of gathering except for one type of symmetric initial configurations, and provide gathering algorithms for initial configurations proved to be gatherable."
]
} |
1905.09035 | 2945026633 | Egocentric action anticipation consists in understanding which objects the camera wearer will interact with in the near future and which actions they will perform. We tackle the problem proposing an architecture able to anticipate actions at multiple temporal scales using two LSTMs to 1) summarize the past, and 2) formulate predictions about the future. The input video is processed considering three complimentary modalities: appearance (RGB), motion (optical flow) and objects (object-based features). Modality-specific predictions are fused using a novel Modality ATTention (MATT) mechanism which learns to weigh modalities in an adaptive fashion. Extensive evaluations on two large-scale benchmark datasets show that our method outperforms prior art by up to +7 on the challenging EPIC-KITCHENS dataset including more than 2500 actions, and generalizes to EGTEA Gaze+. Our approach is also shown to generalize to the tasks of early action recognition and action recognition. At the moment of submission, our method is ranked first in the leaderboard of the EPIC-KITCHENS egocentric action anticipation challenge. | Early action recognition consists in recognizing an ongoing action as early as possible from partial observations @cite_48 . The problem has been widely investigated in the domain of third person vision @cite_47 @cite_10 @cite_23 @cite_48 @cite_45 @cite_36 @cite_8 @cite_40 @cite_52 . Differently from these works, we address the task of actions from egocentric video, i.e., the action should be recognized before it starts, hence it cannot even be partially observed at the time of prediction. Given the similarity between early recognition and anticipation, we consider and evaluate some ideas investigated in the context of early action recognition, such as the use of LSTMs to process streaming video @cite_47 @cite_45 @cite_40 , and the use of dedicated loss functions @cite_40 . Moreover, we show that the proposed architecture also generalizes to the problem of early egocentric action recognition, achieving state-of-the-art performances. | {
"cite_N": [
"@cite_8",
"@cite_36",
"@cite_48",
"@cite_52",
"@cite_40",
"@cite_45",
"@cite_23",
"@cite_47",
"@cite_10"
],
"mid": [
"225975350",
"2016776918",
"2342437993",
"2147615062",
"2472970127",
"2800438594",
"2118527252",
"2605300166",
"2610990652"
],
"abstract": [
"Many applications in computer vision (e.g., games, human computer interaction) require a reliable and early detector of visual events. Existing event detection methods rely on one-versus-all or multi-class classifiers that do not scale well to online detection of large number of events. This paper proposes Sequential Max-Margin Event Detectors (SMMED) to efficiently detect an event in the presence of a large number of event classes. SMMED sequentially discards classes until only one class is identified as the detected class. This approach has two main benefits w.r.t. standard approaches: (1) It provides an efficient solution for early detection of events in the presence of large number of classes, and (2) it is computationally efficient because only a subset of likely classes are evaluated. The benefits of SMMED in comparison with existing approaches is illustrated in three databases using different modalities: MSRDaliy Activity (3D depth videos), UCF101 (RGB videos) and the CMU-Multi-Modal Action Detection (MAD) database (depth, RGB and skeleton). The CMU-MAD was recorded to target the problem of event detection (not classification), and the data and labels are available at http: humansensing.cs.cmu.edu mad .",
"The need for early detection of temporal events from sequential data arises in a wide spectrum of applications ranging from human-robot interaction to video security. While temporal event detection has been extensively studied, early detection is a relatively unexplored problem. This paper proposes a maximum-margin framework for training temporal event detectors to recognize partial events, enabling early detection. Our method is based on Structured Output SVM, but extends it to accommodate sequential data. Experiments on datasets of varying complexity, for detecting facial expressions, hand gestures, and human activities, demonstrate the benefits of our approach.",
"In online action detection, the goal is to detect the start of an action in a video stream as soon as it happens. For instance, if a child is chasing a ball, an autonomous car should recognize what is going on and respond immediately. This is a very challenging problem for four reasons. First, only partial actions are observed. Second, there is a large variability in negative data. Third, the start of the action is unknown, so it is unclear over what time window the information should be integrated. Finally, in real world data, large within-class variability exists. This problem has been addressed before, but only to some extent. Our contributions to online action detection are threefold. First, we introduce a realistic dataset composed of 27 episodes from 6 popular TV series. The dataset spans over 16 h of footage annotated with 30 action classes, totaling 6,231 action instances. Second, we analyze and compare various baseline methods, showing this is a challenging problem for which none of the methods provides a good solution. Third, we analyze the change in performance when there is a variation in viewpoint, occlusion, truncation, etc. We introduce an evaluation protocol for fair comparison. The dataset, the baselines and the models will all be made publicly available to encourage (much needed) further research on online action detection on realistic data.",
"In this paper, we present a novel approach of human activity prediction. Human activity prediction is a probabilistic process of inferring ongoing activities from videos only containing onsets (i.e. the beginning part) of the activities. The goal is to enable early recognition of unfinished activities as opposed to the after-the-fact classification of completed activities. Activity prediction methodologies are particularly necessary for surveillance systems which are required to prevent crimes and dangerous activities from occurring. We probabilistically formulate the activity prediction problem, and introduce new methodologies designed for the prediction. We represent an activity as an integral histogram of spatio-temporal features, efficiently modeling how feature distributions change over time. The new recognition methodology named dynamic bag-of-words is developed, which considers sequential nature of human activities while maintaining advantages of the bag-of-words to handle noisy observations. Our experiments confirm that our approach reliably recognizes ongoing activities from streaming videos with a high accuracy.",
"In this work we improve training of temporal deep models to better learn activity progression for activity detection and early detection tasks. Conventionally, when training a Recurrent Neural Network, specifically a Long Short Term Memory (LSTM) model, the training loss only considers classification error. However, we argue that the detection score of the correct activity category, or the detection score margin between the correct and incorrect categories, should be monotonically non-decreasing as the model observes more of the activity. We design novel ranking losses that directly penalize the model on violation of such monotonicities, which are used together with classification loss in training of LSTM models. Evaluation on ActivityNet shows significant benefits of the proposed ranking losses in both activity detection and early detection tasks.",
"Online action detection is a challenging problem: a system needs to decide what action is happening at the current frame, based on previous frames only. Fortunately in real-life, human actions are not independent from one another: there are strong (long-term) dependencies between them. An online action detection method should be able to capture these dependencies, to enable a more accurate early detection. At first sight, an LSTM seems very suitable for this problem. It is able to model both short-term and long-term patterns. It takes its input one frame at the time, updates its internal state and gives as output the current class probabilities. In practice, however, the detection results obtained with LSTMs are still quite low. In this work, we start from the hypothesis that it may be too difficult for an LSTM to learn both the interpretation of the input and the temporal patterns at the same time. We propose a two-stream feedback network, where one stream processes the input and the other models the temporal relations. We show improved detection accuracy on an artificial toy dataset and on the Breakfast Dataset [21] and the TVSeries Dataset [7], reallife datasets with inherent temporal dependencies between the actions.",
"Recognizing human activities in partially observed videos is a challenging problem and has many practical applications. When the unobserved subsequence is at the end of the video, the problem is reduced to activity prediction from unfinished activity streaming, which has been studied by many researchers. However, in the general case, an unobserved subsequence may occur at any time by yielding a temporal gap in the video. In this paper, we propose a new method that can recognize human activities from partially observed videos in the general case. Specifically, we formulate the problem into a probabilistic framework: 1) dividing each activity into multiple ordered temporal segments, 2) using spatiotemporal features of the training video samples in each segment as bases and applying sparse coding (SC) to derive the activity likelihood of the test video sample at each segment, and 3) finally combining the likelihood at each segment to achieve a global posterior for the activities. We further extend the proposed method to include more bases that correspond to a mixture of segments with different temporal lengths (MSSC), which can better represent the activities with large intra-class variations. We evaluate the proposed methods (SC and MSSC) on various real videos. We also evaluate the proposed methods on two special cases: 1) activity prediction where the unobserved subsequence is at the end of the video, and 2) human activity recognition on fully observed videos. Experimental results show that the proposed methods outperform existing state-of-the-art comparison methods.",
"In contrast to the widely studied problem of recognizing an action given a complete sequence, action anticipation aims to identify the action from only partially available videos. As such, it is therefore key to the success of computer vision applications requiring to react as early as possible, such as autonomous navigation. In this paper, we propose a new action anticipation method that achieves high prediction accuracy even in the presence of a very small percentage of a video sequence. To this end, we develop a multi-stage LSTM architecture that leverages context-aware and action-aware features, and introduce a novel loss function that encourages the model to predict the correct class as early as possible. Our experiments on standard benchmark datasets evidence the benefits of our approach; We outperform the state-of-the-art action anticipation methods for early prediction by a relative increase in accuracy of 22.0 on JHMDB-21, 14.0 on UT-Interaction and 49.9 on UCF-101.",
"In this paper we introduce the problem of predicting action progress in videos. We argue that this is an extremely important task because, on the one hand, it can be valuable for a wide range of applications and, on the other hand, it facilitates better action detection results. To this end we introduce a novel approach, named ProgressNet, capable of predicting when an action takes place in a video, where it is located within the frames, and how far it has progressed during its execution. Motivated by the recent success obtained from the interaction of Convolutional and Recurrent Neural Networks, our model is based on a combination of the Faster R-CNN framework, to make framewise predictions, and LSTM networks, to estimate action progress through time. After introducing two evaluation protocols for the task at hand, we demonstrate the capability of our model to effectively predict action progress on the UCF-101 and J-HMDB datasets. Additionally, we show that exploiting action progress it is also possible to improve spatio-temporal localization."
]
} |
1905.09035 | 2945026633 | Egocentric action anticipation consists in understanding which objects the camera wearer will interact with in the near future and which actions they will perform. We tackle the problem proposing an architecture able to anticipate actions at multiple temporal scales using two LSTMs to 1) summarize the past, and 2) formulate predictions about the future. The input video is processed considering three complimentary modalities: appearance (RGB), motion (optical flow) and objects (object-based features). Modality-specific predictions are fused using a novel Modality ATTention (MATT) mechanism which learns to weigh modalities in an adaptive fashion. Extensive evaluations on two large-scale benchmark datasets show that our method outperforms prior art by up to +7 on the challenging EPIC-KITCHENS dataset including more than 2500 actions, and generalizes to EGTEA Gaze+. Our approach is also shown to generalize to the tasks of early action recognition and action recognition. At the moment of submission, our method is ranked first in the leaderboard of the EPIC-KITCHENS egocentric action anticipation challenge. | Action anticipation is the task of predicting an action it occurs @cite_3 . Previous works investigated different forms of action and activity anticipation from third person video @cite_32 @cite_33 @cite_3 @cite_60 @cite_41 @cite_35 @cite_21 @cite_58 @cite_12 @cite_46 @cite_39 @cite_34 . While we consider the problem of action anticipation from egocentric visual data, our work builds on some ideas explored in past works such as the use of LSTMs to anticipate actions @cite_32 @cite_3 @cite_35 , the use of the encoder-decoder framework to encode past observations and produce hypotheses of future actions @cite_3 , and the use of object specific features @cite_46 to determine which objects are present in the scene. Additionally, we show that other approaches, such as the direct regression of future representations @cite_3 @cite_39 , do not achieve satisfactory performance in the considered scenario. | {
"cite_N": [
"@cite_35",
"@cite_33",
"@cite_60",
"@cite_41",
"@cite_21",
"@cite_32",
"@cite_3",
"@cite_39",
"@cite_34",
"@cite_46",
"@cite_58",
"@cite_12"
],
"mid": [
"2963692464",
"2780253814",
"2099320314",
"1789187189",
"",
"2963570630",
"",
"2422305492",
"",
"2779943048",
"1483019628",
"2185953016"
],
"abstract": [
"Anticipating the future actions of a human is a widely studied problem in robotics that requires spatio-temporal reasoning. In this work we propose a deep learning approach for anticipation in sensory-rich robotics applications. We introduce a sensory-fusion architecture which jointly learns to anticipate and fuse information from multiple sensory streams. Our architecture consists of Recurrent Neural Networks (RNNs) that use Long Short-Term Memory (LSTM) units to capture long temporal dependencies. We train our architecture in a sequence-to-sequence prediction manner, and it explicitly learns to predict the future given only a partial temporal context. We further introduce a novel loss layer for anticipation which prevents over-fitting and encourages early anticipation. We use our architecture to anticipate driving maneuvers several seconds before they happen on a natural driving data set of 1180 miles. The context for maneuver anticipation comes from multiple sensors installed on the vehicle. Our approach shows significant improvement over the state-of-the-art in maneuver anticipation by increasing the precision from 77.4 to 90.5 and recall from 71.2 to 87.4 .",
"A large number of very popular team sports involve the act of one team trying to score a goal against the other. During this game play, defending players constantly try to predict the next move of the attackers to prevent them from scoring, whereas attackers constantly try to predict the next move of the defenders in order to defy them and score. Such behavior is a prime example of the general human faculty to make predictions about the future and is an important facet of human intelligence. An algorithmic solution to learning a model of the external world from sensory inputs in order to make forecasts is an important unsolved problem. In this work we develop a generic framework for forecasting future events in team sports videos directly from visual inputs. We introduce water polo and basketball datasets towards this end and compare the predictions of the proposed methods against expert and non-expert humans.",
"Forecasting human activities from visual evidence is an emerging area of research which aims to allow computational systems to make predictions about unseen human actions. We explore the task of activity forecasting in the context of dual-agent interactions to understand how the actions of one person can be used to predict the actions of another. We model dual-agent interactions as an optimal control problem, where the actions of the initiating agent induce a cost topology over the space of reactive poses – a space in which the reactive agent plans an optimal pose trajectory. The technique developed in this work employs a kernel-based reinforcement learning approximation of the soft maximum value function to deal with the high-dimensional nature of human motion and applies a mean-shift procedure over a continuous cost function to infer a smooth reaction sequence. Experimental results show that our proposed method is able to properly model human interactions in a high dimensional space of human poses. When compared to several baseline models, results show that our method is able to generate highly plausible simulations of human interaction.",
"Advanced Driver Assistance Systems (ADAS) have made driving safer over the last decade. They prepare vehicles for unsafe road conditions and alert drivers if they perform a dangerous maneuver. However, many accidents are unavoidable because by the time drivers are alerted, it is already too late. Anticipating maneuvers beforehand can alert drivers before they perform the maneuver and also give ADAS more time to avoid or prepare for the danger. In this work we anticipate driving maneuvers a few seconds before they occur. For this purpose we equip a car with cameras and a computing device to capture the driving context from both inside and outside of the car. We propose an Autoregressive Input-Output HMM to model the contextual information alongwith the maneuvers. We evaluate our approach on a diverse data set with 1180 miles of natural freeway and city driving and show that we can anticipate maneuvers 3.5 seconds before they occur with over 80 F1-score in real-time.",
"",
"Analyzing human actions in videos has gained increased attention recently. While most works focus on classifying and labeling observed video frames or anticipating the very recent future, making long-term predictions over more than just a few seconds is a task with many practical applications that has not yet been addressed. In this paper, we propose two methods to predict a considerably large amount of future actions and their durations. Both, a CNN and an RNN are trained to learn future video labels based on previously seen content. We show that our methods generate accurate predictions of the future even for long videos with a huge amount of different actions and can even deal with noisy or erroneous input information.",
"",
"Anticipating actions and objects before they start or appear is a difficult problem in computer vision with several real-world applications. This task is challenging partly because it requires leveraging extensive knowledge of the world that is difficult to write down. We believe that a promising resource for efficiently learning this knowledge is through readily available unlabeled video. We present a framework that capitalizes on temporal structure in unlabeled video to learn to anticipate human actions and objects. The key idea behind our approach is that we can train deep networks to predict the visual representation of images in the future. Visual representations are a promising prediction target because they encode images at a higher semantic level than pixels yet are automatic to compute. We then apply recognition algorithms on our predicted representation to anticipate objects and actions. We experimentally validate this idea on two datasets, anticipating actions one second in the future and objects five seconds in the future.",
"",
"Most of the existing works on human activity analysis focus on recognition or early recognition of the activity labels from complete or partial observations. Predicting the labels of future unobserved activities where no frames of the predicted activities have been observed is a challenging problem, with important applications, which has not been explored much. Associated with the future label prediction problem is the problem of predicting the starting time of the next activity. In this work, we propose a system that is able to infer about the labels and the starting times of future activities. Activities are characterized by the previous activity sequence (which is observed), as well as the objects present in the scene during their occurrence. We propose a network similar to a hybrid Siamese network with three branches to jointly learn both the future label and the starting time. The first branch takes visual features from the objects present in the scene using a fully connected network, the second branch takes previous activity features using a LSTM network to model long-term sequential relationships and the third branch captures the last observed activity features to model the context of inter-activity time using another fully connected network. These concatenated features are used for both label and time prediction. Experiments on two challenging datasets demonstrate that our framework for joint prediction of activity label and starting time improves the performance of both, and outperforms the state-of-the-arts.",
"An important aspect of human perception is anticipation, which we use extensively in our day-to-day activities when interacting with other humans as well as with our surroundings. Anticipating which activities will a human do next (and how) can enable an assistive robot to plan ahead for reactive responses. Furthermore, anticipation can even improve the detection accuracy of past activities. The challenge, however, is two-fold: We need to capture the rich context for modeling the activities and object affordances, and we need to anticipate the distribution over a large space of future human activities. In this work, we represent each possible future using an anticipatory temporal conditional random field (ATCRF) that models the rich spatial-temporal relations through object affordances. We then consider each ATCRF as a particle and represent the distribution over the potential futures using a set of particles. In extensive evaluation on CAD-120 human activity RGB-D dataset, we first show that anticipation improves the state-of-the-art detection results. We then show that for new subjects (not seen in the training set), we obtain an activity anticipation accuracy (defined as whether one of top three predictions actually happened) of 84.1, 74.4 and 62.2 percent for an anticipation time of 1, 3 and 10 seconds respectively. Finally, we also show a robot using our algorithm for performing a few reactive responses.",
"We consider inferring the future actions of people from a still image or a short video clip. Predicting future actions before they are actually executed is a critical ingredient for enabling us to effectively interact with other humans on a daily basis. However, challenges are two fold: First, we need to capture the subtle details inherent in human movements that may imply a future action; second, predictions usually should be carried out as quickly as possible in the social world, when limited prior observations are available."
]
} |
1905.09035 | 2945026633 | Egocentric action anticipation consists in understanding which objects the camera wearer will interact with in the near future and which actions they will perform. We tackle the problem proposing an architecture able to anticipate actions at multiple temporal scales using two LSTMs to 1) summarize the past, and 2) formulate predictions about the future. The input video is processed considering three complimentary modalities: appearance (RGB), motion (optical flow) and objects (object-based features). Modality-specific predictions are fused using a novel Modality ATTention (MATT) mechanism which learns to weigh modalities in an adaptive fashion. Extensive evaluations on two large-scale benchmark datasets show that our method outperforms prior art by up to +7 on the challenging EPIC-KITCHENS dataset including more than 2500 actions, and generalizes to EGTEA Gaze+. Our approach is also shown to generalize to the tasks of early action recognition and action recognition. At the moment of submission, our method is ranked first in the leaderboard of the EPIC-KITCHENS egocentric action anticipation challenge. | Past works on anticipation from egocentric visual data have investigated different problems and considered different evaluation frameworks @cite_27 @cite_50 @cite_25 @cite_13 @cite_7 @cite_28 @cite_19 @cite_5 @cite_66 @cite_49 . Instead, we tackle the egocentric action anticipation challenge recently proposed in @cite_24 , which has been little investigated so far @cite_30 . While a direct comparison of the proposed approach with the aforementioned works is unfeasible due to the lack of a common framework, our method incorporates some ideas from past approaches, such as the analysis of past actions @cite_28 and the detection of the objects present in the scene to infer future actions @cite_25 . | {
"cite_N": [
"@cite_30",
"@cite_7",
"@cite_28",
"@cite_66",
"@cite_24",
"@cite_19",
"@cite_27",
"@cite_50",
"@cite_49",
"@cite_5",
"@cite_13",
"@cite_25"
],
"mid": [
"2913636123",
"2777985721",
"2038485832",
"2741156154",
"",
"2396252981",
"2592908675",
"2620265864",
"2232035143",
"2235735853",
"2443846596",
"2763331099"
],
"abstract": [
"Current action anticipation approaches often neglect the intrinsic uncertainty of future predictions when loss functions or evaluation measures are designed. The uncertainty of future observations is especially relevant in the context of egocentric visual data, which is naturally exposed to a great deal of variability. Considering the problem of egocentric action anticipation, we investigate how loss functions and evaluation measures can be designed to explicitly take into account the natural multi-modality of future events. In particular, we discuss suitable measures to evaluate egocentric action anticipation and study how loss functions can be defined to incorporate the uncertainty arising from the prediction of future events. Experiments performed on the EPIC-KITCHENS dataset show that the proposed loss function allows improving the results of both egocentric action anticipation and recognition methods.",
"We address the problem of incrementally modeling and forecasting long-term goals of a first-person camera wearer: what the user will do, where they will go, and what goal they seek. In contrast to prior work in trajectory forecasting, our algorithm, DARKO, goes further to reason about semantic states (will I pick up an object?), and future goal states that are far in terms of both space and time. DARKO learns and forecasts from first-person visual observations of the user’s daily behaviors via an Online Inverse Reinforcement Learning (IRL) approach. Classical IRL discovers only the rewards in a batch setting, whereas DARKO discovers the states, transitions, rewards, and goals of a user from streaming data. Among other results, we show DARKO forecasts goals better than competing methods in both noisy and ideal settings, and our approach is theoretically and empirically no-regret.",
"In this paper, we present a core technology to enable robot recognition of human activities during human-robot interactions. In particular, we propose a methodology for early recognition of activities from robot-centric videos (i.e., first-person videos) obtained from a robot's viewpoint during its interaction with humans. Early recognition, which is also known as activity prediction, is an ability to infer an ongoing activity at its early stage. We present an algorithm to recognize human activities targeting the camera from streaming videos, enabling the robot to predict intended activities of the interacting person as early as possible and take fast reactions to such activities (e.g., avoiding harmful events targeting itself before they actually occur). We introduce the novel concept of'onset' that efficiently summarizes pre-activity observations, and design a recognition approach to consider event history in addition to visual features from first-person videos. We propose to represent an onset using a cascade histogram of time series gradients, and we describe a novel algorithmic setup to take advantage of such onset for early recognition of activities. The experimental results clearly illustrate that the proposed concept of onset enables better earlier recognition of human activities from first-person videos collected with a robot. Categories and Subject Descriptors I.2.10 [Artificial Intelligence]: Vision and Scene Understanding–video analysis; I.4.8 [Image Processing and Computer Vision]: Scene Analysis-motion; I.2.9 [Artificial Intelligence]: Robotics–sensors",
"We introduce a new problem of gaze anticipation on egocentric videos. This substantially extends the conventional gaze prediction problem to future frames by no longer confining it on the current frame. To solve this problem, we propose a new generative adversarial neural network based model, Deep Future Gaze (DFG). DFG generates multiple future frames conditioned on the single current frame and anticipates corresponding future gazes in next few seconds. It consists of two networks: generator and discriminator. The generator uses a two-stream spatial temporal convolution architecture (3D-CNN) explicitly untangling the foreground and the background to generate future frames. It then attaches another 3D-CNN for gaze anticipation based on these synthetic frames. The discriminator plays against the generator by differentiating the synthetic frames of the generator from the real frames. Through competition with discriminator, the generator progressively improves quality of the future frames and thus anticipates future gaze better. Experimental results on the publicly available egocentric datasets show that DFG significantly outperforms all well-established baselines. Moreover, we demonstrate that DFG achieves better performance of gaze prediction on current frames than state-of-the-art methods. This is due to benefiting from learning motion discriminative representations in frame generation. We further contribute a new egocentric dataset (OST) in the object search task. DFG also achieves the best performance for this challenging dataset.",
"",
"We record, and analyze, and present to the community, KrishnaCam, a large (7.6 million frames, 70 hours) egocentric video stream along with GPS position, acceleration and body orientation data spanning nine months of the life of a computer vision graduate student. We explore and exploit the inherent redundancies in this rich visual data stream to answer simple scene understanding questions such as: How much novel visual information does the student see each day? Given a single egocentric photograph of a scene, can we predict where the student might walk next? We find that given our large video database, simple, nearest-neighbor methods are surprisingly adept baselines for these tasks, even in scenes and scenarios where the camera wearer has never been before. For example, we demonstrate the ability to predict the near-future trajectory of the student in broad set of outdoor situations that includes following sidewalks, stopping to wait for a bus, taking a daily path to work, and the lack of movement while eating food.",
"Long-term activity forecasting deals with the problem of predicting how an agent will complete a full activity, defined as a continuous trajectory and a discrete sequence of sub-actions. While previous data-driven methods only dealt with forecasting 2D trajectories, we present a method that leverages common sense prior knowledge and minimal data. In order to forecast the trajectories, we learn a policy function that maps from states to actions the agent should perform next. Through the use of deep reinforcement learning, our method is able to learn a highly non-linear mapping from agent states to actions. We develop the first forecasting framework that uses ego-centric video input, which is an optimal vantage point for understanding human activities over large spaces. Given an annotated first person video sequence for the activity, we construct a 3D point cloud of the environment and activity paths through 3D space. Based on a limited number of examples, we use reinforcement learning to derive a policy for the entire environment, even for areas that have never been visited during the demonstrated examples. We explore the use of deep reinforcement learning to recover a direct mapping from environmental features to best action. Our approach makes it possible to combine a high dimensional continuous state (namely the local point could density surrounding the agent) with a discrete state portion (action stage of an activity) into a single state for behavior forecasting. The result is a policy that generalizes very well from only a few activity samples. We validate our approach on our First-Person Office Behavior Dataset and show that our method of encoding more prior knowledge leads to an increase in forecasting accuracy. We also demonstrate that the deep reinforcement learning approach is able to achieve higher forecasting accuracy than the traditional alternatives.",
"This paper presents an approach to forecast future locations of human hands and objects. Given an image frame, the goal is to predict presence and location of hands and objects in the future frame (e.g., 5 seconds later), even when they are not visible in the current frame. The key idea is that (1) an intermediate representation of a convolutional object recognition model abstracts scene information in its frame and that (2) we can predict (i.e., regress) such representations corresponding to the future frames based on that of the current frame. We design a new two-stream convolutional neural network (CNN) architecture for videos by extending the state-of-the-art convolutional object detection network, and present a new fully convolutional regression network for predicting future scene representations. Our experiments confirm that combining the regressed future representation with our detection network allows reliable estimation of future hands and objects in videos",
"Given a video of an activity, can we predict what will happen next? In this paper we explore two simple tasks related to temporal prediction in egocentric videos of everyday activities. We provide both human experiments to understand how well people can perform on these tasks and computational models for prediction. Experiments indicate that humans and computers can do well on temporal prediction and that personalization to a particular individual or environment provides significantly increased performance. Developing methods for temporal prediction could have far reaching benefits for robots or intelligent agents to anticipate what a person will do, before they do it.",
"We all have experienced forgetting habitual actions among our daily activities. For example, we probably have forgotten to turn the lights off before leaving a room or turn the stove off after cooking. In this paper, we propose a solution to the problem of issuing notifications on actions that may be missed. This involves learning about interdependencies between actions and being able to predict an ongoing action while segmenting the input video stream. In order to show a proof of concept, we collected a new egocentric dataset, in which people wear a camera while making lattes. We show promising results on the extremely challenging task of issuing correct and timely reminders. We also show that our model reliably segments the actions, while predicting the ongoing one when only a few frames from the beginning of the action are observed. The overall prediction accuracy is 46.2 when only 10 frames of an action are seen (2 3 of a sec). Moreover, the overall recognition and segmentation accuracy is shown to be 72.7 when the whole activity sequence is observed. Finally, the online prediction and segmentation accuracy is 68.3 when the prediction is made at every time step.",
"We presents a method for future localization: to predict plausible future trajectories of ego-motion in egocentric stereo images. Our paths avoid obstacles, move between objects, even turn around a corner into space behind objects. As a byproduct of the predicted trajectories, we discover the empty space occluded by foreground objects. One key innovation is the creation of an EgoRetinal map, akin to an illustrated tourist map, that 'rearranges' pixels taking into accounts depth information, the ground plane, and body motion direction, so that it allows motion planning and perception of objects on one image space. We learn to plan trajectories directly on this EgoRetinal map using first person experience of walking around in a variety of scenes. In a testing phase, given an novel scene, we find multiple hypotheses of future trajectories from the learned experience. We refine them by minimizing a cost function that describes compatibility between the obstacles in the EgoRetinal map and trajectories. We quantitatively evaluate our method to show predictive validity and apply to various real world daily activities including walking, shopping, and social interactions.",
"Abstract Although First Person Vision systems can sense the environment from the user’s perspective, they are generally unable to predict his intentions and goals. Since human activities can be decomposed in terms of atomic actions and interactions with objects, intelligent wearable systems would benefit from the ability to anticipate user-object interactions. Even if this task is not trivial, the First Person Vision paradigm can provide important cues to address this challenge. We propose to exploit the dynamics of the scene to recognize next-active-objects before an object interaction begins . We train a classifier to discriminate trajectories leading to an object activation from all others and forecast next-active-objects by analyzing fixed-length trajectory segments within a temporal sliding window. The proposed method compares favorably with respect to several baselines on the Activity of Daily Living (ADL) egocentric dataset comprising 10 h of videos acquired by 20 subjects while performing unconstrained interactions with several objects."
]
} |
1905.09033 | 2953713066 | We propose a network architecture to perform efficient scene understanding. This work presents three main novelties: the first is an Improved Guided Upsampling Module that can replace in toto the decoder part in common semantic segmentation networks. Our second contribution is the introduction of a new module based on spatial sampling to perform Instance Segmentation. It provides a very fast instance segmentation, needing only thresholding as post-processing step at inference time. Finally, we propose a novel efficient network design that includes the new modules and test it against different datasets for outdoor scene understanding. To our knowledge, our network is one of the themost efficient architectures for scene understanding published to date, furthermore being 8.6 more accurate than the fastest competitor on semantic segmentation and almost five times faster than the most efficient network for instance segmentation. | SegNet @cite_31 , one of the first efficiency-oriented architectures together with @cite_21 , introduced an efficient way to exploit high-resolution information by saving max-pooling indices from the encoder and using them during upsampling. ENet authors @cite_27 designed the first highly-efficient architecture by making use of clever design patterns and state-of-the-art building blocks: residual blocks @cite_38 , early downsampling, and 1D factorized convolutions @cite_11 . We think that the use of factorized convolutions is the main reason for the success of this architecture. ERFNet @cite_28 authors implemented an architecture with a very similar overall structure to ENet but with a residual module named Non-Bottleneck-1D module. The overall architecture raises the performance of ENet of a great margin. We will build our fast architecture on top of ERFNet. ESPNet @cite_1 is a very recent, efficient architecture that makes use of a novel module named ESP module. It achieves a very good tradeoff in terms of accuracy and network speed. | {
"cite_N": [
"@cite_38",
"@cite_28",
"@cite_21",
"@cite_1",
"@cite_27",
"@cite_31",
"@cite_11"
],
"mid": [
"2194775991",
"2762439315",
"2907965334",
"2790933182",
"2419448466",
"2963881378",
"2097117768"
],
"abstract": [
"Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57 error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28 relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.",
"Semantic segmentation is a challenging task that addresses most of the perception needs of intelligent vehicles (IVs) in an unified way. Deep neural networks excel at this task, as they can be trained end-to-end to accurately classify multiple object categories in an image at pixel level. However, a good tradeoff between high quality and computational resources is yet not present in the state-of-the-art semantic segmentation approaches, limiting their application in real vehicles. In this paper, we propose a deep architecture that is able to run in real time while providing accurate semantic segmentation. The core of our architecture is a novel layer that uses residual connections and factorized convolutions in order to remain efficient while retaining remarkable accuracy. Our approach is able to run at over 83 FPS in a single Titan X, and 7 FPS in a Jetson TX1 (embedded device). A comprehensive set of experiments on the publicly available Cityscapes data set demonstrates that our system achieves an accuracy that is similar to the state of the art, while being orders of magnitude faster to compute than other architectures that achieve top precision. The resulting tradeoff makes our model an ideal approach for scene understanding in IV applications. The code is publicly available at: https: github.com Eromera erfnet",
"",
"We introduce a fast and efficient convolutional neural network, ESPNet, for semantic segmentation of high resolution images under resource constraints. ESPNet is based on a new convolutional module, efficient spatial pyramid (ESP), which is efficient in terms of computation, memory, and power. ESPNet is 22 times faster (on a standard GPU) and 180 times smaller than the state-of-the-art semantic segmentation network PSPNet, while its category-wise accuracy is only 8 less. We evaluated EPSNet on a variety of semantic segmentation datasets including Cityscapes, PASCAL VOC, and a breast biopsy whole slide image dataset. Under the same constraints on memory and computation, ESPNet outperforms all the current efficient CNN networks such as MobileNet, ShuffleNet, and ENet on both standard metrics and our newly introduced performance metrics that measure efficiency on edge devices. Our network can process high resolution images at a rate of 112 and 9 frames per second on a standard GPU and edge device, respectively.",
"The ability to perform pixel-wise semantic segmentation in real-time is of paramount importance in practical mobile applications. Recent deep neural networks aimed at this task have the disadvantage of requiring a large number of floating point operations and have long run-times that hinder their usability. In this paper, we propose a novel deep neural network architecture named ENet (efficient neural network), created specifically for tasks requiring low latency operation. ENet is up to 18x faster, requires 75x less FLOPs, has 79x less parameters, and provides similar or better accuracy to existing models. We have tested it on CamVid, Cityscapes and SUN datasets and report on comparisons with existing state-of-the-art methods, and the trade-offs between accuracy and processing time of a network. We present performance measurements of the proposed architecture on embedded systems and suggest possible software improvements that could make ENet even faster.",
"We present a novel and practical deep fully convolutional neural network architecture for semantic pixel-wise segmentation termed SegNet. This core trainable segmentation engine consists of an encoder network, a corresponding decoder network followed by a pixel-wise classification layer. The architecture of the encoder network is topologically identical to the 13 convolutional layers in the VGG16 network [1] . The role of the decoder network is to map the low resolution encoder feature maps to full input resolution feature maps for pixel-wise classification. The novelty of SegNet lies is in the manner in which the decoder upsamples its lower resolution input feature map(s). Specifically, the decoder uses pooling indices computed in the max-pooling step of the corresponding encoder to perform non-linear upsampling. This eliminates the need for learning to upsample. The upsampled maps are sparse and are then convolved with trainable filters to produce dense feature maps. We compare our proposed architecture with the widely adopted FCN [2] and also with the well known DeepLab-LargeFOV [3] , DeconvNet [4] architectures. This comparison reveals the memory versus accuracy trade-off involved in achieving good segmentation performance. SegNet was primarily motivated by scene understanding applications. Hence, it is designed to be efficient both in terms of memory and computational time during inference. It is also significantly smaller in the number of trainable parameters than other competing architectures and can be trained end-to-end using stochastic gradient descent. We also performed a controlled benchmark of SegNet and other architectures on both road scenes and SUN RGB-D indoor scene segmentation tasks. These quantitative assessments show that SegNet provides good performance with competitive inference time and most efficient inference memory-wise as compared to other architectures. We also provide a Caffe implementation of SegNet and a web demo at http: mi.eng.cam.ac.uk projects segnet .",
"We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection."
]
} |
1905.09033 | 2953713066 | We propose a network architecture to perform efficient scene understanding. This work presents three main novelties: the first is an Improved Guided Upsampling Module that can replace in toto the decoder part in common semantic segmentation networks. Our second contribution is the introduction of a new module based on spatial sampling to perform Instance Segmentation. It provides a very fast instance segmentation, needing only thresholding as post-processing step at inference time. Finally, we propose a novel efficient network design that includes the new modules and test it against different datasets for outdoor scene understanding. To our knowledge, our network is one of the themost efficient architectures for scene understanding published to date, furthermore being 8.6 more accurate than the fastest competitor on semantic segmentation and almost five times faster than the most efficient network for instance segmentation. | Following @cite_64 we categorize instance segmentation architectures in four different classes: Proposal-based, Recurrent methods, Energy-based and Clustering. @cite_58 @cite_12 are examples of proposal-based approaches. They employ MCG @cite_62 as a class-agnostic predictor of object proposals and they subsequently make use of a classification step to produce instances. A good number of recent works rely on the joint use of an object detector and a semantic segmentation network @cite_24 @cite_20 @cite_52 @cite_7 . As a matter of fact, recent state-of-the-art instance segmentation algorithms @cite_59 @cite_16 are based on enhanced versions of the Faster R-CNN object detector @cite_22 but, from the point-of-view of this work, architectures like PANet @cite_59 or Mask-RCNN @cite_16 , are still computationally too heavy to be employed on edge devices for real-time applications. | {
"cite_N": [
"@cite_64",
"@cite_62",
"@cite_22",
"@cite_7",
"@cite_52",
"@cite_16",
"@cite_24",
"@cite_59",
"@cite_58",
"@cite_20",
"@cite_12"
],
"mid": [
"2744404335",
"1991367009",
"",
"",
"",
"",
"1923115158",
"2963857746",
"1507506748",
"",
"1927486677"
],
"abstract": [
"Semantic instance segmentation remains a challenging task. In this work we propose to tackle the problem with a discriminative loss function, operating at the pixel level, that encourages a convolutional network to produce a representation of the image that can easily be clustered into instances with a simple post-processing step. The loss function encourages the network to map each pixel to a point in feature space so that pixels belonging to the same instance lie close together while different instances are separated by a wide margin. Our approach of combining an off-the-shelf network with a principled loss function inspired by a metric learning objective is conceptually simple and distinct from recent efforts in instance segmentation. In contrast to previous works, our method does not rely on object proposals or recurrent mechanisms. A key contribution of our work is to demonstrate that such a simple setup without bells and whistles is effective and can perform on par with more complex methods. Moreover, we show that it does not suffer from some of the limitations of the popular detect-and-segment approaches. We achieve competitive performance on the Cityscapes and CVPPP leaf segmentation benchmarks.",
"We propose a unified approach for bottom-up hierarchical image segmentation and object candidate generation for recognition, called Multiscale Combinatorial Grouping (MCG). For this purpose, we first develop a fast normalized cuts algorithm. We then propose a high-performance hierarchical segmenter that makes effective use of multiscale information. Finally, we propose a grouping strategy that combines our multiscale regions into highly-accurate object candidates by exploring efficiently their combinatorial space. We conduct extensive experiments on both the BSDS500 and on the PASCAL 2012 segmentation datasets, showing that MCG produces state-of-the-art contours, hierarchical regions and object candidates.",
"",
"",
"",
"",
"The topic of semantic segmentation has witnessed considerable progress due to the powerful features learned by convolutional neural networks (CNNs) [13]. The current leading approaches for semantic segmentation exploit shape information by extracting CNN features from masked image regions. This strategy introduces artificial boundaries on the images and may impact the quality of the extracted features. Besides, the operations on the raw image domain require to compute thousands of networks on a single image, which is time-consuming. In this paper, we propose to exploit shape information via masking convolutional features. The proposal segments (e.g., super-pixels) are treated as masks on the convolutional feature maps. The CNN features of segments are directly masked out from these maps and used to train classifiers for recognition. We further propose a joint method to handle objects and “stuff” (e.g., grass, sky, water) in the same framework. State-of-the-art results are demonstrated on benchmarks of PASCAL VOC and new PASCAL-CONTEXT, with a compelling computational speed.",
"The way that information propagates in neural networks is of great importance. In this paper, we propose Path Aggregation Network (PANet) aiming at boosting information flow in proposal-based instance segmentation framework. Specifically, we enhance the entire feature hierarchy with accurate localization signals in lower layers by bottom-up path augmentation, which shortens the information path between lower layers and topmost feature. We present adaptive feature pooling, which links feature grid and all feature levels to make useful information in each level propagate directly to following proposal subnetworks. A complementary branch capturing different views for each proposal is created to further improve mask prediction. These improvements are simple to implement, with subtle extra computational overhead. Yet they are useful and make our PANet reach the 1st place in the COCO 2017 Challenge Instance Segmentation task and the 2nd place in Object Detection task without large-batch training. PANet is also state-of-the-art on MVD and Cityscapes.",
"We aim to detect all instances of a category in an image and, for each instance, mark the pixels that belong to it. We call this task Simultaneous Detection and Segmentation (SDS). Unlike classical bounding box detection, SDS requires a segmentation and not just a box. Unlike classical semantic segmentation, we require individual object instances. We build on recent work that uses convolutional neural networks to classify category-independent region proposals (R-CNN [16]), introducing a novel architecture tailored for SDS. We then use category-specific, top-down figure-ground predictions to refine our bottom-up proposals. We show a 7 point boost (16 relative) over our baselines on SDS, a 5 point boost (10 relative) over state-of-the-art on semantic segmentation, and state-of-the-art performance in object detection. Finally, we provide diagnostic tools that unpack performance and provide directions for future work.",
"",
"We present a multi-instance object segmentation algorithm to tackle occlusions. As an object is split into two parts by an occluder, it is nearly impossible to group the two separate regions into an instance by purely bottomup schemes. To address this problem, we propose to incorporate top-down category specific reasoning and shape prediction through exemplars into an intuitive energy minimization framework. We perform extensive evaluations of our method on the challenging PASCAL VOC 2012 segmentation set. The proposed algorithm achieves favorable results on the joint detection and segmentation task against the state-of-the-art method both quantitatively and qualitatively."
]
} |
1905.08965 | 2945823804 | Several recent works discussed application-driven image restoration neural networks, which are capable of not only removing noise in images but also preserving their semantic-aware details, making them suitable for various high-level computer vision tasks as the pre-processing step. However, such approaches require extra annotations for their high-level vision tasks, in order to train the joint pipeline using hybrid losses. The availability of those annotations is yet often limited to a few image sets, potentially restricting the general applicability of these methods to denoising more unseen and unannotated images. Motivated by that, we propose a segmentation-aware image denoising model dubbed U-SAID, based on a novel unsupervised approach with a pixel-wise uncertainty loss. U-SAID does not need any ground-truth segmentation map, and thus can be applied to any image dataset. It generates denoised images with comparable or even better quality, and the denoised results show stronger robustness for subsequent semantic segmentation tasks, when compared to either its supervised counterpart or classical "application-agnostic" denoisers. Moreover, we demonstrate the superior generalizability of U-SAID in three-folds, by plugging its "universal" denoiser without fine-tuning: (1) denoising unseen types of images; (2) denoising as pre-processing for segmenting unseen noisy images; and (3) denoising for unseen high-level tasks. Extensive experiments demonstrate the effectiveness, robustness and generalizability of the proposed U-SAID over various popular image sets. | Image denoising has been studied with intensive efforts for decades. Earliest methods refer to various image filters @cite_0 . Later on, many model-based method with various priors have been introduced to this topic, in either spatial or transform domain, or their hybrid, such as spatial smoothness @cite_31 , non-local patch similarity @cite_21 , sparsity @cite_1 @cite_13 @cite_37 and low-rankness @cite_32 . More recently, a number of deep learning models have demonstrated superior performance for image denoising @cite_10 @cite_33 @cite_27 . Despite their encouraging process, most existing denoising algorithms reconstruct images by minimizing the mean square error (MSE), which is well-known to be mis-aligned with human perception quality and often tends to over-smooth textures @cite_40 . Moreover, while image denoising algorithms are often needed as the pre-processing step for the acquired noisy visual data before subsequent high-level visual analytics, their impact on the semantic visual information was much less explored. | {
"cite_N": [
"@cite_37",
"@cite_33",
"@cite_10",
"@cite_21",
"@cite_1",
"@cite_32",
"@cite_0",
"@cite_27",
"@cite_40",
"@cite_31",
"@cite_13"
],
"mid": [
"2075157914",
"2964046669",
"2037642501",
"2056370875",
"2153663612",
"2048695508",
"2099244020",
"2508457857",
"2064076387",
"2103559027",
"2536599074"
],
"abstract": [
"In recent years, sparse signal modeling, especially using the synthesis model has been popular. Sparse coding in the synthesis model is however, NP-hard. Recently, interest has turned to the sparsifying transform model, for which sparse coding is cheap. However, natural images typically contain diverse textures that cannot be sparsified well by a single transform. Hence, in this work, we propose a union of sparsifying transforms model. Sparse coding in this model reduces to a form of clustering. The proposed model is also equivalent to a structured overcomplete sparsifying transform model with block cosparsity, dubbed OCTOBOS. The alternating algorithm introduced for learning such transforms involves simple closed-form solutions. A theoretical analysis provides a convergence guarantee for this algorithm. It is shown to be globally convergent to the set of partial minimizers of the non-convex learning problem. We also show that under certain conditions, the algorithm converges to the set of stationary points of the overall objective. When applied to images, the algorithm learns a collection of well-conditioned square transforms, and a good clustering of patches or textures. The resulting sparse representations for the images are much better than those obtained with a single learned transform, or with analytical transforms. We show the promising performance of the proposed approach in image denoising, which compares quite favorably with approaches involving a single learned square transform or an overcomplete synthesis dictionary, or gaussian mixture models. The proposed denoising method is also faster than the synthesis dictionary based approach.",
"In this paper, we propose a very deep fully convolutional encoding-decoding framework for image restoration such as denoising and super-resolution. The network is composed of multiple layers of convolution and deconvolution operators, learning end-to-end mappings from corrupted images to the original ones. The convolutional layers act as the feature extractor, which capture the abstraction of image contents while eliminating noises corruptions. Deconvolutional layers are then used to recover the image details. We propose to symmetrically link convolutional and deconvolutional layers with skip-layer connections, with which the training converges much faster and attains a higher-quality local optimum. First, the skip connections allow the signal to be back-propagated to bottom layers directly, and thus tackles the problem of gradient vanishing, making training deep networks easier and achieving restoration performance gains consequently. Second, these skip connections pass image details from convolutional layers to deconvolutional layers, which is beneficial in recovering the original image. Significantly, with the large capacity, we can handle different levels of noises using a single model. Experimental results show that our network achieves better performance than recent state-of-the-art methods.",
"Image denoising can be described as the problem of mapping from a noisy image to a noise-free image. The best currently available denoising methods approximate this mapping with cleverly engineered algorithms. In this work we attempt to learn this mapping directly with a plain multi layer perceptron (MLP) applied to image patches. While this has been done before, we will show that by training on large image databases we are able to compete with the current state-of-the-art image denoising methods. Furthermore, our approach is easily adapted to less extensively studied types of noise (by merely exchanging the training data), for which we achieve excellent results as well.",
"We propose a novel image denoising strategy based on an enhanced sparse representation in transform domain. The enhancement of the sparsity is achieved by grouping similar 2D image fragments (e.g., blocks) into 3D data arrays which we call \"groups.\" Collaborative Altering is a special procedure developed to deal with these 3D groups. We realize it using the three successive steps: 3D transformation of a group, shrinkage of the transform spectrum, and inverse 3D transformation. The result is a 3D estimate that consists of the jointly filtered grouped image blocks. By attenuating the noise, the collaborative filtering reveals even the finest details shared by grouped blocks and, at the same time, it preserves the essential unique features of each individual block. The filtered blocks are then returned to their original positions. Because these blocks are overlapping, for each pixel, we obtain many different estimates which need to be combined. Aggregation is a particular averaging procedure which is exploited to take advantage of this redundancy. A significant improvement is obtained by a specially developed collaborative Wiener filtering. An algorithm based on this novel denoising strategy and its efficient implementation are presented in full detail; an extension to color-image denoising is also developed. The experimental results demonstrate that this computationally scalable algorithm achieves state-of-the-art denoising performance in terms of both peak signal-to-noise ratio and subjective visual quality.",
"We address the image denoising problem, where zero-mean white and homogeneous Gaussian additive noise is to be removed from a given image. The approach taken is based on sparse and redundant representations over trained dictionaries. Using the K-SVD algorithm, we obtain a dictionary that describes the image content effectively. Two training options are considered: using the corrupted image itself, or training on a corpus of high-quality image database. Since the K-SVD is limited in handling small image patches, we extend its deployment to arbitrary image sizes by defining a global image prior that forces sparsity over patches in every location in the image. We show how such Bayesian treatment leads to a simple and effective denoising algorithm. This leads to a state-of-the-art denoising performance, equivalent and sometimes surpassing recently published leading alternative denoising methods",
"As a convex relaxation of the low rank matrix factorization problem, the nuclear norm minimization has been attracting significant research interest in recent years. The standard nuclear norm minimization regularizes each singular value equally to pursue the convexity of the objective function. However, this greatly restricts its capability and flexibility in dealing with many practical problems (e.g., denoising), where the singular values have clear physical meanings and should be treated differently. In this paper we study the weighted nuclear norm minimization (WNNM) problem, where the singular values are assigned different weights. The solutions of the WNNM problem are analyzed under different weighting conditions. We then apply the proposed WNNM algorithm to image denoising by exploiting the image nonlocal self-similarity. Experimental results clearly show that the proposed WNNM algorithm outperforms many state-of-the-art denoising algorithms such as BM3D in terms of both quantitative measure and visual perception quality.",
"Bilateral filtering smooths images while preserving edges, by means of a nonlinear combination of nearby image values. The method is noniterative, local, and simple. It combines gray levels or colors based on both their geometric closeness and their photometric similarity, and prefers near values to distant values in both domain and range. In contrast with filters that operate on the three bands of a color image separately, a bilateral filter can enforce the perceptual metric underlying the CIE-Lab color space, and smooth colors and preserve edges in a way that is tuned to human perception. Also, in contrast with standard filtering, bilateral filtering produces no phantom colors along edges in color images, and reduces phantom colors where they appear in the original image.",
"The discriminative model learning for image denoising has been recently attracting considerable attentions due to its favorable denoising performance. In this paper, we take one step forward by investigating the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image denoising. Specifically, residual learning and batch normalization are utilized to speed up the training process as well as boost the denoising performance. Different from the existing discriminative denoising models which usually train a specific model for additive white Gaussian noise at a certain noise level, our DnCNN model is able to handle Gaussian denoising with unknown noise level (i.e., blind Gaussian denoising). With the residual learning strategy, DnCNN implicitly removes the latent clean image in the hidden layers. This property motivates us to train a single DnCNN model to tackle with several general image denoising tasks, such as Gaussian denoising, single image super-resolution, and JPEG image deblocking. Our extensive experiments demonstrate that our DnCNN model can not only exhibit high effectiveness in several general image denoising tasks, but also be efficiently implemented by benefiting from GPU computing.",
"In this paper, we analyse two well-known objective image quality metrics, the peak-signal-to-noise ratio (PSNR) as well as the structural similarity index measure (SSIM), and we derive a simple mathematical relationship between them which works for various kinds of image degradations such as Gaussian blur, additive Gaussian white noise, jpeg and jpeg2000 compression. A series of tests realized on images extracted from the Kodak database gives a better understanding of the similarity and difference between the SSIM and the PSNR.",
"A constrained optimization type of numerical algorithm for removing noise from images is presented. The total variation of the image is minimized subject to constraints involving the statistics of the noise. The constraints are imposed using Lagrange multipliers. The solution is obtained using the gradient-projection method. This amounts to solving a time dependent partial differential equation on a manifold determined by the constraints. As t--- 0o the solution converges to a steady state which is the denoised image. The numerical algorithm is simple and relatively fast. The results appear to be state-of-the-art for very noisy images. The method is noninvasive, yielding sharp edges in the image. The technique could be interpreted as a first step of moving each level set of the image normal to itself with velocity equal to the curvature of the level set divided by the magnitude of the gradient of the image, and a second step which projects the image back onto the constraint set.",
"We propose in this paper to unify two different approaches to image restoration: On the one hand, learning a basis set (dictionary) adapted to sparse signal descriptions has proven to be very effective in image reconstruction and classification tasks. On the other hand, explicitly exploiting the self-similarities of natural images has led to the successful non-local means approach to image restoration. We propose simultaneous sparse coding as a framework for combining these two approaches in a natural manner. This is achieved by jointly decomposing groups of similar signals on subsets of the learned dictionary. Experimental results in image denoising and demosaicking tasks with synthetic and real noise show that the proposed method outperforms the state of the art, making it possible to effectively restore raw images from digital cameras at a reasonable speed and memory cost."
]
} |
1905.08965 | 2945823804 | Several recent works discussed application-driven image restoration neural networks, which are capable of not only removing noise in images but also preserving their semantic-aware details, making them suitable for various high-level computer vision tasks as the pre-processing step. However, such approaches require extra annotations for their high-level vision tasks, in order to train the joint pipeline using hybrid losses. The availability of those annotations is yet often limited to a few image sets, potentially restricting the general applicability of these methods to denoising more unseen and unannotated images. Motivated by that, we propose a segmentation-aware image denoising model dubbed U-SAID, based on a novel unsupervised approach with a pixel-wise uncertainty loss. U-SAID does not need any ground-truth segmentation map, and thus can be applied to any image dataset. It generates denoised images with comparable or even better quality, and the denoised results show stronger robustness for subsequent semantic segmentation tasks, when compared to either its supervised counterpart or classical "application-agnostic" denoisers. Moreover, we demonstrate the superior generalizability of U-SAID in three-folds, by plugging its "universal" denoiser without fine-tuning: (1) denoising unseen types of images; (2) denoising as pre-processing for segmenting unseen noisy images; and (3) denoising for unseen high-level tasks. Extensive experiments demonstrate the effectiveness, robustness and generalizability of the proposed U-SAID over various popular image sets. | Following @cite_20 @cite_26 , we also adopt segmentation as our high-level task, because it can supply pixel-wise feedbacks and is thus considered to be more helpful for dense regression tasks. As pointed out by @cite_28 , the availability of segmentation information can compromise the over-smoothening effects of CNNs across regions and increases their spatial precision. However, we would like to emphasize (again) that while @cite_20 @cite_28 @cite_26 all exploit GT segmentation maps as extra strong information during training, we have only a weaker form of available from the segmentation task, due to the absence of its GT as extra information. Straightforwardly, our methodology is applicable when cascaded with other high-level tasks as well. | {
"cite_N": [
"@cite_28",
"@cite_26",
"@cite_20"
],
"mid": [
"2963321359",
"2796427356",
"2953280900"
],
"abstract": [
"We introduce an approach to integrate segmentation information within a convolutional neural network (CNN). This counter-acts the tendency of CNNs to smooth information across regions and increases their spatial precision. To obtain segmentation information, we set up a CNN to provide an embedding space where region co-membership can be estimated based on Euclidean distance. We use these embeddings to compute a local attention mask relative to every neuron position. We incorporate such masks in CNNs and replace the convolution operation with a “segmentation-aware” variant that allows a neuron to selectively attend to inputs coming from its own region. We call the resulting network a segmentation-aware CNN because it adapts its filters at each image point according to local segmentation cues, while at the same time remaining fully-convolutional. We demonstrate the merit of our method on two widely different dense prediction tasks, that involve classification (semantic segmentation) and regression (optical flow). Our results show that in semantic segmentation we can replace DenseCRF inference with a cascade of segmentation-aware filters, and in optical flow we obtain clearly sharper responses than the ones obtained with comparable networks that do not use segmentation. In both cases segmentation-aware convolution yields systematic improvements over strong baselines.",
"Compressed sensing MRI is a classic inverse problem in the field of computational imaging, accelerating the MR imaging by measuring less k-space data. The deep neural network models provide the stronger representation ability and faster reconstruction compared with \"shallow\" optimization-based methods. However, in the existing deep-based CS-MRI models, the high-level semantic supervision information from massive segmentation-labels in MRI dataset is overlooked. In this paper, we proposed a segmentation-aware deep fusion network called SADFN for compressed sensing MRI. The multilayer feature aggregation (MLFA) method is introduced here to fuse all the features from different layers in the segmentation network. Then, the aggregated feature maps containing semantic information are provided to each layer in the reconstruction network with a feature fusion strategy. This guarantees the reconstruction network is aware of the different regions in the image it reconstructs, simplifying the function mapping. We prove the utility of the cross-layer and cross-task information fusion strategy by comparative study. Extensive experiments on brain segmentation benchmark MRBrainS validated that the proposed SADFN model achieves state-of-the-art accuracy in compressed sensing MRI. This paper provides a novel approach to guide the low-level visual task using the information from mid- or high-level task.",
"Conventionally, image denoising and high-level vision tasks are handled separately in computer vision. In this paper, we cope with the two jointly and explore the mutual influence between them. First we propose a convolutional neural network for image denoising which achieves the state-of-the-art performance. Second we propose a deep neural network solution that cascades two modules for image denoising and various high-level tasks, respectively, and use the joint loss for updating only the denoising network via back-propagation. We demonstrate that on one hand, the proposed denoiser has the generality to overcome the performance degradation of different high-level vision tasks. On the other hand, with the guidance of high-level vision information, the denoising network can generate more visually appealing results. To the best of our knowledge, this is the first work investigating the benefit of exploiting image semantics simultaneously for image denoising and high-level vision tasks via deep learning. The code is available online this https URL."
]
} |
1905.08965 | 2945823804 | Several recent works discussed application-driven image restoration neural networks, which are capable of not only removing noise in images but also preserving their semantic-aware details, making them suitable for various high-level computer vision tasks as the pre-processing step. However, such approaches require extra annotations for their high-level vision tasks, in order to train the joint pipeline using hybrid losses. The availability of those annotations is yet often limited to a few image sets, potentially restricting the general applicability of these methods to denoising more unseen and unannotated images. Motivated by that, we propose a segmentation-aware image denoising model dubbed U-SAID, based on a novel unsupervised approach with a pixel-wise uncertainty loss. U-SAID does not need any ground-truth segmentation map, and thus can be applied to any image dataset. It generates denoised images with comparable or even better quality, and the denoised results show stronger robustness for subsequent semantic segmentation tasks, when compared to either its supervised counterpart or classical "application-agnostic" denoisers. Moreover, we demonstrate the superior generalizability of U-SAID in three-folds, by plugging its "universal" denoiser without fine-tuning: (1) denoising unseen types of images; (2) denoising as pre-processing for segmenting unseen noisy images; and (3) denoising for unseen high-level tasks. Extensive experiments demonstrate the effectiveness, robustness and generalizability of the proposed U-SAID over various popular image sets. | Our work is also broadly related to training deep network with noisy or uncertain annotations @cite_5 @cite_6 . Especially for the segmentation task, existing supervised models require manually labeled segmentations for training. But pixel-based labeling for high-resolution images is often time-consuming and error-prone, causing incorrect pixel-wise annotations. Existing works often consider them as label noise @cite_15 . For example, @cite_25 proposed a noise-tolerant deep model for histopathological image segmentation, using the label-flip noise models proposed in @cite_39 . However, those algorithms still need to be given segmentation maps (though inaccurate), and often demand more statistical estimations of the label noise. | {
"cite_N": [
"@cite_6",
"@cite_39",
"@cite_5",
"@cite_15",
"@cite_25"
],
"mid": [
"2312198368",
"1866072925",
"2577784528",
"2121056381",
"2755446223"
],
"abstract": [
"A weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these ‘free’ tags labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel @math -optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the @math -optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.",
"The availability of large labeled datasets has allowed Convolutional Network models to achieve impressive recognition results. However, in many settings manual annotation of the data is impractical; instead our data has noisy labels, i.e. there is some freely available label for each image which may or may not be accurate. In this paper, we explore the performance of discriminatively-trained Convnets when trained on such noisy data. We introduce an extra noise layer into the network which adapts the network outputs to match the noisy label distribution. The parameters of this noise layer can be estimated as part of the training process and involve simple modifications to current training infrastructures for deep networks. We demonstrate the approaches on several datasets, including large scale experiments on the ImageNet classification benchmark.",
"We present an approach to effectively use millions of images with noisy annotations in conjunction with a small subset of cleanly-annotated images to learn powerful image representations. One common approach to combine clean and noisy data is to first pre-train a network using the large noisy dataset and then fine-tune with the clean dataset. We show this approach does not fully leverage the information contained in the clean set. Thus, we demonstrate how to use the clean annotations to reduce the noise in the large dataset before fine-tuning the network using both the clean set and the full set with reduced noise. The approach comprises a multi-task network that jointly learns to clean noisy annotations and to accurately classify images. We evaluate our approach on the recently released Open Images dataset, containing 9 million images, multiple annotations per image and over 6000 unique classes. For the small clean set of annotations we use a quarter of the validation set with 40k images. Our results demonstrate that the proposed approach clearly outperforms direct fine-tuning across all major categories of classes in the Open Image dataset. Further, our approach is particularly effective for a large number of classes with wide range of noise in annotations (20-80 false positive annotations).",
"Current state-of-the-art deep learning systems for visual object recognition and detection use purely supervised training with regularization such as dropout to avoid overfitting. The performance depends critically on the amount of labeled examples, and in current practice the labels are assumed to be unambiguous and accurate. However, this assumption often does not hold; e.g. in recognition, class labels may be missing; in detection, objects in the image may not be localized; and in general, the labeling may be subjective. In this work we propose a generic way to handle noisy and incomplete labeling by augmenting the prediction objective with a notion of consistency. We consider a prediction consistent if the same prediction is made given similar percepts, where the notion of similarity is between deep network features computed from the input data. In experiments we demonstrate that our approach yields substantial robustness to label noise on several datasets. On MNIST handwritten digits, we show that our model is robust to label corruption. On the Toronto Face Database, we show that our model handles well the case of subjective labels in emotion recognition, achieving state-of-the- art results, and can also benefit from unlabeled face images with no modification to our method. On the ILSVRC2014 detection challenge data, we show that our approach extends to very deep networks, high resolution images and structured outputs, and results in improved scalable detection.",
"Inhomogeneous color distribution and intensity impose major difficulty in fully automated histopathological image (histo-image) segmentation. In this paper, we propose a novel deep learning framework for histo-image segmentation. We innovate a noise-tolerant layer to the output layer of a deep learning image segmentation framework U-Net, which alleviates the requirement of accurately segmented training images and enables “unsupervised” histo-image segmentation by taking noisy segmentation results of traditional image segmentation algorithms as the training outputs. We implement noise-tolerant U-Net for histo-image segmentation to study Duchenne Muscular Dystrophy (DMD), a muscle degenerative disease. Performance comparison with traditional algorithms and the original U-Net demonstrates the great potential of the proposed noise-tolerant U-Net for histo-image segmentation."
]
} |
1905.09010 | 2946510061 | Generative adversarial network (GAN)-based image inpainting methods which utilize coarse-to-fine network with a contextual attention module (CAM) have shown remarkable performance. However, they require numerous computational resources such as convolution operations and network parameters due to two stacked generative networks, which results in a low speed. To address this problem, we propose a novel network structure called PEPSI: parallel extended-decoder path for semantic inpainting network, which aims at not only reducing hardware costs but also improving the inpainting performance. The PEPSI consists of a single shared encoding network and parallel decoding networks with coarse and inpainting paths. The coarse path generates a preliminary inpainting result to train the encoding network for prediction of features for the CAM. At the same time, the inpainting path results in higher inpainting quality with refined features reconstructed using the CAM. In addition, we propose a Diet-PEPSI which significantly reduces the network parameters while maintaining the performance. In the proposed method, we present a Diet-PEPSI unit (DPU) which effectively aggregates the global contextual information with a small number of parameters. Extensive experiments and comparisons with state-of-the-art image inpainting methods demonstrate that both PEPSI and Diet-PEPSI achieve significant improvements in qualitative scores and reduced computation cost. | To inpaint the images with the square hole in arbitrary locations, Iizuka @cite_36 proposed an improved network structure which can sample features in wider receptive fields using multiple dilated convolutional layers. In addition, they use two sibling discriminators: global and local discriminators. The local discriminator focuses on the inpainted region to distinguish local texture consistency while the global discriminator inspects if the result is coherent in a whole image. Yu @cite_38 have extended this work using the coarse-to-fine network and the contextual attention module (CAM). The CAM learns the relation among background and foreground feature patches by computing the cosine similarity. To collect the background features involved with the missing region, this method requires the features at the missing region encoded from roughly completed images. To this end, two stacked generative networks (coarse and refinement) were used to generate an intermediate result of roughly restored image. This method achieved a remarkable performance compared with the recent state-of-the-art inpainting methods; however, it requires considerable computational resources due to the use of the two-stage network structure. | {
"cite_N": [
"@cite_36",
"@cite_38"
],
"mid": [
"2738588019",
"2784790939"
],
"abstract": [
"We present a novel approach for image completion that results in images that are both locally and globally consistent. With a fully-convolutional neural network, we can complete images of arbitrary resolutions by filling-in missing regions of any shape. To train this image completion network to be consistent, we use global and local context discriminators that are trained to distinguish real images from completed ones. The global discriminator looks at the entire image to assess if it is coherent as a whole, while the local discriminator looks only at a small area centered at the completed region to ensure the local consistency of the generated patches. The image completion network is then trained to fool the both context discriminator networks, which requires it to generate images that are indistinguishable from real ones with regard to overall consistency as well as in details. We show that our approach can be used to complete a wide variety of scenes. Furthermore, in contrast with the patch-based approaches such as PatchMatch, our approach can generate fragments that do not appear elsewhere in the image, which allows us to naturally complete the images of objects with familiar and highly specific structures, such as faces.",
"Recent deep learning based approaches have shown promising results on image inpainting for the challenging task of filling in large missing regions in an image. These methods can generate visually plausible image structures and textures, but often create distorted structures or blurry textures inconsistent with surrounding areas. This is mainly due to ineffectiveness of convolutional neural networks in explicitly borrowing or copying information from distant spatial locations. On the other hand, traditional texture and patch synthesis approaches are particularly suitable when it needs to borrow textures from the surrounding regions. Motivated by these observations, we propose a new deep generative model-based approach which can not only synthesize novel image structures but also explicitly utilize surrounding image features as references during network training to make better predictions. The model is a feed-forward, fully convolutional neural network which can process images with multiple holes at arbitrary locations and with variable sizes during the test time. Experiments on multiple datasets including faces, textures and natural images demonstrate that the proposed approach generates higher-quality inpainting results than existing ones. Code and trained models will be released."
]
} |
1905.08945 | 2946260846 | Neural Machine Translation (NMT) has been proven to achieve impressive results. The NMT system translation results depend strongly on the size and quality of parallel corpora. Nevertheless, for many language pairs, no rich-resource parallel corpora exist. As described in this paper, we propose a corpus augmentation method by segmenting long sentences in a corpus using back-translation and generating pseudo-parallel sentence pairs. The experiment results of the Japanese-Chinese and Chinese-Japanese translation with Japanese-Chinese scientific paper excerpt corpus (ASPEC-JC) show that the method improves translation performance. | For back-translation method, implemented their NMT system with iteratively applying back-translation. explored the use of generated back-translated data, aided by denoising with a language model trained on the target side. Translation performance can also be improved by iterative back-translation in both high-resource and low-resource scenarios @cite_5 . A more refined idea of back-translation is the dual learning approach of , which integrates training on parallel data and training on monolingual data via round-tripping. | {
"cite_N": [
"@cite_5"
],
"mid": [
"2963708445"
],
"abstract": [
"A prerequisite for training corpus-based machine translation (MT) systems – either Statistical MT (SMT) or Neural MT (NMT) – is the availability of high-quality parallel data. This is arguably more important today than ever before, as NMT has been shown in many studies to outperform SMT, but mostly when large parallel corpora are available; in cases where data is limited, SMT can still outperform NMT. Recently researchers have shown that back-translating monolingual data can be used to create synthetic parallel corpora, which in turn can be used in combination with authentic parallel data to train a high-quality NMT system. Given that large collections of new parallel text become available only quite rarely, back-translation has become the norm when building state-of-the-art NMT systems, especially in resource-poor scenarios. However, we assert that there are many unknown factors regarding the actual effects of back-translated data on the translation capabilities of an NMT model. Accordingly, in this work we investigate how using back-translated data as a training corpus – both as a separate standalone dataset as well as combined with human-generated parallel data – affects the performance of an NMT model. We use incrementally larger amounts of back-translated data to train a range of NMT systems for German-to-English, and analyse the resulting translation performance."
]
} |
1905.08848 | 2946584487 | When concept drift is detected during classification in a data stream, a common remedy is to retrain a framework's classifier. However, this loses useful information if the classifier has learnt the current concept well, and this concept will recur again in the future. Some frameworks retain and reuse classifiers, but it can be time-consuming to select an appropriate classifier to reuse. These frameworks rarely match the accuracy of state-of-the-art ensemble approaches. For many data stream tasks, speed is important: fast, accurate frameworks are needed for time-dependent applications. We propose the Enhanced Concept Profiling Framework (ECPF), which aims to recognise recurring concepts and reuse a classifier trained previously, enabling accurate classification immediately following a drift. The novelty of ECPF is in how it uses similarity of classifications on new data, between a new classifier and existing classifiers, to quickly identify the best classifier to reuse. It always trains both a new classifier and a reused classifier, and retains the more accurate classifier when concept drift occurs. Finally, it creates a copy of reused classifiers, so a classifier well-suited for a recurring concept will not be impacted by being trained on a different concept. In our experiments, ECPF classifies significantly more accurately than a state-of-the-art classifier reuse framework (Diversity Pool) and a state-of-the-art ensemble technique (Adaptive Random Forest) on synthetic datasets with recurring concepts. It classifies real-world datasets five times faster than Diversity Pool, and six times faster than Adaptive Random Forest and is not significantly less accurate than either. | Ensemble approaches to recurrent learning use multiple classifiers examining a stream, allowing greater chance to find one that functions well at a given time. As a result, they are known to generally outperform single-classifier techniques in terms of accuracy, but to require more time and memory to run. AUE is one example of a current ensemble technique for classifying data streams. It maintains a collection of ten Hoeffding Trees (though other component classifiers can be used), with the weakest performing member regularly being replaced by a new tree, based on performance on recent data. Adaptive Random Forests (ARF) @cite_0 has been shown to be a leading stream classification technique in terms of classification accuracy. This approach builds a diverse set of classifiers, each which can be replaced by a new tree trained on data seen since drift was last detected, according to ADWIN which is run as part of the ensemble. | {
"cite_N": [
"@cite_0"
],
"mid": [
"2626498001"
],
"abstract": [
"Random forests is currently one of the most used machine learning algorithms in the non-streaming (batch) setting. This preference is attributable to its high learning performance and low demands with respect to input preparation and hyper-parameter tuning. However, in the challenging context of evolving data streams, there is no random forests algorithm that can be considered state-of-the-art in comparison to bagging and boosting based algorithms. In this work, we present the adaptive random forest (ARF) algorithm for classification of evolving data streams. In contrast to previous attempts of replicating random forests for data stream learning, ARF includes an effective resampling method and adaptive operators that can cope with different types of concept drifts without complex optimizations for different data sets. We present experiments with a parallel implementation of ARF which has no degradation in terms of classification performance in comparison to a serial implementation, since trees and adaptive operators are independent from one another. Finally, we compare ARF with state-of-the-art algorithms in a traditional test-then-train evaluation and a novel delayed labelling evaluation, and show that ARF is accurate and uses a feasible amount of resources."
]
} |
1905.09000 | 2946761349 | Visual inspection of underwater structures by vehicles, e.g. remotely operated vehicles (ROVs), plays an important role in scientific, military, and commercial sectors. However, the automatic extraction of information using software tools is hindered by the characteristics of water which degrade the quality of captured videos. As a contribution for restoring the color of underwater images, Underwater Denoising Autoencoder (UDAE) model is developed using a denoising autoencoder with U-Net architecture. The proposed network takes into consideration the accuracy and the computation cost to enable real-time implementation on underwater visual tasks using end-to-end autoencoder network. Underwater vehicles perception is improved by reconstructing captured frames; hence obtaining better performance in underwater tasks. Related learning methods use generative adversarial networks (GANs) to generate color corrected underwater images, and to our knowledge this paper is the first to deal with a single autoencoder capable of producing same or better results. Moreover, image pairs are constructed for training the proposed network, where it is hard to obtain such dataset from underwater scenery. At the end, the proposed model is compared to a state-of-the-art method. | Numerous attempts have been made with different image improvement methods for restoring the color of raw underwater images. These methods fall into two categories @cite_0 : hardware-based methods @cite_13 @cite_6 and software-based methods @cite_4 @cite_9 @cite_10 . Software-based methods invert the formation of underwater images and construct physical models for image enhancement in addition to modifying the image pixel values. Hardware-based methods capture multiple images with help of polarization filters, stereo setups or specialized hardware devices and use the obtained additional information @cite_11 @cite_0 . | {
"cite_N": [
"@cite_4",
"@cite_10",
"@cite_9",
"@cite_6",
"@cite_0",
"@cite_13",
"@cite_11"
],
"mid": [
"2763503841",
"1976263166",
"1928779108",
"2155586404",
"2591930867",
"2135001643",
"2145885328"
],
"abstract": [
"We introduce an effective technique to enhance the images captured underwater and degraded due to the medium scattering and absorption. Our method is a single image approach that does not require specialized hardware or knowledge about the underwater conditions or scene structure. It builds on the blending of two images that are directly derived from a color-compensated and white-balanced version of the original degraded image. The two images to fusion, as well as their associated weight maps, are defined to promote the transfer of edges and color contrast to the output image. To avoid that the sharp weight map transitions create artifacts in the low frequency components of the reconstructed image, we also adapt a multiscale fusion strategy. Our extensive qualitative and quantitative evaluation reveals that our enhanced images and videos are characterized by better exposedness of the dark regions, improved global contrast, and edges sharpness. Our validation also proves that our algorithm is reasonably independent of the camera settings, and improves the accuracy of several image processing applications, such as image segmentation and keypoint matching.",
"Light scattering and color change are two major sources of distortion for underwater photography. Light scattering is caused by light incident on objects reflected and deflected multiple times by particles present in the water before reaching the camera. This in turn lowers the visibility and contrast of the image captured. Color change corresponds to the varying degrees of attenuation encountered by light traveling in the water with different wavelengths, rendering ambient underwater environments dominated by a bluish tone. No existing underwater processing techniques can handle light scattering and color change distortions suffered by underwater images, and the possible presence of artificial lighting simultaneously. This paper proposes a novel systematic approach to enhance underwater images by a dehazing algorithm, to compensate the attenuation discrepancy along the propagation path, and to take the influence of the possible presence of an artifical light source into consideration. Once the depth map, i.e., distances between the objects and the camera, is estimated, the foreground and background within a scene are segmented. The light intensities of foreground and background are compared to determine whether an artificial light source is employed during the image capturing process. After compensating the effect of artifical light, the haze phenomenon and discrepancy in wavelength attenuation along the underwater propagation path to camera are corrected. Next, the water depth in the image scene is estimated according to the residual energy ratios of different color channels existing in the background light. Based on the amount of attenuation corresponding to each light wavelength, color change compensation is conducted to restore color balance. The performance of the proposed algorithm for wavelength compensation and image dehazing (WCID) is evaluated both objectively and subjectively by utilizing ground-truth color patches and video downloaded from the Youtube website. Both results demonstrate that images with significantly enhanced visibility and superior color fidelity are obtained by the WCID proposed.",
"Blurring and color cast are two of the most challenging problems for underwater imaging. The poor quality hinders the automatic segmentation or analysis of images. In this paper, we describe an image enhancement method to reduce the blurring and color cast of the underwater medium. It is a two-folded approach; First, a color correction algorithm is applied to correct the color cast and produce a natural appearance of the sub-sea images. Second, a pair of learned dictionaries based on sparse representation are applied to sharpen the image and enhance the details. Our strategy is a single image approach that does not require additional knowledge of environment such as depth, distance object camera or water quality. The experimental results show that the proposed method can efficiently enhance almost every underwater image; And offers a quality that is typically sufficient for the high level computer vision algorithms.",
"Vision in scattering media is important but challenging. Images suffer from poor visibility due to backscattering and attenuation. Most prior methods for scene recovery use active illumination scanners (structured and gated), which can be slow and cumbersome, while natural illumination is inapplicable to dark environments. The current paper addresses the need for a non-scanning recovery method, that uses active scene irradiance. We study the formation of images under widefield artificial illumination. Based on the formation model, the paper presents an approach for recovering the object signal. It also yields rough information about the 3D scene structure. The approach can work with compact, simple hardware, having active widefield, polychromatic polarized illumination. The camera is fitted with a polarization analyzer. Two frames of the scene are taken, with different states of the analyzer or polarizer. A recovery algorithm follows the acquisition. It allows both the backscatter and the object reflection to be partially polarized. It thus unifies and generalizes prior polarization-based methods, which had assumed exclusive polarization of either of these components. The approach is limited to an effective range, due to image noise and illumination falloff. Thus, the limits and noise sensitivity are analyzed. We demonstrate the approach in underwater field experiments.",
"Underwater cameras are widely used to observe the sea floor. They are usually included in autonomous underwater vehicles (AUVs), unmanned underwater vehicles (UUVs), and in situ ocean sensor networks. Despite being an important sensor for monitoring underwater scenes, there exist many issues with recent underwater camera sensors. Because of light’s transportation characteristics in water and the biological activity at the sea floor, the acquired underwater images often suffer from scatters and large amounts of noise. Over the last five years, many methods have been proposed to overcome traditional underwater imaging problems. This paper aims to review the state-of-the-art techniques in underwater image processing by highlighting the contributions and challenges presented in over 40 papers. We present an overview of various underwater image-processing approaches, such as underwater image de-scattering, underwater image color restoration, and underwater image quality assessments. Finally, we summarize the future trends and challenges in designing and processing underwater imaging sensors.",
"Underwater imaging is important for scientific research and technology as well as for popular activities, yet it is plagued by poor visibility conditions. In this paper, we present a computer vision approach that removes degradation effects in underwater vision. We analyze the physical effects of visibility degradation. It is shown that the main degradation effects can be associated with partial polarization of light. Then, an algorithm is presented, which inverts the image formation process for recovering good visibility in images of scenes. The algorithm is based on a couple of images taken through a polarizer at different orientations. As a by-product, a distance map of the scene is also derived. In addition, this paper analyzes the noise sensitivity of the recovery. We successfully demonstrated our approach in experiments conducted in the sea. Great improvements of scene contrast and color correction were obtained, nearly doubling the underwater visibility range.",
"The underwater image processing area has received considerable attention within the last decades, showing important achievements. In this paper we review some of the most recent methods that have been specifically developed for the underwater environment. These techniques are capable of extending the range of underwater imaging, improving image contrast and resolution. After considering the basic physics of the light propagation in the water medium, we focus on the different algorithms available in the literature. The conditions for which each of them have been originally developed are highlighted as well as the quality assessment methods used to evaluate their performance."
]
} |
1905.09000 | 2946761349 | Visual inspection of underwater structures by vehicles, e.g. remotely operated vehicles (ROVs), plays an important role in scientific, military, and commercial sectors. However, the automatic extraction of information using software tools is hindered by the characteristics of water which degrade the quality of captured videos. As a contribution for restoring the color of underwater images, Underwater Denoising Autoencoder (UDAE) model is developed using a denoising autoencoder with U-Net architecture. The proposed network takes into consideration the accuracy and the computation cost to enable real-time implementation on underwater visual tasks using end-to-end autoencoder network. Underwater vehicles perception is improved by reconstructing captured frames; hence obtaining better performance in underwater tasks. Related learning methods use generative adversarial networks (GANs) to generate color corrected underwater images, and to our knowledge this paper is the first to deal with a single autoencoder capable of producing same or better results. Moreover, image pairs are constructed for training the proposed network, where it is hard to obtain such dataset from underwater scenery. At the end, the proposed model is compared to a state-of-the-art method. | , 2018 @cite_8 used two types of networks: Water Generative Adversarial Network (WaterGAN) for generating realistic underwater images and Underwater Image Restoration Network for correcting the color. The generator of WaterGAN models the formation of underwater image using three stages: Attenuation, Scattering, and Camera Model. After that, the learned generator is used to generate training image samples for the color restoration network. First, a relative depth map is estimated and reconstructed from the input image and are both used for color restoration. They showed efficiency for real-time applications, however, their network is limited to certain degradation type appearance due to the way of generating images. Figure shows the images that were used for training the network which do not reflect underwater structures. The clean images consist of in-air images, whereas the corruption process is limited to certain degradation types (e.g. greenish mask). | {
"cite_N": [
"@cite_8"
],
"mid": [
"2593797484"
],
"abstract": [
"This letter reports on WaterGAN, a generative adversarial network (GAN) for generating realistic underwater images from in-air image and depth pairings in an unsupervised pipeline used for color correction of monocular underwater images. Cameras onboard autonomous and remotely operated vehicles can capture high-resolution images to map the seafloor; however, underwater image formation is subject to the complex process of light propagation through the water column. The raw images retrieved are characteristically different than images taken in air due to effects, such as absorption and scattering, which cause attenuation of light at different rates for different wavelengths. While this physical process is well described theoretically, the model depends on many parameters intrinsic to the water column as well as the structure of the scene. These factors make recovery of these parameters difficult without simplifying assumptions or field calibration; hence, restoration of underwater images is a nontrivial problem. Deep learning has demonstrated great success in modeling complex nonlinear systems but requires a large amount of training data, which is difficult to compile in deep sea environments. Using WaterGAN, we generate a large training dataset of corresponding depth, in-air color images, and realistic underwater images. These data serve as input to a two-stage network for color correction of monocular underwater images. Our proposed pipeline is validated with testing on real data collected from both a pure water test tank and from underwater surveys collected in the field. Source code, sample datasets, and pretrained models are made publicly available."
]
} |
1905.09000 | 2946761349 | Visual inspection of underwater structures by vehicles, e.g. remotely operated vehicles (ROVs), plays an important role in scientific, military, and commercial sectors. However, the automatic extraction of information using software tools is hindered by the characteristics of water which degrade the quality of captured videos. As a contribution for restoring the color of underwater images, Underwater Denoising Autoencoder (UDAE) model is developed using a denoising autoencoder with U-Net architecture. The proposed network takes into consideration the accuracy and the computation cost to enable real-time implementation on underwater visual tasks using end-to-end autoencoder network. Underwater vehicles perception is improved by reconstructing captured frames; hence obtaining better performance in underwater tasks. Related learning methods use generative adversarial networks (GANs) to generate color corrected underwater images, and to our knowledge this paper is the first to deal with a single autoencoder capable of producing same or better results. Moreover, image pairs are constructed for training the proposed network, where it is hard to obtain such dataset from underwater scenery. At the end, the proposed model is compared to a state-of-the-art method. | As an improvement over the aforementioned data generation method, , 2018 @cite_1 and , 2018 @cite_15 used CycleGAN @cite_14 for generating underwater images. After synthesizing the data, it was later used for training their color restoration model. | {
"cite_N": [
"@cite_15",
"@cite_14",
"@cite_1"
],
"mid": [
"2964088115",
"2962793481",
"2767009334"
],
"abstract": [
"Autonomous underwater vehicles (AUVs) rely on a variety of sensors - acoustic, inertial and visual - for intelligent decision making. Due to its non-intrusive, passive nature and high information content, vision is an attractive sensing modality, particularly at shallower depths. However, factors such as light refraction and absorption, suspended particles in the water, and color distortion affect the quality of visual data, resulting in noisy and distorted images. AUVs that rely on visual sensing thus face difficult challenges and consequently exhibit poor performance on vision-driven tasks. This paper proposes a method to improve the quality of visual underwater scenes using Generative Adversarial Networks (GANs), with the goal of improving input to vision-driven behaviors further down the autonomy pipeline. Furthermore, we show how recently proposed methods are able to generate a dataset for the purpose of such underwater image restoration. For any visually-guided underwater robots, this improvement can result in increased safety and reliability through robust visual perception. To that effect, we present quantitative and qualitative data which demonstrates that images corrected through the proposed approach generate more visually appealing images, and also provide increased accuracy for a diver tracking algorithm.",
"Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available. We present an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples. Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss. Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa). Qualitative results are presented on several tasks where paired training data does not exist, including collection style transfer, object transfiguration, season transfer, photo enhancement, etc. Quantitative comparisons against several prior methods demonstrate the superiority of our approach.",
"Underwater vision suffers from severe effects due to selective attenuation and scattering when light propagates through water. Such degradation not only affects the quality of underwater images, but limits the ability of vision tasks. Different from existing methods that either ignore the wavelength dependence on the attenuation or assume a specific spectral profile, we tackle color distortion problem of underwater images from a new view. In this letter, we propose a weakly supervised color transfer method to correct color distortion. The proposed method relaxes the need for paired underwater images for training and allows the underwater images being taken in unknown locations. Inspired by cycle-consistent adversarial networks, we design a multiterm loss function including adversarial loss, cycle consistency loss, and structural similarity index measure loss, which makes the content and structure of the outputs same as the inputs, meanwhile the color is similar to the images that were taken without the water. Experiments on underwater images captured under diverse scenes show that our method produces visually pleasing results, even outperforms the state-of-the-art methods. Besides, our method can improve the performance of vision tasks."
]
} |
1905.09000 | 2946761349 | Visual inspection of underwater structures by vehicles, e.g. remotely operated vehicles (ROVs), plays an important role in scientific, military, and commercial sectors. However, the automatic extraction of information using software tools is hindered by the characteristics of water which degrade the quality of captured videos. As a contribution for restoring the color of underwater images, Underwater Denoising Autoencoder (UDAE) model is developed using a denoising autoencoder with U-Net architecture. The proposed network takes into consideration the accuracy and the computation cost to enable real-time implementation on underwater visual tasks using end-to-end autoencoder network. Underwater vehicles perception is improved by reconstructing captured frames; hence obtaining better performance in underwater tasks. Related learning methods use generative adversarial networks (GANs) to generate color corrected underwater images, and to our knowledge this paper is the first to deal with a single autoencoder capable of producing same or better results. Moreover, image pairs are constructed for training the proposed network, where it is hard to obtain such dataset from underwater scenery. At the end, the proposed model is compared to a state-of-the-art method. | The previously mentioned deep learning methods showed good performance in restoring the color. However in certain scenarios, they led to an unrealistic color correction of underwater images as in , 2018 @cite_1 . The training dataset lacked true colors of underwater structures such as coral reefs and fish. Furthermore, a drawback in the color restoration model, Underwater Generative Adversarial Network (UGAN), of , 2018 @cite_15 is the efficiency of real-time implementation with high resolution images, as the model's architecture makes it computationally costly. | {
"cite_N": [
"@cite_15",
"@cite_1"
],
"mid": [
"2964088115",
"2767009334"
],
"abstract": [
"Autonomous underwater vehicles (AUVs) rely on a variety of sensors - acoustic, inertial and visual - for intelligent decision making. Due to its non-intrusive, passive nature and high information content, vision is an attractive sensing modality, particularly at shallower depths. However, factors such as light refraction and absorption, suspended particles in the water, and color distortion affect the quality of visual data, resulting in noisy and distorted images. AUVs that rely on visual sensing thus face difficult challenges and consequently exhibit poor performance on vision-driven tasks. This paper proposes a method to improve the quality of visual underwater scenes using Generative Adversarial Networks (GANs), with the goal of improving input to vision-driven behaviors further down the autonomy pipeline. Furthermore, we show how recently proposed methods are able to generate a dataset for the purpose of such underwater image restoration. For any visually-guided underwater robots, this improvement can result in increased safety and reliability through robust visual perception. To that effect, we present quantitative and qualitative data which demonstrates that images corrected through the proposed approach generate more visually appealing images, and also provide increased accuracy for a diver tracking algorithm.",
"Underwater vision suffers from severe effects due to selective attenuation and scattering when light propagates through water. Such degradation not only affects the quality of underwater images, but limits the ability of vision tasks. Different from existing methods that either ignore the wavelength dependence on the attenuation or assume a specific spectral profile, we tackle color distortion problem of underwater images from a new view. In this letter, we propose a weakly supervised color transfer method to correct color distortion. The proposed method relaxes the need for paired underwater images for training and allows the underwater images being taken in unknown locations. Inspired by cycle-consistent adversarial networks, we design a multiterm loss function including adversarial loss, cycle consistency loss, and structural similarity index measure loss, which makes the content and structure of the outputs same as the inputs, meanwhile the color is similar to the images that were taken without the water. Experiments on underwater images captured under diverse scenes show that our method produces visually pleasing results, even outperforms the state-of-the-art methods. Besides, our method can improve the performance of vision tasks."
]
} |
1905.09000 | 2946761349 | Visual inspection of underwater structures by vehicles, e.g. remotely operated vehicles (ROVs), plays an important role in scientific, military, and commercial sectors. However, the automatic extraction of information using software tools is hindered by the characteristics of water which degrade the quality of captured videos. As a contribution for restoring the color of underwater images, Underwater Denoising Autoencoder (UDAE) model is developed using a denoising autoencoder with U-Net architecture. The proposed network takes into consideration the accuracy and the computation cost to enable real-time implementation on underwater visual tasks using end-to-end autoencoder network. Underwater vehicles perception is improved by reconstructing captured frames; hence obtaining better performance in underwater tasks. Related learning methods use generative adversarial networks (GANs) to generate color corrected underwater images, and to our knowledge this paper is the first to deal with a single autoencoder capable of producing same or better results. Moreover, image pairs are constructed for training the proposed network, where it is hard to obtain such dataset from underwater scenery. At the end, the proposed model is compared to a state-of-the-art method. | We follow the same procedure as in , 2018 @cite_15 for generating synthetic images. However, a different set of images is used for the training of CycleGAN. , 2018 @cite_15 collected clear underwater images and style-transferred the characteristics of degradation from distorted underwater images to them. Our generated dataset is composed of various underwater locations with different degradation types, leading to a better generalization than their network. | {
"cite_N": [
"@cite_15"
],
"mid": [
"2964088115"
],
"abstract": [
"Autonomous underwater vehicles (AUVs) rely on a variety of sensors - acoustic, inertial and visual - for intelligent decision making. Due to its non-intrusive, passive nature and high information content, vision is an attractive sensing modality, particularly at shallower depths. However, factors such as light refraction and absorption, suspended particles in the water, and color distortion affect the quality of visual data, resulting in noisy and distorted images. AUVs that rely on visual sensing thus face difficult challenges and consequently exhibit poor performance on vision-driven tasks. This paper proposes a method to improve the quality of visual underwater scenes using Generative Adversarial Networks (GANs), with the goal of improving input to vision-driven behaviors further down the autonomy pipeline. Furthermore, we show how recently proposed methods are able to generate a dataset for the purpose of such underwater image restoration. For any visually-guided underwater robots, this improvement can result in increased safety and reliability through robust visual perception. To that effect, we present quantitative and qualitative data which demonstrates that images corrected through the proposed approach generate more visually appealing images, and also provide increased accuracy for a diver tracking algorithm."
]
} |
1905.08561 | 2946748368 | Dynamic searchable symmetric encryption (DSSE) is a useful cryptographic tool in encrypted cloud storage. However, it has been reported that DSSE usually suffers from file-injection attacks and content leak of deleted documents. To mitigate these attacks, forward privacy and backward privacy have been proposed. Nevertheless, the existing forward backward-private DSSE schemes can only support single keyword queries. To address this problem, in this paper, we propose two DSSE schemes supporting range queries. One is forward-private and supports a large number of documents. The other can achieve backward privacy, while it can only support a limited number of documents. Finally, we also give the security proofs of the proposed DSSE schemes in the random oracle model. | @cite_7 were the first using symmetric encryption to facilitate keyword search over the encrypted data. Later, @cite_6 gave a formal definition for SSE and the corresponding security model in the static setting. To make SSE more scalable and expressive, @cite_16 proposed a new scalable SSE supporting Boolean queries. Following this construction, many extensions have been proposed. @cite_10 extended it to process a much richer collection of queries. For instance, they used a binary tree with keywords labelled according to the tree levels to support range queries. @cite_2 made another extension to support general Boolean queries. 's construction has also been extended into multi-user setting @cite_8 @cite_21 @cite_4 . However, the above schemes cannot support data update. To solve this problem, some DSSE schemes have been proposed @cite_15 @cite_1 . | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_21",
"@cite_1",
"@cite_6",
"@cite_2",
"@cite_15",
"@cite_16",
"@cite_10"
],
"mid": [
"2766544523",
"2147929033",
"",
"2620403523",
"2086763678",
"2146828512",
"2586817631",
"2152516507",
"1502708590",
"2294724888"
],
"abstract": [
"The single-writer multi-reader searchable encryption (SMSE) allows an arbitrary authorized user to submit a valid search token and get the corresponding encrypted identifiers. In order to achieve fine-grained access control, the identifiers are encrypted by the attribute-based encryption. In this case, the user can decrypt a ciphertext only when the access policy in it matches the user’s attribute set. However, the server unable to determine whether the user can decrypt a certain ciphertext without the knowledge of the user’s attribute set. As a result, all the ciphertexts based on a search token have to be returned to the user, which causes unnecessary communication and decryption costs. In this paper, we propose a new SMSE scheme, in which the server just needs to return the ones which can be decrypted by the user rather than the whole search results. In order to achieve this goal, we present a server-side match technique with which the server can test whether the user can decrypt a ciphertext without knowing the user’s attribute set. Furthermore, the decryption computation is very efficient, irrespective of the structure of access policy. Therefore, both the communication and decryption overheads are dramatically reduced in our scheme.",
"It is desirable to store data on data storage servers such as mail servers and file servers in encrypted form to reduce security and privacy risks. But this usually implies that one has to sacrifice functionality for security. For example, if a client wishes to retrieve only documents containing certain words, it was not previously known how to let the data storage server perform the search and answer the query, without loss of data confidentiality. We describe our cryptographic schemes for the problem of searching on encrypted data and provide proofs of security for the resulting crypto systems. Our techniques have a number of crucial advantages. They are provably secure: they provide provable secrecy for encryption, in the sense that the untrusted server cannot learn anything about the plaintext when only given the ciphertext; they provide query isolation for searches, meaning that the untrusted server cannot learn anything more about the plaintext than the search result; they provide controlled searching, so that the untrusted server cannot search for an arbitrary word without the user's authorization; they also support hidden queries, so that the user may ask the untrusted server to search for a secret word without revealing the word to the server. The algorithms presented are simple, fast (for a document of length n, the encryption and search algorithms only need O(n) stream cipher and block cipher operations), and introduce almost no space and communication overhead, and hence are practical to use today.",
"",
"We propose a multi-user Symmetric Searchable Encryption (SSE) scheme based on the single-user Oblivious Cross Tags (OXT) protocol (, CRYPTO 2013). The scheme allows any user to perform a search query by interacting with the server and any ( -1 ) ‘helping’ users, and preserves the privacy of database content against the server even assuming leakage of up to ( -1 ) users’ keys to the server (for a threshold parameter ( )), while hiding the query from the ( -1 ) ‘helping users’. To achieve the latter query privacy property, we design a new distributed key-homomorphic pseudorandom function (PRF) that hides the PRF input (search keyword) from the ‘helping’ key share holders. By distributing the utilized keys among the users, the need of constant online presence of the data owner to provide services to the users is eliminated, while providing resilience against user key exposure.",
"We design and implement dynamic symmetric searchable encryption schemes that efficiently and privately search server-held encrypted databases with tens of billions of record-keyword pairs. Our basic theoretical construction supports single-keyword searches and offers asymptotically optimal server index size, fully parallel searching, and minimal leakage. Our implementation effort brought to the fore several factors ignored by earlier coarse-grained theoretical performance analyses, including lowlevel space utilization, I O parallelism and goodput. We accordingly introduce several optimizations to our theoretically optimal construction that model the prototype’s characteristics designed to overcome these factors. All of our schemes and optimizations are proven secure and the information leaked to the untrusted server is precisely quantified. We evaluate the performance of our prototype using two very large datasets: a synthesized census database with 100 million records and hundreds of keywords per record and a multi-million webpage collection that includes Wikipedia as a subset. Moreover, we report on an implementation that uses the dynamic SSE schemes developed here as the basis for supporting recent SSE advances, including complex search queries (e.g., Boolean queries) and richer operational settings (e.g., query delegation), in the above terabyte-scale databases.",
"Searchable symmetric encryption (SSE) allows a party to outsource the storage of its data to another party (a server) in a private manner, while maintaining the ability to selectively search over it. This problem has been the focus of active research in recent years. In this paper we show two solutions to SSE that simultaneously enjoy the following properties: Both solutions are more efficient than all previous constant-round schemes. In particular, the work performed by the server per returned document is constant as opposed to linear in the size of the data. Both solutions enjoy stronger security guarantees than previous constant-round schemes. In fact, we point out subtle but serious problems with previous notions of security for SSE, and show how to design constructions which avoid these pitfalls. Further, our second solution also achieves what we call adaptive SSE security, where queries to the server can be chosen adaptively (by the adversary) during the execution of the search; this notion is both important in practice and has not been previously considered. Surprisingly, despite being more secure and more efficient, our SSE schemes are remarkably simple. We consider the simplicity of both solutions as an important step towards the deployment of SSE technologies.As an additional contribution, we also consider multi-user SSE. All prior work on SSE studied the setting where only the owner of the data is capable of submitting search queries. We consider the natural extension where an arbitrary group of parties other than the owner can submit search queries. We formally define SSE in the multi-user setting, and present an efficient construction that achieves better performance than simply using access control mechanisms.",
"A Searchable Symmetric Encryption (SSE) scheme allows a server to search a user's data without having to decrypt the data. This provides the user with a high degree of privacy and is particularly useful when data is stored on Cloud. Numerous SSE schemes have already been proposed and while most have excellent security properties, few meet high performance requirements and most only support searching for a single keyword at a time. The SSE scheme of (CRYPTO 2013) is notable for its high efficiency on restricted forms of Boolean queries, but has low efficiency for, or does not support, other common forms of boolean queries. In this paper, we propose a generalization and optimization of the SSE scheme, which extends to support a much larger class of boolean queries, and performs no worse, and often with much higher efficiency than the scheme for the remaining queries.",
"Searchable symmetric encryption (SSE) allows a client to encrypt its data in such a way that this data can still be searched. The most immediate application of SSE is to cloud storage, where it enables a client to securely outsource its data to an untrusted cloud provider without sacrificing the ability to search over it. SSE has been the focus of active research and a multitude of schemes that achieve various levels of security and efficiency have been proposed. Any practical SSE scheme, however, should (at a minimum) satisfy the following properties: sublinear search time, security against adaptive chosen-keyword attacks, compact indexes and the ability to add and delete files efficiently. Unfortunately, none of the previously-known SSE constructions achieve all these properties at the same time. This severely limits the practical value of SSE and decreases its chance of deployment in real-world cloud storage systems. To address this, we propose the first SSE scheme to satisfy all the properties outlined above. Our construction extends the inverted index approach (, CCS 2006) in several non-trivial ways and introduces new techniques for the design of SSE. In addition, we implement our scheme and conduct a performance evaluation, showing that our approach is highly efficient and ready for deployment.",
"This work presents the design and analysis of the first searchable symmetric encryption (SSE) protocol that supports conjunctive search and general Boolean queries on outsourced symmetrically- encrypted data and that scales to very large databases and arbitrarily-structured data including free text search. To date, work in this area has focused mainly on single-keyword search. For the case of conjunctive search, prior SSE constructions required work linear in the total number of documents in the database and provided good privacy only for structured attribute-value data, rendering these solutions too slow and inflexible for large practical databases.",
"We extend the searchable symmetric encryption (SSE) protocol of [, Crypto’13] adding support for range, substring, wildcard, and phrase queries, in addition to the Boolean queries supported in the original protocol. Our techniques apply to the basic single-client scenario underlying the common SSE setting as well as to the more complex Multi-Client and Outsourced Symmetric PIR extensions of [, CCS’13]. We provide performance information based on our prototype implementation, showing the practicality and scalability of our techniques to very large databases, thus extending the performance results of [, NDSS’14] to these rich and comprehensive query types."
]
} |
1905.08561 | 2946748368 | Dynamic searchable symmetric encryption (DSSE) is a useful cryptographic tool in encrypted cloud storage. However, it has been reported that DSSE usually suffers from file-injection attacks and content leak of deleted documents. To mitigate these attacks, forward privacy and backward privacy have been proposed. Nevertheless, the existing forward backward-private DSSE schemes can only support single keyword queries. To address this problem, in this paper, we propose two DSSE schemes supporting range queries. One is forward-private and supports a large number of documents. The other can achieve backward privacy, while it can only support a limited number of documents. Finally, we also give the security proofs of the proposed DSSE schemes in the random oracle model. | However, designing a secure DSSE scheme is not an easy job. @cite_11 pointed out that only a small leakage leveraged by the adversary would be enough to compromise the privacy of clients' queries. A concrete attack named file-injection attack was proposed by @cite_13 . In this attack, the adversary can infer the concept of aclient queries by injecting a small portion of new documents into encrypted database. This attack also highlights the need for forward privacy which protects security of new added parts. Accordingly, we have backward privacy that protects security of new added parts and later deleted. These two security notions were first introduced by @cite_0 . The formal definitions of forward backward privacy for DSSE were given by Bost @cite_19 and @cite_14 , respectively. In @cite_19 , Bost also proposed a concrete forward-private DSSE scheme, it does not support physical deletion. Later on, @cite_5 proposed a forward-private DSSE scheme supporting physical deletion. Meanwhile, @cite_14 proposed a forward backward-private DSSE to reduce leakage during deletion. Unfortunately, all the existing forward backward-private DSSE schemes only support single keyword queries. Hence, forward backward-private DSSE supporting more expressive queries, such as range queries, are quite desired. | {
"cite_N": [
"@cite_14",
"@cite_0",
"@cite_19",
"@cite_5",
"@cite_13",
"@cite_11"
],
"mid": [
"2765463836",
"2016575873",
"2539385377",
"2766707306",
"2472177816",
"2033165262"
],
"abstract": [
"Using dynamic Searchable Symmetric Encryption, a user with limited storage resources can securely outsource a database to an untrusted server, in such a way that the database can still be searched and updated efficiently. For these schemes, it would be desirable that updates do not reveal any information a priori about the modifications they carry out, and that deleted results remain inaccessible to the server a posteriori. If the first property, called forward privacy, has been the main motivation of recent works, the second one, backward privacy, has been overlooked. In this paper, we study for the first time the notion of backward privacy for searchable encryption. After giving formal definitions for different flavors of backward privacy, we present several schemes achieving both forward and backward privacy, with various efficiency trade-offs. Our constructions crucially rely on primitives such as constrained pseudo-random functions and puncturable encryption schemes. Using these advanced cryptographic primitives allows for a fine-grained control of the power of the adversary, preventing her from evaluating functions on selected inputs, or decrypting specific ciphertexts. In turn, this high degree of control allows our SSE constructions to achieve the stronger forms of privacy outlined above. As an example, we present a framework to construct forward-private schemes from range-constrained pseudo-random functions. Finally, we provide experimental results for implementations of our schemes, and study their practical efficiency.",
"In this paper we revisit the DSSE problem. We propose the first DSSE scheme that achieves the best of both worlds, i.e., both small leakage and efficiency. In particular, our DSSE scheme leaks significantly less information than any other previous DSSE construction and supports both updates and searches in sublinear time in the worst case, maintaining at the same time a data structure of only linear size. We finally provide an implementation of our construction, showing its practical efficiency.",
"Searchable Symmetric Encryption aims at making possible searching over an encrypted database stored on an untrusted server while keeping privacy of both the queries and the data, by allowing some small controlled leakage to the server. Recent work shows that dynamic schemes -- in which the data is efficiently updatable -- leaking some information on updated keywords are subject to devastating adaptative attacks breaking the privacy of the queries. The only way to thwart this attack is to design forward private schemes whose update procedure does not leak if a newly inserted element matches previous search queries. This work proposes Sophos as a forward private SSE scheme with performance similar to existing less secure schemes, and that is conceptually simpler (and also more efficient) than previous forward private constructions. In particular, it only relies on trapdoor permutations and does not use an ORAM-like construction. We also explain why Sophos is an optimal point of the security performance tradeoff for SSE. Finally, an implementation and evaluation results demonstrate its practical efficiency.",
"The recently proposed file-injection type attacks are highlighting the importance of forward security in dynamic searchable symmetric encryption (DSSE). Forward security enables to thwart those attacks by hiding the information about the newly added files matching a previous search query. However, there are still only a few DSSE schemes that provide forward security, and they have factors that hinder efficiency. In particular, all of these schemes do not support actual data deletion, which increments both storage space and computational complexity. In this paper, we design and implement a forward secure DSSE scheme with optimal search and update complexity, for both computation and communication point of view. As a starting point, we propose a new, simple, theoretical data structure, called dual dictionary that can take advantage of both the inverted and the forward indexes at the same time. This data structure allows to delete data explicitly and in real time, which greatly improves efficiency compared to previous works. In addition, our scheme provides forward security by encrypting the newly added data with fresh keys not related with the previous search tokens. We implemented our scheme for Enron email and Wikipedia datasets and measured its performance. The comparison with Sophos shows that our scheme is very efficient in practice, for both searches and updates in dynamic environments.",
"The goal of searchable encryption (SE) is to enable a client to execute searches over encrypted files stored on an untrusted server while ensuring some measure of privacy for both the encrypted files and the search queries. Most recent research has focused on developing efficient SE schemes at the expense of allowing some small, wellcharacterized “(information) leakage” to the server about the files and or the queries. The practical impact of this leakage, however, remains unclear. We thoroughly study file-injection attacks—in which the server sends files to the client that the client then encrypts and stores—on the query privacy of singlekeyword and conjunctive SE schemes. We show such attacks can reveal the client’s queries in their entirety using very few injected files, even for SE schemes having low leakage. We also demonstrate that natural countermeasures for preventing file-injection attacks can be easily circumvented. Our attacks outperform prior work significantly in terms of their effectiveness as well as in terms of their assumptions about the attacker’s prior knowledge.",
"Schemes for secure outsourcing of client data with search capability are being increasingly marketed and deployed. In the literature, schemes for accomplishing this efficiently are called Searchable Encryption (SE). They achieve high efficiency with provable security by means of a quantifiable leakage profile. However, the degree to which SE leakage can be exploited by an adversary is not well understood. To address this, we present a characterization of the leakage profiles of in-the-wild searchable encryption products and SE schemes in the literature, and present attack models based on an adversarial server's prior knowledge. Then we empirically investigate the security of searchable encryption by providing query recovery and plaintext recovery attacks that exploit these leakage profiles. We term these leakage-abuse attacks and demonstrate their effectiveness for varying leakage profiles and levels of server knowledge, for realistic scenarios. Amongst our contributions are realistic active attacks which have not been previously explored."
]
} |
1905.08561 | 2946748368 | Dynamic searchable symmetric encryption (DSSE) is a useful cryptographic tool in encrypted cloud storage. However, it has been reported that DSSE usually suffers from file-injection attacks and content leak of deleted documents. To mitigate these attacks, forward privacy and backward privacy have been proposed. Nevertheless, the existing forward backward-private DSSE schemes can only support single keyword queries. To address this problem, in this paper, we propose two DSSE schemes supporting range queries. One is forward-private and supports a large number of documents. The other can achieve backward privacy, while it can only support a limited number of documents. Finally, we also give the security proofs of the proposed DSSE schemes in the random oracle model. | Apart from the binary tree technique, order preserving encryption (OPE) can also be used to support range queries. The concept of OPE was proposed by @cite_17 , and it allows the order of the plaintexts to be preserved in the ciphertexts. It is easy to see that this kind of encryption would lead to the leakage in @cite_23 @cite_12 . To reduce this leakage, @cite_18 proposed another concept named order revealing encryption (ORE), where the order of the ciphertexts are revealed by using an algorithm rather than comparing the ciphertexts (in OPE) directly. More efficient ORE schemes were proposed later @cite_9 . However, ORE-based SSE still leaks much information about the underlying plaintexts. To avoid this, in this paper, we focus on how to use the binary tree structure to achieve range queries. | {
"cite_N": [
"@cite_18",
"@cite_9",
"@cite_23",
"@cite_12",
"@cite_17"
],
"mid": [
"754106230",
"2400905995",
"1895952394",
"2163992091",
"2154496743"
],
"abstract": [
"Deciding “greater-than” relations among data items just given their encryptions is at the heart of search algorithms on encrypted data, most notably, non-interactive binary search on encrypted data. Order-preserving encryption provides one solution, but provably provides only limited security guarantees. Two-input functional encryption is another approach, but requires the full power of obfuscation machinery and is currently not implementable.",
"In an order-preserving encryption scheme, the encryption algorithm produces ciphertexts that preserve the order of their plaintexts. Order-preserving encryption schemes have been studied intensely in the last decade, and yet not much is known about the security of these schemes. Very recently, Boneh eti¾?al. Eurocrypti¾?2015 introduced a generalization of order-preserving encryption, called order-revealing encryption, and presented a construction which achieves this notion with best-possible security. Because their construction relies on multilinear maps, it is too impractical for most applications and therefore remains a theoretical result. In this work, we build efficiently implementable order-revealing encryption from pseudorandom functions. We present the first efficient order-revealing encryption scheme which achieves a simulation-based security notion with respect to a leakage function that precisely quantifies what is leaked by the scheme. In fact, ciphertexts in our scheme are only about 1.6 times longer than their plaintexts. Moreover, we show how composing our construction with existing order-preserving encryption schemes results in order-revealing encryption that is strictly more secure than all preceding order-preserving encryption schemes.",
"We initiate the cryptographic study of order-preserving symmetric encryption (OPE), a primitive suggested in the database community by (SIGMOD '04) for allowing efficient range queries on encrypted data. Interestingly, we first show that a straightforward relaxation of standard security notions for encryption such as indistinguishability against chosen-plaintext attack (IND-CPA) is unachievable by a practical OPE scheme. Instead, we propose a security notion in the spirit of pseudorandom functions (PRFs) and related primitives asking that an OPE scheme look \"as-random-as-possible\" subject to the order-preserving constraint. We then design an efficient OPE scheme and prove its security under our notion based on pseudorandomness of an underlying blockcipher. Our construction is based on a natural relation we uncover between a random order-preserving function and the hypergeometric probability distribution. In particular, it makes black-box use of an efficient sampling algorithm for the latter.",
"We further the study of order-preserving symmetric encryption (OPE), a primitive for allowing efficient range queries on encrypted data, recently initiated (from a cryptographic perspective) by (Eurocrypt'09). First, we address the open problem of characterizing what encryption via a random order-preserving function (ROPF) leaks about underlying data (ROPF being the \"ideal object\" in the security definition, POPF, satisfied by their scheme.) In particular, we show that, for a database of randomly distributed plaintexts and appropriate choice of parameters, ROPF encryption leaks neither the precise value of any plaintext nor the precise distance between any two of them. The analysis here is quite technically non-trivial and introduces useful new techniques. On the other hand, we also show that ROPF encryption does leak both the value of any plaintext as well as the distance between any two plaintexts to within a range of possibilities roughly the square root of the domain size. We then study schemes that are not order-preserving, but which nevertheless allow efficient range queries and achieve security notions stronger than POPF. In a setting where the entire database is known in advance of key-generation (considered in several prior works), we show that recent constructions of \"monotone minimal perfect hash functions\" allow to efficiently achieve (an adaptation of) the notion of IND-O(rdered) CPA also considered by , which asks that only the order relations among the plaintexts is leaked. Finally, we introduce modular order-preserving encryption (MOPE), in which the scheme of is prepended with a shift cipher. MOPE improves the security of OPE in a sense, as it does not leak any information about plaintext location. We clarify that our work should not be interpreted as saying the original scheme of , or the variants that we introduce, are \"secure\" or \"insecure.\" Rather, the goal of this line of research is to help practitioners decide whether the options provide a suitable security-functionality tradeoff for a given application.",
"Encryption is a well established technology for protecting sensitive data. However, once encrypted, data can no longer be easily queried aside from exact matches. We present an order-preserving encryption scheme for numeric data that allows any comparison operation to be directly applied on encrypted data. Query results produced are sound (no false hits) and complete (no false drops). Our scheme handles updates gracefully and new values can be added without requiring changes in the encryption of other values. It allows standard databse indexes to be built over encrypted tables and can easily be integrated with existing database systems. The proposed scheme has been designed to be deployed in application environments in which the intruder can get access to the encrypted database, but does not have prior domain information such as the distribution of values and annot encrypt or decrypt arbitrary values of his choice. The encryption is robust against estimation of the true value in such environments."
]
} |
1905.08581 | 2944872360 | In this paper, we demonstrate a data-driven methodology for modelling the local similarity measures of various attributes in a dataset. We analyse the spread in the numerical attributes and estimate their distribution using polynomial function to showcase an approach for deriving strong initial value ranges of numerical attributes and use a non-overlapping distribution for categorical attributes such that the entire similarity range [0,1] is utilized. We use an open source dataset for demonstrating modelling and development of the similarity measures and will present a case-based reasoning (CBR) system that can be used to search for the most relevant similar cases. | Similar to the preference-based similarity measure development framework presented by authors in @cite_1 @cite_3 , we are presenting a framework for modelling local similarity measures based on the data set available. Therewith we can tailor each similarity measure to the application domain. Using a data-driven approach for automatic similarity learning and feature weighting has been presented by Gabel and Godehardt @cite_5 where they trained a neural network to induce local and global similarity measures @cite_2 . While we are not automatically assigning the similarity measures, we use the existing cases to derive them. | {
"cite_N": [
"@cite_5",
"@cite_1",
"@cite_3",
"@cite_2"
],
"mid": [
"2253522548",
"2136629812",
"2182643186",
"37417951"
],
"abstract": [
"The automatic acquisition of a similarity measure for a CBR system is appealing as it frees the system designer from the tedious task of defining it manually. However, acquiring similarity measures with some machine learning approach typically results in some black box representation of similarity whose magic-like combination of high precision and low explainability may decrease a human user’s trust in the system. In this paper, we target this problem by suggesting a method to induce a human-readable and easily understandable – and thus potentially trustworthy – representation of similarity from a previously learned black box-like representation of similarity measures. Our experimental evaluations support the claim that, given some highly precise learned similarity measure, we can induce a less powerful, but human-understandable representation of it while its corresponding level of accuracy is only marginally impaired.",
"Building on recent research on preference handling in artificial intelligence and related fields, our general goal is to develop a coherent and universally applicable methodological framework for CBR on the basis of formal concepts and methods for knowledge representation and reasoning with preferences. A preference-based approach to CBR appears to be appealing for several reasons, notably because case-based experiences naturally lend themselves to representations in terms of preference relations, even when not dealing with preference information in a literal sense. Moreover, the flexibility and expressiveness of a preference-based formalism well accommodate the uncertain and approximate nature of case-based problem solving. In this paper, we make a first step toward a preference-based formalization of CBR. Apart from providing a general outline of the framework as a whole, we specifically address the step of case-based inference. The latter consists of inferring preferences for candidate solutions in the context of a new problem, given such preferences in similar situations. Our case-based approach to predicting preference models is concretely realized for a scenario in which solutions are represented in the form of subsets of a reference set. First experimental results are presented to demonstrate the effectiveness of this approach.",
"This paper is a continuation of our recent work on preference-based CBR, or Pref-CBR for short. The latter is conceived as a case-based reasoning methodology in which problem solving experience is represented in the form of contextualized preferences, namely preferences for candidate solutions in the context of a target problem to be solved. In our Pref-CBR framework, case-based problem solving is formalized as a preference-guided search process in the space of candidate solutions, which is equipped with a similarity (or, equivalently, a distance) measure. Since the efficacy of Pref-CBR is influenced by the adequacy of this measure, we propose a learning method for adapting solution similarity on the basis of experience gathered by the CBR system in the course of time. More specifically, our method makes use of an underlying probabilistic model and realizes adaptation as Bayesian inference. The effectiveness of this method is illustrated in a case study that deals with the case-based recommendation of red wines.",
"Of which nature is the knowledge a similarity measure can contain? How to bring the knowledge into the measure? How to retrieve and use the knowledge for actual problems?"
]
} |
1905.08609 | 2958161068 | We address a problem of estimating pose of a person's head from its RGB image. The employment of CNNs for the problem has contributed to significant improvement in accuracy in recent works. However, we show that the following two methods, despite their simplicity, can attain further improvement: (i) proper adjustment of the margin of bounding box of a detected face, and (ii) choice of loss functions. We show that the integration of these two methods achieve the new state-of-the-art on standard benchmark datasets for in-the-wild head pose estimation. | The head pose estimation problem has been investigated using various approaches. Murphy- @cite_13 reviewed methods which use additional sensors such as laser pointers, magnetic and inertial sensors to get a stable estimation in controlled situations. In addition, some methods apply appearance template models to compare test images with a set of pose exemplars @cite_24 @cite_26 @cite_11 or train multiple face detectors for different head poses @cite_17 @cite_29 . The most recently proposed head pose estimation methods can be categorized into two groups by input image types: RGB-D image based methods and RGB image based methods. | {
"cite_N": [
"@cite_26",
"@cite_11",
"@cite_29",
"@cite_24",
"@cite_13",
"@cite_17"
],
"mid": [
"2051439328",
"1982214249",
"1734489760",
"2079054988",
"",
"2131227845"
],
"abstract": [
"Identity-independent estimation of head pose from prototype images is a perplexing task, requiring pose-invariant face detection. The problem is exacerbated by changes in illumination, identity and facial position. Facial images must be transformed in such a way as to emphasise di erences in pose, while suppressing di erences in identity. We investigate appropriate transformations for use with a similarityto-prototypes philosophy. The results show that orientation-selective Gabor lters enhance di erences in pose, and that di erent lter orientations are optimal at di erent poses. In contrast, PCA was found to provide an identity-invariant representation in which similarities can be calculated more robustly. We also investigate the angular resolution at which pose changes can be resolved using our methods. An angular resolution of 10 was found to be su ciently discriminable at some poses but not at others, while 20 is quite acceptable at most poses.",
"Abstract Real-time identity-independent estimation of head pose from prototype images is a perplexing task requiring pose-invariant face detection. The problem is exacerbated by changes in illumination, identity and facial position. We approach the problem using a view-based statistical learning technique based on similarity of images to prototypes. For this method to be effective, facial images must be transformed in such a way as to emphasise differences in pose while suppressing differences in identity. We investigate appropriate transformations for use with a similarity-to-prototypes philosophy. The results show that orientation-selective Gabor filters enhance differences in pose and that different filter orientations are optimal at different poses. In contrast, principal component analysis (PCA) was found to provide an identity-invariant representation in which similarities can be calculated more robustly. We also investigate the angular resolution at which pose changes can be resolved using our methods. An angular resolution of 10° was found to be sufficiently discriminable at some poses but not at others, while 20° is quite acceptable at most poses.",
"Head pose estimation in low resolution is a challenge problem. Traditional pose estimation algorithms, which assume faces have been well aligned before pose estimation, would face much difficulty in this situation, since face alignment itself does not work well in this low resolution scenario. In this paper, we propose to estimate head pose using view-based multi-view face detectors directly. Naive Bayesian classifier is then applied to fuse the information of head pose from multiple camera views. To model the temporal changing of head pose, Hidden Markov Model is used to obtain the optimal sequence of head pose with greatest likelihood.",
"Abstract Support vector machines (SVMs) have shown great potential for learning classification functions that can be applied to object recognition. In this work, we extend SVMs to model the appearance of human faces which undergo non-linear change across multiple views. The approach uses inherent factors in the nature of the input images and the SVM classification algorithm to perform both multi-view face detection and pose estimation.",
"",
"This paper describes an approach for the problem of face pose discrimination using support vector machines (SVM). Face pose discrimination means that one can label the face image as one of several known poses. Face images are drawn from the standard FERET database. The training set consists of 150 images equally distributed among frontal, approximately 33.75 spl deg rotated left and right poses, respectively, and the test set consists of 450 images again equally distributed among the three different types of poses. SVM achieved perfect accuracy-100 -discriminating between the three possible face poses on unseen test data, using either polynomials of degree 3 or radial basis functions (RBF) as kernel approximation functions."
]
} |
1905.08609 | 2958161068 | We address a problem of estimating pose of a person's head from its RGB image. The employment of CNNs for the problem has contributed to significant improvement in accuracy in recent works. However, we show that the following two methods, despite their simplicity, can attain further improvement: (i) proper adjustment of the margin of bounding box of a detected face, and (ii) choice of loss functions. We show that the integration of these two methods achieve the new state-of-the-art on standard benchmark datasets for in-the-wild head pose estimation. | @cite_29 proposed a system for estimating location and orientation of head from depth data. @cite_25 , they used random forests for real-time head pose estimation and to localize a set of facial features in 3D. @cite_16 used a convolutional neural network (CNN) to perform head pose estimation from depth data. However, all these methods require depth information acquired by special sensors. In addition, these devices are not always usable while operating in a complex environment like outdoor due to technical problems, and computational complexity of methods which use depth images is higher than that of those using RGB images. So, data requirements of RGB-D image based estimation methods are not satisfied in various industrial products. | {
"cite_N": [
"@cite_29",
"@cite_16",
"@cite_25"
],
"mid": [
"1734489760",
"2963670480",
"2047875689"
],
"abstract": [
"Head pose estimation in low resolution is a challenge problem. Traditional pose estimation algorithms, which assume faces have been well aligned before pose estimation, would face much difficulty in this situation, since face alignment itself does not work well in this low resolution scenario. In this paper, we propose to estimate head pose using view-based multi-view face detectors directly. Naive Bayesian classifier is then applied to fuse the information of head pose from multiple camera views. To model the temporal changing of head pose, Hidden Markov Model is used to obtain the optimal sequence of head pose with greatest likelihood.",
"",
"We present a random forest-based framework for real time head pose estimation from depth images and extend it to localize a set of facial features in 3D. Our algorithm takes a voting approach, where each patch extracted from the depth image can directly cast a vote for the head pose or each of the facial features. Our system proves capable of handling large rotations, partial occlusions, and the noisy depth data acquired using commercial sensors. Moreover, the algorithm works on each frame independently and achieves real time performance without resorting to parallel computations on a GPU. We present extensive experiments on publicly available, challenging datasets and present a new annotated head pose database recorded using a Microsoft Kinect."
]
} |
1905.08749 | 2946345950 | We consider the problem of making a quick decision in favor of one of two possible physical models while the numerical measurements are acquired by sensing devices featuring minimal digitization complexity. Therefore, the digital data streams available for statistical processing are binary and exhibit temporal and spatial dependencies. Such a setting is of relevance when designing radio systems that have to perform time-critical monitoring tasks in noisy environments with low energy consumption. We assess the latency penalty in sequential decision-making resulting from using analog-to-binary instead of high-resolution analog-to-digital conversion at the front-ends. To handle the intractable multivariate binary data model, we first consider sequential tests for exponential family distributions. Within this generic probabilistic framework, we identify approximations for the log-likelihood ratio and the Kullback-Leibler divergence. The results allow designing sequential detectors for binary radio systems and analyzing their average run-time along classical arguments of Wald. In particular, the derived tests exploit the structure of the spatio-temporal correlation of the analog sensor signals engraved into the binary radio data stream. As an application, we consider the specification of binary sensing architectures in the context of cognitive radio and satellite-based synchronization where our results characterize the detection latency as a function of the temporal oversampling and the number of antennas. Finally, we evaluate the efficiency of the proposed algorithms and illustrate the accuracy of our analysis via Monte-Carlo simulations. | Quantized sequential decision-making has been studied predominantly for sensor networks where measurement nodes forward compressed statistics of their observations to minimize the communication overhead @cite_16 @cite_38 @cite_29 @cite_34 @cite_9 @cite_42 @cite_41 . In such a setup, it is usually assumed that the sensing nodes have access to the unquantized digital observations and compress them, for example, by quantizing the likelihood ratio. Here we discuss sequential detection when employing binary sensing front-ends. Therefore, the nodes do not have access to the high-resolution observations, and the central decision is exclusively based on hard-limited measurements. This case is less well studied, and the focus of existing works is on the design of optimal quantizers @cite_44 @cite_18 @cite_33 @cite_15 . | {
"cite_N": [
"@cite_38",
"@cite_18",
"@cite_33",
"@cite_41",
"@cite_29",
"@cite_9",
"@cite_42",
"@cite_34",
"@cite_44",
"@cite_15",
"@cite_16"
],
"mid": [
"",
"2130896996",
"2137865494",
"",
"",
"",
"",
"",
"2106147267",
"2054171087",
"2044313711"
],
"abstract": [
"",
"Optimum detection schemes based on quantized data are of great interest in radar and sonar applications. The design and properties of multisensor schemes are considered here for detection of weak random signals in additive, possibly non-Gaussian, noise. Signal-to-noise ratios are assumed unknown and the signals at the different sensors may be statistically dependent. Analytical expressions describing the best way to fuse the quantized observations for cases with any given observation sample size are provided. The best schemes for originally quantizing the observations are also studied for the case of asymptotically large observation sample sizes. These schemes are shown to minimize the mean-squared error between the best weak-signal test statistic based on unquantized observations and the best weak-signal test statistic based on quantized observations (under signal absent). Numerical results indicate it is sometimes best for each quantizer to use different size alphabets when a quantizer is located at each sensor. >",
"We consider the problem of sequential decentralized detection, a problem that entails the choice of a stopping rule (specifying the sample size), a global decision function (a choice between two competing hypotheses), and a set of quantization rules (the local decisions on the basis of which the global decision is made). The main result of this paper is to resolve an open problem posed by (1993) concerning whether optimal local decision functions for the Bayesian formulation of sequential decentralized detection can be found within the class of stationary rules. We provide a negative answer to this question by exploiting an asymptotic approximation to the optimal cost of stationary quantization rules, and the asymmetry of the Kullback-Leibler divergences. In addition, we show that asymptotically optimal quantizers, when restricted to the space of blockwise stationary quantizers, are likelihood-based threshold rules",
"",
"",
"",
"",
"",
"The quantization of the observed data for sequential signal detection is studied. The criteria used are the minimizations of the average sample number under the hypothesis, the average sample number under the alternative, and the maximum average sample number. Numerical results show that the performance is not very sensitive to different criteria. Using a sequential probability ratio test (SPRT), the performances of optimum quantizers are compared to systems with unquantized data. The asymptotic relative efficiencies of the quantizerSPRT's with respect to the SPRT for unquantized data are derived for symmetric noise densities. The relation between these asymptotic relative efficiencies and those of fixed-sample-size detectors is noted.",
"Distributed sensor systems composed of spatially distributed micro sensor nodes have been proposed for large scale monitoring applications. In these systems, nodes aggregate their sensor data to provide real time information about the underlying state. To extend the lifetime each node of the system has to limit the complexity of the sequential fusion algorithm. In this paper we derive optimal likelihood quantization rules for maximizing sequential detection performance. The resulting sequential detection algorithm is in the form of a finite state machine ideal for implementation in low complexity low power devices.",
"A class of decentralized sequential detection problems is investigated. Under appropriate independence assumptions, it is shown that at each time instant the optimal local strategies are given by threshold tests on the likelihood of ratios. Furthermore, local decisions depend not only on the present and past observations, but on the past local decision as well. That is, for each local processor a different threshold exists for every different sequence of past decisions. Examples, computational techniques, discussions of the difficulties at hand, and suggestions for further exploration of the problem are presented. >"
]
} |
1905.08608 | 2945122711 | With the thriving of deep learning, 3D Convolutional Neural Networks have become a popular choice in volumetric image analysis due to their impressive 3D contexts mining ability. However, the 3D convolutional kernels will introduce a significant increase in the amount of trainable parameters. Considering the training data is often limited in biomedical tasks, a tradeoff has to be made between model size and its representational power. To address this concern, in this paper, we propose a novel 3D Dense Separated Convolution (3D-DSC) module to replace the original 3D convolutional kernels. The 3D-DSC module is constructed by a series of densely connected 1D filters. The decomposition of 3D kernel into 1D filters reduces the risk of over-fitting by removing the redundancy of 3D kernels in a topologically constrained manner, while providing the infrastructure for deepening the network. By further introducing nonlinear layers and dense connections between 1D filters, the network's representational power can be significantly improved while maintaining a compact architecture. We demonstrate the superiority of 3D-DSC on volumetric image classification and segmentation, which are two challenging tasks often encountered in biomedical image computing. | Rather than designing a parameter efficient network, another group of works resort to exploring the redundancy of network in a post-processing manner. Among these efforts, the Low-Rank Approximation (LRA) methods are most relevant to ours. By viewing the convolutional layers as high-order tensors, these methods compress convolutional layers of pre-trained networks by finding their appropriate LRA. Using low-rank decomposition to accelerate convolution was first suggested by @cite_4 in codebook learning. In the context of CNNs, @cite_5 proposed a canonical polyadic (CP) decomposition and clustering scheme for the convolutional kernels. Pre-trained 3D filters are approximated by a consecutive 1D filters and the error is minimized by using clustering and post training. @cite_22 suggested using different tensor decomposition schemes, an iterative schemes was employed to get an approximate local solution. @cite_18 further extend the use of CP decomposition and propose a different low-rank architecture that enable both approximating an already trained network and training from scratch. Since LRA methods are mainly aim to speed up CNNs, and by approximating the weights of pretrained convolutional layers, the improvement of network’s nonlinear representational power are generally disregarded or even scarified. | {
"cite_N": [
"@cite_5",
"@cite_18",
"@cite_4",
"@cite_22"
],
"mid": [
"2167215970",
"2963048316",
"2121775913",
"1996901117"
],
"abstract": [
"We present techniques for speeding up the test-time evaluation of large convolutional networks, designed for object recognition tasks. These models deliver impressive accuracy, but each image evaluation requires millions of floating point operations, making their deployment on smartphones and Internet-scale clusters problematic. The computation is dominated by the convolution operations in the lower layers of the model. We exploit the redundancy present within the convolutional filters to derive approximations that significantly reduce the required computation. Using large state-of-the-art models, we demonstrate speedups of convolutional layers on both CPU and GPU by a factor of 2 x, while keeping the accuracy within 1 of the original model.",
"Abstract: We propose a simple two-step approach for speeding up convolution layers within large convolutional neural networks based on tensor decomposition and discriminative fine-tuning. Given a layer, we use non-linear least squares to compute a low-rank CP-decomposition of the 4D convolution kernel tensor into a sum of a small number of rank-one tensors. At the second step, this decomposition is used to replace the original convolutional layer with a sequence of four convolutional layers with small kernels. After such replacement, the entire network is fine-tuned on the training data using standard backpropagation process. We evaluate this approach on two CNNs and show that it is competitive with previous approaches, leading to higher obtained CPU speedups at the cost of lower accuracy drops for the smaller of the two networks. Thus, for the 36-class character classification CNN, our approach obtains a 8.5x CPU speedup of the whole network with only minor accuracy drop (1 from 91 to 90 ). For the standard ImageNet architecture (AlexNet), the approach speeds up the second convolution layer by a factor of 4x at the cost of @math increase of the overall top-5 classification error.",
"Learning filters to produce sparse image representations in terms of over complete dictionaries has emerged as a powerful way to create image features for many different purposes. Unfortunately, these filters are usually both numerous and non-separable, making their use computationally expensive. In this paper, we show that such filters can be computed as linear combinations of a smaller number of separable ones, thus greatly reducing the computational complexity at no cost in terms of performance. This makes filter learning approaches practical even for large images or 3D volumes, and we show that we significantly outperform state-of-the-art methods on the linear structure extraction task, in terms of both accuracy and speed. Moreover, our approach is general and can be used on generic filter banks to reduce the complexity of the convolutions.",
"The focus of this paper is speeding up the application of convolutional neural networks. While delivering impressive results across a range of computer vision and machine learning tasks, these networks are computationally demanding, limiting their deployability. Convolutional layers generally consume the bulk of the processing time, and so in this work we present two simple schemes for drastically speeding up these layers. This is achieved by exploiting cross-channel or filter redundancy to construct a low rank basis of filters that are rank-1 in the spatial domain. Our methods are architecture agnostic, and can be easily applied to existing CPU and GPU convolutional frameworks for tuneable speedup performance. We demonstrate this with a real world network designed for scene text character recognition [15], showing a possible 2.5× speedup with no loss in accuracy, and 4.5× speedup with less than 1 drop in accuracy, still achieving state-of-the-art on standard benchmarks."
]
} |
1905.08645 | 2946695070 | In this work we present a new framework for the analysis and design of randomized gossip algorithms for solving the average consensus problem. We show how classical randomized iterative methods for solving linear systems can be interpreted as gossip algorithms when applied to special systems encoding the underlying network and explain in detail their decentralized nature. Our general framework recovers a comprehensive array of well-known gossip algorithms as special cases, including the pairwise randomized gossip algorithm and path averaging gossip, and allows for the development of provably faster variants. The flexibility of the new approach enables the design of a number of new specific gossip methods. For instance, we propose and analyze novel block and the first provably accelerated randomized gossip protocols, and dual randomized gossip algorithms. From a numerical analysis viewpoint, our work is the first that explores in depth the decentralized nature of randomized iterative methods for linear systems and proposes them as methods for solving the average consensus problem. We evaluate the performance of the proposed gossip protocols by performing extensive experimental testing on typical wireless network topologies. | The problem of average consensus has been extensively studied in the automatic control and signal processing literature for the past two decades @cite_63 , and was first introduced for decentralized processing in the seminal work @cite_53 . A clear connection between the rate of convergence and spectral characteristics of the underlying network topology over which message passing occurs was first established in @cite_72 for pairwise randomized gossip algorithms. | {
"cite_N": [
"@cite_53",
"@cite_72",
"@cite_63"
],
"mid": [
"2154834860",
"2117905067",
"2120286056"
],
"abstract": [
"We present a model for asynchronous distributed computation and then proceed to analyze the convergence of natural asynchronous distributed versions of a large class of deterministic and stochastic gradient-like algorithms. We show that such algorithms retain the desirable convergence properties of their centralized counterparts, provided that the time between consecutive interprocessor communications and the communication delays are not too large.",
"Motivated by applications to sensor, peer-to-peer, and ad hoc networks, we study distributed algorithms, also known as gossip algorithms, for exchanging information and for computing in an arbitrarily connected network of nodes. The topology of such networks changes continuously as new nodes join and old nodes leave the network. Algorithms for such networks need to be robust against changes in topology. Additionally, nodes in sensor networks operate under limited computational, communication, and energy resources. These constraints have motivated the design of \"gossip\" algorithms: schemes which distribute the computational burden and in which a node communicates with a randomly chosen neighbor. We analyze the averaging problem under the gossip constraint for an arbitrary network graph, and find that the averaging time of a gossip algorithm depends on the second largest eigenvalue of a doubly stochastic matrix characterizing the algorithm. Designing the fastest gossip algorithm corresponds to minimizing this eigenvalue, which is a semidefinite program (SDP). In general, SDPs cannot be solved in a distributed fashion; however, exploiting problem structure, we propose a distributed subgradient method that solves the optimization problem over the network. The relation of averaging time to the second largest eigenvalue naturally relates it to the mixing time of a random walk with transition probabilities derived from the gossip algorithm. We use this connection to study the performance and scaling of gossip algorithms on two popular networks: Wireless Sensor Networks, which are modeled as Geometric Random Graphs, and the Internet graph under the so-called Preferential Connectivity (PC) model.",
"Gossip algorithms are attractive for in-network processing in sensor networks because they do not require any specialized routing, there is no bottleneck or single point of failure, and they are robust to unreliable wireless network conditions. Recently, there has been a surge of activity in the computer science, control, signal processing, and information theory communities, developing faster and more robust gossip algorithms and deriving theoretical performance guarantees. This paper presents an overview of recent work in the area. We describe convergence rate results, which are related to the number of transmitted messages and thus the amount of energy consumed in the network for gossiping. We discuss issues related to gossiping over wireless links, including the effects of quantization and noise, and we illustrate the use of gossip algorithms for canonical signal processing tasks including distributed estimation, source localization, and compression."
]
} |
1905.08645 | 2946695070 | In this work we present a new framework for the analysis and design of randomized gossip algorithms for solving the average consensus problem. We show how classical randomized iterative methods for solving linear systems can be interpreted as gossip algorithms when applied to special systems encoding the underlying network and explain in detail their decentralized nature. Our general framework recovers a comprehensive array of well-known gossip algorithms as special cases, including the pairwise randomized gossip algorithm and path averaging gossip, and allows for the development of provably faster variants. The flexibility of the new approach enables the design of a number of new specific gossip methods. For instance, we propose and analyze novel block and the first provably accelerated randomized gossip protocols, and dual randomized gossip algorithms. From a numerical analysis viewpoint, our work is the first that explores in depth the decentralized nature of randomized iterative methods for linear systems and proposes them as methods for solving the average consensus problem. We evaluate the performance of the proposed gossip protocols by performing extensive experimental testing on typical wireless network topologies. | Motivated by network topologies with salient properties of wireless networks (e.g., nodes can communicate directly only with other nearby nodes), several methods were proposed to accelerate the convergence of gossip algorithms. For instance, @cite_59 proposed averaging among a set of nodes forming a path in the network (this protocol can be seen as special case of our block variants in ). Broadcast gossip algorithms have also been analyzed @cite_82 where the nodes communicate with more than one of their neighbors by broadcasting their values. | {
"cite_N": [
"@cite_59",
"@cite_82"
],
"mid": [
"2127850304",
"2161305486"
],
"abstract": [
"Gossip algorithms have recently received significant attention, mainly because they constitute simple and robust message-passing schemes for distributed information processing over networks. However, for many topologies that are realistic for wireless ad-hoc and sensor networks (like grids and random geometric graphs), the standard nearest-neighbor gossip converges as slowly as flooding (O(n2) messages). A recently proposed algorithm called geographic gossip improves gossip efficiency by a √n factor, by exploiting geographic information to enable multihop long-distance communications. This paper proves that a variation of geographic gossip that averages along routed paths, improves efficiency by an additional √n factor, and is order optimal (O(n) messages) for grids and random geometric graphs with high probability. We develop a general technique (travel agency method) based on Markov chain mixing time inequalities which can give bounds on the performance of randomized message-passing algorithms operating over various graph topologies.",
"Motivated by applications to wireless sensor, peer-to-peer, and ad hoc networks, we study distributed broadcasting algorithms for exchanging information and computing in an arbitrarily connected network of nodes. Specifically, we study a broadcasting-based gossiping algorithm to compute the (possibly weighted) average of the initial measurements of the nodes at every node in the network. We show that the broadcast gossip algorithm converges almost surely to a consensus. We prove that the random consensus value is, in expectation, the average of initial node measurements and that it can be made arbitrarily close to this value in mean squared error sense, under a balanced connectivity model and by trading off convergence speed with accuracy of the computation. We provide theoretical and numerical results on the mean square error performance, on the convergence rate and study the effect of the ldquomixing parameterrdquo on the convergence rate of the broadcast gossip algorithm. The results indicate that the mean squared error strictly decreases through iterations until the consensus is achieved. Finally, we assess and compare the communication cost of the broadcast gossip algorithm to achieve a given distance to consensus through theoretical and numerical results."
]
} |
1905.08645 | 2946695070 | In this work we present a new framework for the analysis and design of randomized gossip algorithms for solving the average consensus problem. We show how classical randomized iterative methods for solving linear systems can be interpreted as gossip algorithms when applied to special systems encoding the underlying network and explain in detail their decentralized nature. Our general framework recovers a comprehensive array of well-known gossip algorithms as special cases, including the pairwise randomized gossip algorithm and path averaging gossip, and allows for the development of provably faster variants. The flexibility of the new approach enables the design of a number of new specific gossip methods. For instance, we propose and analyze novel block and the first provably accelerated randomized gossip protocols, and dual randomized gossip algorithms. From a numerical analysis viewpoint, our work is the first that explores in depth the decentralized nature of randomized iterative methods for linear systems and proposes them as methods for solving the average consensus problem. We evaluate the performance of the proposed gossip protocols by performing extensive experimental testing on typical wireless network topologies. | While the gossip algorithms studied in @cite_72 @cite_59 @cite_82 are all first-order (the update of @math only depends on @math ), a faster randomized pairwise gossip protocol was proposed in @cite_61 which suggested to incorporate additional memory to accelerate convergence. The first analysis of this protocol was later proposed in @cite_47 under strong conditions. It is worth to mention that in the setting of deterministic gossip algorithms theoretical guarantees for accelerated convergence were obtained in @cite_65 @cite_25 . In we propose fast and provably accelerated randomized gossip algorithms with memory and compare them in more detail with the fast randomized algorithm proposed in @cite_61 @cite_0 . | {
"cite_N": [
"@cite_61",
"@cite_65",
"@cite_0",
"@cite_72",
"@cite_59",
"@cite_47",
"@cite_25",
"@cite_82"
],
"mid": [
"1605258304",
"2116066327",
"",
"2117905067",
"2127850304",
"2105333851",
"2122624829",
"2161305486"
],
"abstract": [
"We introduce a technique for accelerating the gos- sip algorithm of Boyd et. al. (INFOCOM 2005) for distributed averaging in a network. By employing memory in the form of a small shift-register in the computation at each node, we can speed up the algorithm’s convergence by a factor of 10. Our accelerated algorithm is inspired by the observation that the original gossip algorithm is analogous to the power method in Numerical Analysis, which can be accelerated by a shift-register based recurrence.",
"Distributed averaging describes a class of network algorithms for the decentralized computation of aggregate statistics. Initially, each node has a scalar data value, and the goal is to compute the average of these values at every node (the so-called average consensus problem). Nodes iteratively exchange information with their neighbors and perform local updates until the value at every node converges to the initial network average. Much previous work has focused on algorithms where each node maintains and updates a single value; every time an update is performed, the previous value is forgotten. Convergence to the average consensus is achieved asymptotically. The convergence rate is fundamentally limited by network connectivity, and it can be prohibitively slow on topologies such as grids and random geometric graphs, even if the update rules are optimized. In this paper, we provide the first theoretical demonstration that adding a local prediction component to the update rule can significantly improve the convergence rate of distributed averaging algorithms. We focus on the case where the local predictor is a linear combination of the node's current and previous values (i.e., two memory taps), and our update rule computes a combination of the predictor and the usual weighted linear combination of values received from neighboring nodes. We derive the optimal mixing parameter for combining the predictor with the neighbors' values, and conduct a theoretical analysis of the improvement in convergence rate that can be achieved using this acceleration methodology. For a chain topology on N nodes, this leads to a factor of N improvement over standard consensus, and for a two-dimensional grid, our approach achieves a factor of ?N improvement.",
"",
"Motivated by applications to sensor, peer-to-peer, and ad hoc networks, we study distributed algorithms, also known as gossip algorithms, for exchanging information and for computing in an arbitrarily connected network of nodes. The topology of such networks changes continuously as new nodes join and old nodes leave the network. Algorithms for such networks need to be robust against changes in topology. Additionally, nodes in sensor networks operate under limited computational, communication, and energy resources. These constraints have motivated the design of \"gossip\" algorithms: schemes which distribute the computational burden and in which a node communicates with a randomly chosen neighbor. We analyze the averaging problem under the gossip constraint for an arbitrary network graph, and find that the averaging time of a gossip algorithm depends on the second largest eigenvalue of a doubly stochastic matrix characterizing the algorithm. Designing the fastest gossip algorithm corresponds to minimizing this eigenvalue, which is a semidefinite program (SDP). In general, SDPs cannot be solved in a distributed fashion; however, exploiting problem structure, we propose a distributed subgradient method that solves the optimization problem over the network. The relation of averaging time to the second largest eigenvalue naturally relates it to the mixing time of a random walk with transition probabilities derived from the gossip algorithm. We use this connection to study the performance and scaling of gossip algorithms on two popular networks: Wireless Sensor Networks, which are modeled as Geometric Random Graphs, and the Internet graph under the so-called Preferential Connectivity (PC) model.",
"Gossip algorithms have recently received significant attention, mainly because they constitute simple and robust message-passing schemes for distributed information processing over networks. However, for many topologies that are realistic for wireless ad-hoc and sensor networks (like grids and random geometric graphs), the standard nearest-neighbor gossip converges as slowly as flooding (O(n2) messages). A recently proposed algorithm called geographic gossip improves gossip efficiency by a √n factor, by exploiting geographic information to enable multihop long-distance communications. This paper proves that a variation of geographic gossip that averages along routed paths, improves efficiency by an additional √n factor, and is order optimal (O(n) messages) for grids and random geometric graphs with high probability. We develop a general technique (travel agency method) based on Markov chain mixing time inequalities which can give bounds on the performance of randomized message-passing algorithms operating over various graph topologies.",
"Gossiping is a distributed process whose purpose is to enable the members of a group of n>1 autonomous agents to asymptotically determine in a decentralized manner, the average of the initial values of their scalar gossip variables. This paper analyzes the accelerated gossip algorithms, first proposed in Cao, Spielman, and Yeh (2006), in which local memory is exploited by installing shift-registers at each agent. For the two-register case, the existence of the desired convergence is established under a symmetry assumption by separately studying the convergence in expectation and in mean square. In particular, the optimal rate of convergence in expectation is derived which is faster than that of the standard gossip algorithm, and a sufficient condition on the adjustable parameter for the convergence in mean square is provided. These theoretical results are validated for some classes of networks by comparison with existing empirical data. More general multi-register cases are also discussed.",
"In the past few years, the problem of distributed consensus has received a lot of attention, particularly in the framework of ad hoc sensor networks. Most methods proposed in the literature address the consensus averaging problem by distributed linear iterative algorithms, with asymptotic convergence of the consensus solution. The convergence rate of such distributed algorithms typically depends on the network topology and the weights given to the edges between neighboring sensors, as described by the network matrix. In this paper, we propose to accelerate the convergence rate for given network matrices by the use of polynomial filtering algorithms. The main idea of the proposed methodology is to apply a polynomial filter on the network matrix that will shape its spectrum in order to increase the convergence rate. Such an algorithm is equivalent to periodic updates in each of the sensors by aggregating a few of its previous estimates. We formulate the computation of the coefficients of the optimal polynomial as a semidefinite program that can be efficiently and globally solved for both static and dynamic network topologies. We finally provide simulation results that demonstrate the effectiveness of the proposed solutions in accelerating the convergence of distributed consensus averaging problems.",
"Motivated by applications to wireless sensor, peer-to-peer, and ad hoc networks, we study distributed broadcasting algorithms for exchanging information and computing in an arbitrarily connected network of nodes. Specifically, we study a broadcasting-based gossiping algorithm to compute the (possibly weighted) average of the initial measurements of the nodes at every node in the network. We show that the broadcast gossip algorithm converges almost surely to a consensus. We prove that the random consensus value is, in expectation, the average of initial node measurements and that it can be made arbitrarily close to this value in mean squared error sense, under a balanced connectivity model and by trading off convergence speed with accuracy of the computation. We provide theoretical and numerical results on the mean square error performance, on the convergence rate and study the effect of the ldquomixing parameterrdquo on the convergence rate of the broadcast gossip algorithm. The results indicate that the mean squared error strictly decreases through iterations until the consensus is achieved. Finally, we assess and compare the communication cost of the broadcast gossip algorithm to achieve a given distance to consensus through theoretical and numerical results."
]
} |
1905.08616 | 2944989142 | We describe a method to infer dense depth from camera motion and sparse depth as estimated using a visual-inertial odometry system. Unlike other scenarios using point clouds from lidar or structured light sensors, we have few hundreds to few thousand points, insufficient to inform the topology of the scene. Our method first constructs a piecewise planar scaffolding of the scene, and then uses it to infer dense depth using the image along with the sparse points. We use a predictive cross-modal criterion, akin to self-supervision,' measuring photometric consistency across time, forward-backward pose consistency, and geometric compatibility with the sparse point cloud. We also launch the first visual-inertial + depth dataset, which we hope will foster additional exploration into combining the complementary strengths of visual and inertial sensors. To compare our method to prior work, we adopt the unsupervised KITTI depth completion benchmark, and show state-of-the-art performance on it. | Rotation Parameterization To construct the photometric consistency loss during training, an auxiliary pose network is needed if no camera poses are available. While the translational part of the relative pose can be modeled as @math , the rotational part belongs to the special orthogonal group @math @cite_2 , which is represented by a @math matrix. @cite_3 uses quaternions, which require an additional norm constraint; this is a soft constraint imposed in the loss function, and thus is not guaranteed. @cite_14 @cite_5 @cite_29 use Euler angles which requires the composition of several matrices that may result in the rotation matrix to no longer be orthogonal. We use the exponential map on @math to map the output of the pose network to a rotation matrix. Though theoretically similar, we empirically found that the exponential map is more beneficial than the Euler angles in sec:experiments_results . | {
"cite_N": [
"@cite_14",
"@cite_29",
"@cite_3",
"@cite_2",
"@cite_5"
],
"mid": [
"2883362496",
"2609883120",
"2200124539",
"1984190176",
"2963583471"
],
"abstract": [
"We propose using global orientation from inertial measurements, and the bias it induces on the shape of objects populating the scene, to inform visual three-dimensional reconstruction. We test the effect of using the resulting prior in-depth prediction from a single image, where the normal vectors to surfaces of objects of certain classes tend to align with gravity or be orthogonal to it. Adding such a prior to baseline methods for monocular depth prediction yields improvements beyond the state-of-the-art and illustrates the power of gravity as a supervisory signal.",
"We present an unsupervised learning framework for the task of monocular depth and camera motion estimation from unstructured video sequences. In common with recent work [10, 14, 16], we use an end-to-end learning approach with view synthesis as the supervisory signal. In contrast to the previous work, our method is completely unsupervised, requiring only monocular video sequences for training. Our method uses single-view depth and multiview pose networks, with a loss based on warping nearby views to the target using the computed depth and pose. The networks are thus coupled by the loss during training, but can be applied independently at test time. Empirical evaluation on the KITTI dataset demonstrates the effectiveness of our approach: 1) monocular depth performs comparably with supervised methods that use either ground-truth pose or depth for training, and 2) pose estimation performs favorably compared to established SLAM systems under comparable input settings.",
"We present a robust and real-time monocular six degree of freedom relocalization system. Our system trains a convolutional neural network to regress the 6-DOF camera pose from a single RGB image in an end-to-end manner with no need of additional engineering or graph optimisation. The algorithm can operate indoors and outdoors in real time, taking 5ms per frame to compute. It obtains approximately 2m and 3 degrees accuracy for large scale outdoor scenes and 0.5m and 5 degrees accuracy indoors. This is achieved using an efficient 23 layer deep convnet, demonstrating that convnets can be used to solve complicated out of image plane regression problems. This was made possible by leveraging transfer learning from large scale classification data. We show that the PoseNet localizes from high level features and is robust to difficult lighting, motion blur and different camera intrinsics where point based SIFT registration fails. Furthermore we show how the pose feature that is produced generalizes to other scenes allowing us to regress pose with only a few dozen training examples.",
"This book introduces the geometry of 3-D vision, that is, the reconstruction of 3-D models of objects from a collection of 2-D images. It details the classic theory of two view geometry and shows that a more proper tool for studying the geometry of multiple views is the so-called rank consideration of the multiple view matrix. It also develops practical reconstruction algorithms and discusses possible extensions of the theory.",
"We propose GeoNet, a jointly unsupervised learning framework for monocular depth, optical flow and egomotion estimation from videos. The three components are coupled by the nature of 3D scene geometry, jointly learned by our framework in an end-to-end manner. Specifically, geometric relationships are extracted over the predictions of individual modules and then combined as an image reconstruction loss, reasoning about static and dynamic scene parts separately. Furthermore, we propose an adaptive geometric consistency loss to increase robustness towards outliers and non-Lambertian regions, which resolves occlusions and texture ambiguities effectively. Experimentation on the KITTI driving dataset reveals that our scheme achieves state-of-the-art results in all of the three tasks, performing better than previously unsupervised methods and comparably with supervised ones."
]
} |
1905.08768 | 2945918727 | Hybrid quantum-classical algorithms such as the quantum approximate optimization algorithm (QAOA) are considered one of the most promising approaches for leveraging near-term quantum computers for practical applications. Such algorithms are often implemented in a variational form, combining classical optimization methods with a quantum machine to find parameters to maximize performance. The quality of the QAOA solution depends heavily on quality of the parameters produced by the classical optimizer. Moreover, multiple local optima in the space of parameters make it harder for the classical optimizer. In this paper we study the use of a multistart optimization approach within a QAOA framework to improve the performance of quantum machines on important graph clustering problems. We also demonstrate that reusing the optimal parameters from similar problems can improve the performance of classical optimization methods, expanding on similar results for MAXCUT. | Despite the recent advances in gradient-based methods @cite_31 @cite_73 @cite_14 @cite_20 @cite_70 @cite_43 , gradient-free black-box methods remain the most common approach for optimizing parameters in hybrid quantum-classical algorithms. A variety of methods have been used, including the Nelder-Mead method @cite_15 (for both QAOA parameter optimization @cite_26 @cite_36 and training quantum Boltzmann machines @cite_61 ), Bayesian methods @cite_23 , Powell's method @cite_8 , and an interior-point minimization method @cite_37 . Researchers resort to derivative-free methods because analytic gradients for quantum circuits may not be available and approximating gradients can be computationally expensive @cite_56 . (In some cases, algorithmic differentiation techniques may provide gradient information @cite_63 .) Since gradient-based methods can be sensitive to noise @cite_50 , they may be less suitable for noisy intermediate-scale quantum hardware. | {
"cite_N": [
"@cite_61",
"@cite_37",
"@cite_14",
"@cite_26",
"@cite_15",
"@cite_70",
"@cite_36",
"@cite_8",
"@cite_56",
"@cite_43",
"@cite_23",
"@cite_63",
"@cite_50",
"@cite_31",
"@cite_73",
"@cite_20"
],
"mid": [
"2775342956",
"2504500777",
"",
"2903998642",
"2171074980",
"",
"",
"2394697995",
"2901444101",
"",
"2774424698",
"2901859177",
"2905869472",
"2573722252",
"",
""
],
"abstract": [
"The question has remained open if near-term gate model quantum computers will offer a quantum advantage for practical applications in the pre-fault tolerance noise regime. A class of algorithms which have shown some promise in this regard are the so-called classical-quantum hybrid variational algorithms. Here we develop a low-depth quantum algorithm to train quantum Boltzmann machine neural networks using such variational methods. We introduce a method which employs the quantum approximate optimization algorithm as a subroutine in order to approximately sample from Gibbs states of Ising Hamiltonians. We use this approximate Gibbs sampling to train neural networks for which we demonstrate training convergence for numerically simulated noisy circuits with depolarizing errors of rates of up to 4 .",
"We use Pontryagin's minimum principle to optimize variational quantum algorithms. We show that for a fixed computation time, the optimal evolution has a bang-bang (square pulse) form, both for closed and open quantum systems with Markovian decoherence. Our findings support the choice of evolution ansatz in the recently proposed Quantum Approximate Optimization Algorithm. Focusing on the Sherrington-Kirkpatrick spin-glass as an example, we find a system-size independent distribution of the duration of pulses, with characteristic time scale set by the inverse of the coupling constants in the Hamiltonian. The optimality of the bang-bang protocols and the characteristic time scale of the pulses provide an efficient parameterization of the protocol and inform the search for effective hybrid (classical and quantum) schemes for tackling combinatorial optimization problems. For the particular systems we study, we find numerically that the optimal nonadiabatic bang-bang protocols outperform conventional quantum annealing in the presence of weak white additive external noise and weak coupling to a thermal bath modeled with the Redfield master equation.",
"",
"Computational quantum technologies are entering a new phase in which noisy intermediate-scale quantum computers are available, but are still too small to benefit from active error correction. Even with a finite coherence budget to invest in quantum information processing, noisy devices with about 50 qubits are expected to experimentally demonstrate quantum supremacy in the next few years. Defined in terms of artificial tasks, current proposals for quantum supremacy, even if successful, will not help to provide solutions to practical problems. Instead, we believe that future users of quantum computers are interested in actual applications and that noisy quantum devices may still provide value by approximately solving hard combinatorial problems via hybrid classical-quantum algorithms. To lower bound the size of quantum computers with practical utility, we perform realistic simulations of the Quantum Approximate Optimization Algorithm and conclude that quantum speedup will not be attainable, at least for a representative combinatorial problem, until several hundreds of qubits are available.",
"A method is described for the minimization of a function of n variables, which depends on the comparison of function values at the (n 41) vertices of a general simplex, followed by the replacement of the vertex with the highest value by another point. The simplex adapts itself to the local landscape, and contracts on to the final minimum. The method is shown to be effective and computationally compact. A procedure is given for the estimation of the Hessian matrix in the neighbourhood of the minimum, needed in statistical estimation problems.",
"",
"",
"Among the embodiments disclosed herein are variants of the quantum approximate optimization algorithm with different parametrization. In particular embodiments, a different objective is used: rather than looking for a state which approximately solves an optimization problem, embodiments of the disclosed technology find a quantum algorithm that will produce a state with high overlap with the optimal state (given an instance, for example, of MAX-2-SAT). In certain embodiments, a machine learning approach is used in which a “training set” of problems is selected and the parameters optimized to produce large overlap for this training set. The problem was then tested on a larger problem set. When tested on the full set, the parameters that were found produced significantly larger overlap than optimized annealing times. Testing on other random instances (e.g., from 20 to 28 bits) continued to show improvement over annealing, with the improvement being most notable on the hardest problems. Embodiments of the disclosed technology can be used, for example, for near-term quantum computers with limited coherence times.",
"The Quantum Approximate Optimization Algorithm (QAOA) is a promising approach for programming a near-term gate-based hybrid quantum computer to find good approximate solutions of hard combinatorial problems. However, little is currently know about the capabilities of QAOA, or of the difficulty of the requisite parameters optimization. Here, we study the performance of QAOA on the MaxCut combinatorial optimization problem, optimizing the quantum circuits on a classical computer using automatic differentiation and stochastic gradient descent, using QuantumFlow, a quantum circuit simulator implemented with TensorFlow. We find that we can amortize the training cost by optimizing on batches of problems instances; that QAOA can exceed the performance of the classical polynomial time Goemans-Williamson algorithm with modest circuit depth, and that performance with fixed circuit depth is insensitive to problem size. Moreover, MaxCut QAOA can be efficiently implemented on a gate-based quantum computer with limited qubit connectivity, using a qubit swap network. These observations support the prospects that QAOA will be an effective method for solving interesting problems on near-term quantum computers.",
"",
"Machine learning techniques have led to broad adoption of a statistical model of computing. The statistical distributions natively available on quantum processors are a superset of those available classically. Harnessing this attribute has the potential to accelerate or otherwise improve machine learning relative to purely classical performance. A key challenge toward that goal is learning to hybridize classical computing resources and traditional learning techniques with the emerging capabilities of general purpose quantum processors. Here, we demonstrate such hybridization by training a 19-qubit gate model processor to solve a clustering problem, a foundational challenge in unsupervised learning. We use the quantum approximate optimization algorithm in conjunction with a gradient-free Bayesian optimization to train the quantum machine. This quantum classical hybrid algorithm shows robustness to realistic noise, and we find evidence that classical optimization can be used to train around both coherent and incoherent imperfections.",
"PennyLane is a Python 3 software framework for optimization and machine learning of quantum and hybrid quantum-classical computations. The library provides a unified architecture for near-term quantum computing devices, supporting both qubit and continuous-variable paradigms. PennyLane's core feature is the ability to compute gradients of variational quantum circuits in a way that is compatible with classical techniques such as backpropagation. PennyLane thus extends the automatic differentiation algorithms common in optimization and machine learning to include quantum and hybrid computations. A plugin system makes the framework compatible with any gate-based quantum simulator or hardware. We provide plugins for StrawberryFields and ProjectQ (including a IBMQE device interface). PennyLane can be used for the optimization of variational quantum eigensolvers, quantum approximate optimization, quantum machine learning models, and many other applications.",
"Generative modeling is a flavor of machine learning with applications ranging from computer vision to chemical design. It is expected to be one of the techniques most suited to take advantage of the additional resources provided by near-term quantum computers. We implement a data-driven quantum circuit training algorithm on the canonical Bars-and-Stripes data set using a quantum-classical hybrid machine. The training proceeds by running parameterized circuits on a trapped ion quantum computer, and feeding the results to a classical optimizer. We apply two separate strategies, Particle Swarm and Bayesian optimization to this task. We show that the convergence of the quantum circuit to the target distribution depends critically on both the quantum hardware and classical optimization strategy. Our study represents the first successful training of a high-dimensional universal quantum circuit, and highlights the promise and challenges associated with hybrid learning schemes.",
"A novel class of hybrid quantum-classical algorithms based on the variational approach have recently emerged from separate proposals addressing, for example, quantum chemistry and combinatorial problems. These algorithms provide an approximate solution to the problem at hand by encoding it in the state of a quantum computer. The operations used to prepare the state are not a priori fixed but, quite the opposite, are subjected to a classical optimization procedure that modifies the quantum gates and improves the quality of the approximate solution. While the quantum hardware determines the size of the problem and what states are achievable (limited, respectively, by the number of qubits and by the kind and number of possible quantum gates), it is the classical optimization procedure that determines the way in which the quantum states are explored and whether the best available solution is actually reached. In addition, the quantities required in the optimization, for example the objective function itself, have to be estimated with finite precision in any experimental implementation. While it is desirable to have very precise estimates, this comes at the cost of repeating the state preparation and measurement multiple times. Here we analyze the competing requirements of high precision and low number of repetitions and study how the overall performance of the variational algorithm is affected by the precision level and the choice of the optimization method. Finally, this study introduces quasi-Newton optimization methods in the general context of hybrid variational algorithms and presents quantitative results for the Quantum Approximate Optimization Algorithm.",
"",
""
]
} |
1905.08768 | 2945918727 | Hybrid quantum-classical algorithms such as the quantum approximate optimization algorithm (QAOA) are considered one of the most promising approaches for leveraging near-term quantum computers for practical applications. Such algorithms are often implemented in a variational form, combining classical optimization methods with a quantum machine to find parameters to maximize performance. The quality of the QAOA solution depends heavily on quality of the parameters produced by the classical optimizer. Moreover, multiple local optima in the space of parameters make it harder for the classical optimizer. In this paper we study the use of a multistart optimization approach within a QAOA framework to improve the performance of quantum machines on important graph clustering problems. We also demonstrate that reusing the optimal parameters from similar problems can improve the performance of classical optimization methods, expanding on similar results for MAXCUT. | @cite_32 show that even at small depth @math the schedule defined by optimal QAOA parameters is reminiscent of adiabatic quantum annealing, where @math is gradually turned on while @math is gradually turned off (see Sec. ). Similar results were found by Crooks @cite_56 . Additionally, @cite_32 demonstrate that the optimal values @math have small variation between similar problem instances, a finding that we confirm in this work for a different graph problem (see Sec. ). use these insights to introduce a novel parameterization of QAOA and a heuristic optimization scheme based on it. | {
"cite_N": [
"@cite_32",
"@cite_56"
],
"mid": [
"2903058119",
"2901444101"
],
"abstract": [
"The Quantum Approximate Optimization Algorithm (QAOA) is a hybrid quantum-classical variational algorithm designed to tackle combinatorial optimization problems. Despite its promise for near-term quantum applications, not much is currently understood about QAOA's performance beyond its lowest-depth variant. An essential but missing ingredient for understanding and deploying QAOA is a constructive approach to carry out the outer-loop classical optimization. We provide an in-depth study of the performance of QAOA on MaxCut problems by developing an efficient parameter-optimization procedure and revealing its ability to exploit non-adiabatic operations. Building on observed patterns in optimal parameters, we propose heuristic optimization strategies that find quasi-optimal parameters in @math time without the @math resources required by brute-force approaches. We then benchmark QAOA and compare it with quantum annealing, especially on difficult instances where adiabatic quantum annealing fails due to small spectral gaps. The comparison reveals that QAOA can learn via optimization to utilize non-adiabatic mechanisms to circumvent the challenges associated with vanishing spectral gaps. Finally, we provide a realistic resource analysis on the experimental implementation of QAOA. When quantum fluctuations in measurements are accounted for, we illustrate that optimization will be important only for problem sizes beyond numerical simulations, but accessible on near-term devices. We propose a feasible implementation of large MaxCut problems with a few hundred vertices in a system of 2D neutral atoms, reaching the regime to challenge the best classical algorithms.",
"The Quantum Approximate Optimization Algorithm (QAOA) is a promising approach for programming a near-term gate-based hybrid quantum computer to find good approximate solutions of hard combinatorial problems. However, little is currently know about the capabilities of QAOA, or of the difficulty of the requisite parameters optimization. Here, we study the performance of QAOA on the MaxCut combinatorial optimization problem, optimizing the quantum circuits on a classical computer using automatic differentiation and stochastic gradient descent, using QuantumFlow, a quantum circuit simulator implemented with TensorFlow. We find that we can amortize the training cost by optimizing on batches of problems instances; that QAOA can exceed the performance of the classical polynomial time Goemans-Williamson algorithm with modest circuit depth, and that performance with fixed circuit depth is insensitive to problem size. Moreover, MaxCut QAOA can be efficiently implemented on a gate-based quantum computer with limited qubit connectivity, using a qubit swap network. These observations support the prospects that QAOA will be an effective method for solving interesting problems on near-term quantum computers."
]
} |
1905.08768 | 2945918727 | Hybrid quantum-classical algorithms such as the quantum approximate optimization algorithm (QAOA) are considered one of the most promising approaches for leveraging near-term quantum computers for practical applications. Such algorithms are often implemented in a variational form, combining classical optimization methods with a quantum machine to find parameters to maximize performance. The quality of the QAOA solution depends heavily on quality of the parameters produced by the classical optimizer. Moreover, multiple local optima in the space of parameters make it harder for the classical optimizer. In this paper we study the use of a multistart optimization approach within a QAOA framework to improve the performance of quantum machines on important graph clustering problems. We also demonstrate that reusing the optimal parameters from similar problems can improve the performance of classical optimization methods, expanding on similar results for MAXCUT. | @cite_38 show that for MAXCUT on 3-regular graphs, the objective function value is concentrated; that is, typical instances have nearly the same value of objective function. They make a case that the same holds for any combinatorial search problem where the number of clauses with a given variable is bounded (e.g., MAXCUT on a bounded-degree graph). They propose reusing optimal parameters between problems that come from the same distribution and refining them using a local optimization heuristic. In this work, we successfully apply this strategy to modularity clustering, a problem where the number of clauses in which a variable can appear grows with @math (see Sec. ). | {
"cite_N": [
"@cite_38"
],
"mid": [
"2904281284"
],
"abstract": [
"The Quantum Approximate Optimization Algorithm, QAOA, uses a shallow depth quantum circuit to produce a parameter dependent state. For a given combinatorial optimization problem instance, the quantum expectation of the associated cost function is the parameter dependent objective function of the QAOA. We demonstrate that if the parameters are fixed and the instance comes from a reasonable distribution then the objective function value is concentrated in the sense that typical instances have (nearly) the same value of the objective function. This applies not just for optimal parameters as the whole landscape is instance independent. We can prove this is true for low depth quantum circuits for instances of MaxCut on large 3-regular graphs. Our results generalize beyond this example. We support the arguments with numerical examples that show remarkable concentration. For higher depth circuits the numerics also show concentration and we argue for this using the Law of Large Numbers. We also observe by simulation that if we find parameters which result in good performance at say 10 bits these same parameters result in good performance at say 24 bits. These findings suggest ways to run the QAOA that reduce or eliminate the use of the outer loop optimization and may allow us to find good solutions with fewer calls to the quantum computer."
]
} |
1905.08768 | 2945918727 | Hybrid quantum-classical algorithms such as the quantum approximate optimization algorithm (QAOA) are considered one of the most promising approaches for leveraging near-term quantum computers for practical applications. Such algorithms are often implemented in a variational form, combining classical optimization methods with a quantum machine to find parameters to maximize performance. The quality of the QAOA solution depends heavily on quality of the parameters produced by the classical optimizer. Moreover, multiple local optima in the space of parameters make it harder for the classical optimizer. In this paper we study the use of a multistart optimization approach within a QAOA framework to improve the performance of quantum machines on important graph clustering problems. We also demonstrate that reusing the optimal parameters from similar problems can improve the performance of classical optimization methods, expanding on similar results for MAXCUT. | Periodicity of the objective function with respect to QAOA parameters, visible on heatmaps in Fig. , has been demonstrated for MAXCUT @cite_33 @cite_20 . The periodicity was also observed for quasi-maximum-likelihood decoding of classical channel codes @cite_3 . This can potentially allow for further restriction of the domain, eliminating some of the local optima and making the optimization problem easier. However, the theoretical results so far are problem specific. Therefore, we restrict our optimization domain to @math , @math , following @cite_34 . Note that this differs from the approach in @cite_56 , where the values of @math and @math were not constrained. A recent result shows that exploiting the periodicity of variational parameters of certain ansatzes for QAOA and other variational algorithms can improve optimization performance @cite_30 . | {
"cite_N": [
"@cite_30",
"@cite_33",
"@cite_3",
"@cite_56",
"@cite_34",
"@cite_20"
],
"mid": [
"2924355162",
"2962767882",
"2919333331",
"2901444101",
"1568345435",
""
],
"abstract": [
"We propose a sequential minimal optimization method for quantum-classical hybrid algorithms, which converges faster, is robust against statistical error, and is hyperparameter-free. Specifically, the optimization problem of the parameterized quantum circuits is divided into solvable subproblems by considering only a subset of the parameters. In fact, if we choose a single parameter, the cost function becomes a simple sine curve with period @math , and hence we can exactly minimize with respect to the chosen parameter. Furthermore, even in general cases, the cost function is given by a simple sum of trigonometric functions with certain periods and hence can be minimized by using a classical computer. By repeatedly performing this procedure, we can optimize the parameterized quantum circuits so that the cost function becomes as small as possible. We perform numerical simulations and compare the proposed method with existing gradient-free and gradient-based optimization algorithms. We find that the proposed method substantially outperforms the existing optimization algorithms and converges to a solution almost independent of the initial choice of the parameters. This accelerates almost all quantum-classical hybrid algorithms readily and would be a key tool for harnessing near-term quantum devices.",
"",
"Motivated by the recent advancement of quantum processors, we investigate quantum approximate optimization algorithm (QAOA) to employ quasi-maximum-likelihood (ML) decoding of classical channel codes. QAOA is a hybrid quantum-classical variational algorithm, which is advantageous for the near-term noisy intermediate-scale quantum (NISQ) devices, where the fidelity of quantum gates is limited by noise and decoherence. We first describe how to construct Ising Hamiltonian model to realize quasi-ML decoding with QAOA. For level-1 QAOA, we derive the systematic way to generate theoretical expressions of cost expectation for arbitrary binary linear codes. Focusing on (7, 4) Hamming code as an example, we analyze the impact of the degree distribution in associated generator matrix on the quantum decoding performance. The excellent performance of higher-level QAOA decoding is verified when Pauli rotation angles are optimized through meta-heuristic variational quantum eigensolver (VQE). Furthermore, we demonstrate the QAOA decoding performance in a real quantum device.",
"The Quantum Approximate Optimization Algorithm (QAOA) is a promising approach for programming a near-term gate-based hybrid quantum computer to find good approximate solutions of hard combinatorial problems. However, little is currently know about the capabilities of QAOA, or of the difficulty of the requisite parameters optimization. Here, we study the performance of QAOA on the MaxCut combinatorial optimization problem, optimizing the quantum circuits on a classical computer using automatic differentiation and stochastic gradient descent, using QuantumFlow, a quantum circuit simulator implemented with TensorFlow. We find that we can amortize the training cost by optimizing on batches of problems instances; that QAOA can exceed the performance of the classical polynomial time Goemans-Williamson algorithm with modest circuit depth, and that performance with fixed circuit depth is insensitive to problem size. Moreover, MaxCut QAOA can be efficiently implemented on a gate-based quantum computer with limited qubit connectivity, using a qubit swap network. These observations support the prospects that QAOA will be an effective method for solving interesting problems on near-term quantum computers.",
"We introduce a quantum algorithm that produces approximate solutions for combinatorial optimization problems. The algorithm depends on a positive integer p and the quality of the approximation improves as p is increased. The quantum circuit that implements the algorithm consists of unitary gates whose locality is at most the locality of the objective function whose optimum is sought. The depth of the circuit grows linearly with p times (at worst) the number of constraints. If p is fixed, that is, independent of the input size, the algorithm makes use of efficient classical preprocessing. If p grows with the input size a different strategy is proposed. We study the algorithm as applied to MaxCut on regular graphs and analyze its performance on 2-regular and 3-regular graphs for fixed p. For p = 1, on 3-regular graphs the quantum algorithm always finds a cut that is at least 0.6924 times the size of the optimal cut.",
""
]
} |
1905.08768 | 2945918727 | Hybrid quantum-classical algorithms such as the quantum approximate optimization algorithm (QAOA) are considered one of the most promising approaches for leveraging near-term quantum computers for practical applications. Such algorithms are often implemented in a variational form, combining classical optimization methods with a quantum machine to find parameters to maximize performance. The quality of the QAOA solution depends heavily on quality of the parameters produced by the classical optimizer. Moreover, multiple local optima in the space of parameters make it harder for the classical optimizer. In this paper we study the use of a multistart optimization approach within a QAOA framework to improve the performance of quantum machines on important graph clustering problems. We also demonstrate that reusing the optimal parameters from similar problems can improve the performance of classical optimization methods, expanding on similar results for MAXCUT. | The multilevel single linkage method (MLSL) @cite_47 @cite_57 , uniformly samples points over the domain @math and starts runs from those points that do not have a better point within a ball of some radius. They show favorable results for a specific approach for updating the radius as the number of sampled points increases, although such results are only asymptotic. MLSL was generalized by APOSMM @cite_41 @cite_67 to consider all points generated by an ensemble of local optimization runs, and not just those sampled from the domain. | {
"cite_N": [
"@cite_57",
"@cite_41",
"@cite_47",
"@cite_67"
],
"mid": [
"2080162860",
"2285921046",
"1969144558",
"2794440401"
],
"abstract": [
"In Part II of our paper, two stochastic methods for global optimization are described that, with probability 1, find all relevant local minima of the objective function with the smallest possible number of local searches. The computational performance of these methods is examined both analytically and empirically.",
"We propose a derivative-free algorithm for finding high-quality local minima for functions that require significant computational resources to evaluate. Our algorithm efficiently utilizes the computational resources allocated to it and also has strong theoretical results, almost surely starting a finite number of local optimization runs and identifying all local minima. We propose metrics for measuring how efficiently an algorithm finds local minima, and we benchmark our algorithm on synthetic problems (with known local minima) and two real-world applications.",
"In this stochastic approach to global optimization, clustering techniques are applied to identify local minima of a real valued objective function that are potentially global. Three different methods of this type are described; their accuracy and efficiency are analyzed in detail.",
"We propose and analyze an asynchronously parallel optimization algorithm for finding multiple, high-quality minima of nonlinear optimization problems. Our multistart algorithm considers all previously evaluated points when determining where to start or continue a local optimization run. Theoretical results show that when there are finitely many minima, the algorithm almost surely starts a finite number of local optimization runs and identifies every minimum. The algorithm is applicable to general optimization settings, but our numerical results focus on the case when derivatives are unavailable. In numerical tests, a Python implementation of the algorithm is shown to yield good approximations of many minima (including a global minimum), and this ability is shown to scale well with additional resources. Our implementation’s time to solution is shown also to scale well even when the time to perform the function evaluation is highly variable. An implementation of the algorithm is available in the libEnsemble library at https: github.com Libensemble libensemble."
]
} |
1905.08711 | 2945947917 | In this work we present a new efficient approach to Human Action Recognition called Video Transformer Network (VTN). It leverages the latest advances in Computer Vision and Natural Language Processing and applies them to video understanding. The proposed method allows us to create lightweight CNN models that achieve high accuracy and real-time speed using just an RGB mono camera and general purpose CPU. Furthermore, we explain how to improve accuracy by distilling from multiple models with different modalities into a single model. We conduct a comparison with state-of-the-art methods and show that our approach performs on par with most of them on famous Action Recognition datasets. We benchmark the inference time of the models using the modern inference framework and argue that our approach compares favorably with other methods in terms of speed accuracy trade-off, running at 56 FPS on CPU. The models and the training code are available. | Currently, there are multiple methods that solve the AR problem with certain quality. One of the examples is the two-stream framework that fuses information from spatial and temporal nets @cite_39 @cite_13 . Spatial net uses RGB frame as input and represents an ordinary classification CNN working on a frame level, whereas temporal net receives multiple stacked optical flow (OF) frames. Calculating OF with traditional algorithms, such as TVL1 @cite_3 , requires extra resources, but there are several ways to avoid it. For example, OF can be extracted with additional sub-network @cite_49 or RGB difference @cite_51 can be used as an alternative motion representation. | {
"cite_N": [
"@cite_13",
"@cite_3",
"@cite_39",
"@cite_49",
"@cite_51"
],
"mid": [
"",
"1578985305",
"2156303437",
"2964094092",
"2507009361"
],
"abstract": [
"",
"Variational methods are among the most successful approaches to calculate the optical flow between two image frames. A particularly appealing formulation is based on total variation (TV) regularization and the robust L1 norm in the data fidelity term. This formulation can preserve discontinuities in the flow field and offers an increased robustness against illumination changes, occlusions and noise. In this work we present a novel approach to solve the TV-L1 formulation. Our method results in a very efficient numerical scheme, which is based on a dual formulation of the TV energy and employs an efficient point-wise thresholding step. Additionally, our approach can be accelerated by modern graphics processing units. We demonstrate the real-time performance (30 fps) of our approach for video inputs at a resolution of 320 × 240 pixels.",
"We investigate architectures of discriminatively trained deep Convolutional Networks (ConvNets) for action recognition in video. The challenge is to capture the complementary information on appearance from still frames and motion between frames. We also aim to generalise the best performing hand-crafted features within a data-driven learning framework. Our contribution is three-fold. First, we propose a two-stream ConvNet architecture which incorporates spatial and temporal networks. Second, we demonstrate that a ConvNet trained on multi-frame dense optical flow is able to achieve very good performance in spite of limited training data. Finally, we show that multitask learning, applied to two different action classification datasets, can be used to increase the amount of training data and improve the performance on both. Our architecture is trained and evaluated on the standard video actions benchmarks of UCF-101 and HMDB-51, where it is competitive with the state of the art. It also exceeds by a large margin previous attempts to use deep nets for video classification.",
"Motion representation plays a vital role in human action recognition in videos. In this study, we introduce a novel compact motion representation for video action recognition, named Optical Flow guided Feature (OFF), which enables the network to distill temporal information through a fast and robust approach. The OFF is derived from the definition of optical flow and is orthogonal to the optical flow. The derivation also provides theoretical support for using the difference between two frames. By directly calculating pixel-wise spatio-temporal gradients of the deep feature maps, the OFF could be embedded in any existing CNN based video action recognition framework with only a slight additional cost. It enables the CNN to extract spatiotemporal information, especially the temporal information between frames simultaneously. This simple but powerful idea is validated by experimental results. The network with OFF fed only by RGB inputs achieves a competitive accuracy of 93.3 on UCF-101, which is comparable with the result obtained by two streams (RGB and optical flow), but is 15 times faster in speed. Experimental results also show that OFF is complementary to other motion modalities such as optical flow. When the proposed method is plugged into the state-of-the-art video action recognition framework, it has 96.0 and 74.2 accuracy on UCF-101 and HMDB-51 respectively. The code for this project is available at: https: github.com kevin-ssy Optical-Flow-Guided-Feature",
"Deep convolutional networks have achieved great success for visual recognition in still images. However, for action recognition in videos, the advantage over traditional methods is not so evident. This paper aims to discover the principles to design effective ConvNet architectures for action recognition in videos and learn these models given limited training samples. Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition. which is based on the idea of long-range temporal structure modeling. It combines a sparse temporal sampling strategy and video-level supervision to enable efficient and effective learning using the whole action video. The other contribution is our study on a series of good practices in learning ConvNets on video data with the help of temporal segment network. Our approach obtains the state-the-of-art performance on the datasets of HMDB51 ( ( 69.4 , )) and UCF101 ( ( 94.2 , )). We also visualize the learned ConvNet models, which qualitatively demonstrates the effectiveness of temporal segment network and the proposed good practices (Models and code at https: github.com yjxiong temporal-segment-networks)."
]
} |
1905.08711 | 2945947917 | In this work we present a new efficient approach to Human Action Recognition called Video Transformer Network (VTN). It leverages the latest advances in Computer Vision and Natural Language Processing and applies them to video understanding. The proposed method allows us to create lightweight CNN models that achieve high accuracy and real-time speed using just an RGB mono camera and general purpose CPU. Furthermore, we explain how to improve accuracy by distilling from multiple models with different modalities into a single model. We conduct a comparison with state-of-the-art methods and show that our approach performs on par with most of them on famous Action Recognition datasets. We benchmark the inference time of the models using the modern inference framework and argue that our approach compares favorably with other methods in terms of speed accuracy trade-off, running at 56 FPS on CPU. The models and the training code are available. | Another popular group of methods is related to the use of 3D primitives like 3D Convolution, 3D Batch Normalization, 3D Pooling, and others. They generalize original operations introducing an additional dimension @math , which indicates the sequence of frames. One of the first architectures that leveraged these primitives for the application to AR, is C3D @cite_29 . Another famous 3D CNN, which saturated UCF-101 benchmark @cite_45 , is I3D @cite_52 . It benefits from pre-training on a large-scale ImageNet @cite_35 dataset by inflating trained 2D filters into 3D. Although methods based on 3D convolutions allow improving results in terms of accuracy, the computational expenses may achieve dozens of GFLOPs. Another substantial drawback is that at some level of the network only a small number of weights inside the convolutional kernels have a significant impact on the output signal regarding their contribution to the absolute value of activations making utilization of resources ineffective. This problem was mentioned in @cite_43 @cite_37 where authors proposed decomposition techniques and mixed architectures that combine 3D and 2D operations on different levels of the network. | {
"cite_N": [
"@cite_35",
"@cite_37",
"@cite_29",
"@cite_52",
"@cite_43",
"@cite_45"
],
"mid": [
"2108598243",
"",
"1522734439",
"2619082050",
"2963155035",
"24089286"
],
"abstract": [
"The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called “ImageNet”, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.",
"",
"We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets, 2) A homogeneous architecture with small 3x3x3 convolution kernels in all layers is among the best performing architectures for 3D ConvNets, and 3) Our learned features, namely C3D (Convolutional 3D), with a simple linear classifier outperform state-of-the-art methods on 4 different benchmarks and are comparable with current best methods on the other 2 benchmarks. In addition, the features are compact: achieving 52.8 accuracy on UCF101 dataset with only 10 dimensions and also very efficient to compute due to the fast inference of ConvNets. Finally, they are conceptually very simple and easy to train and use.",
"The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.9 on HMDB-51 and 98.0 on UCF-101.",
"In this paper we discuss several forms of spatiotemporal convolutions for video analysis and study their effects on action recognition. Our motivation stems from the observation that 2D CNNs applied to individual frames of the video have remained solid performers in action recognition. In this work we empirically demonstrate the accuracy advantages of 3D CNNs over 2D CNNs within the framework of residual learning. Furthermore, we show that factorizing the 3D convolutional filters into separate spatial and temporal components yields significantly gains in accuracy. Our empirical study leads to the design of a new spatiotemporal convolutional block \"R(2+1)D\" which produces CNNs that achieve results comparable or superior to the state-of-the-art on Sports-1M, Kinetics, UCF101, and HMDB51.",
"We introduce UCF101 which is currently the largest dataset of human actions. It consists of 101 action classes, over 13k clips and 27 hours of video data. The database consists of realistic user uploaded videos containing camera motion and cluttered background. Additionally, we provide baseline action recognition results on this new dataset using standard bag of words approach with overall performance of 44.5 . To the best of our knowledge, UCF101 is currently the most challenging dataset of actions due to its large number of classes, large number of clips and also unconstrained nature of such clips."
]
} |
1905.08711 | 2945947917 | In this work we present a new efficient approach to Human Action Recognition called Video Transformer Network (VTN). It leverages the latest advances in Computer Vision and Natural Language Processing and applies them to video understanding. The proposed method allows us to create lightweight CNN models that achieve high accuracy and real-time speed using just an RGB mono camera and general purpose CPU. Furthermore, we explain how to improve accuracy by distilling from multiple models with different modalities into a single model. We conduct a comparison with state-of-the-art methods and show that our approach performs on par with most of them on famous Action Recognition datasets. We benchmark the inference time of the models using the modern inference framework and argue that our approach compares favorably with other methods in terms of speed accuracy trade-off, running at 56 FPS on CPU. The models and the training code are available. | Recurrent neural networks, LSTMs @cite_11 , and GRUs @cite_16 have been regarded as the default starting point for many sequence modeling problems, such as machine translation or language modeling @cite_53 . Many significant results have been achieved in several challenging tasks by means of employing recurrent networks and attention mechanism @cite_6 @cite_4 . Not surprisingly, several approaches to video classification that model sequences with recurrent connections or gated units have been proposed @cite_47 @cite_59 @cite_32 . These models, while showing comparable results on many benchmarks @cite_52 , seem to be more suitable for online prediction and thus real-time applications, because feature vector computed for the frame can be reused for predicting classification label for multiple time-windows containing this frame. | {
"cite_N": [
"@cite_4",
"@cite_53",
"@cite_32",
"@cite_6",
"@cite_52",
"@cite_59",
"@cite_47",
"@cite_16",
"@cite_11"
],
"mid": [
"2133564696",
"2531409750",
"2951183276",
"2581624817",
"2619082050",
"2172806452",
"1923404803",
"1924770834",
"2064675550"
],
"abstract": [
"Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.",
"We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution). In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers. This observation leads us to propose a novel deep convolutional neural network architecture inspired by Inception, where Inception modules have been replaced with depthwise separable convolutions. We show that this architecture, dubbed Xception, slightly outperforms Inception V3 on the ImageNet dataset (which Inception V3 was designed for), and significantly outperforms Inception V3 on a larger image classification dataset comprising 350 million images and 17,000 classes. Since the Xception architecture has the same number of parameters as Inception V3, the performance gains are not due to increased capacity but rather to a more efficient use of model parameters.",
"Models based on deep convolutional networks have dominated recent image interpretation tasks; we investigate whether models which are also recurrent, or \"temporally deep\", are effective for tasks involving sequences, visual and otherwise. We develop a novel recurrent convolutional architecture suitable for large-scale visual learning which is end-to-end trainable, and demonstrate the value of these models on benchmark video recognition tasks, image description and retrieval problems, and video narration challenges. In contrast to current models which assume a fixed spatio-temporal receptive field or simple temporal averaging for sequential processing, recurrent convolutional models are \"doubly deep\"' in that they can be compositional in spatial and temporal \"layers\". Such models may have advantages when target concepts are complex and or training data are limited. Learning long-term dependencies is possible when nonlinearities are incorporated into the network state updates. Long-term RNN models are appealing in that they directly can map variable-length inputs (e.g., video frames) to variable length outputs (e.g., natural language text) and can model complex temporal dynamics; yet they can be optimized with backpropagation. Our recurrent long-term models are directly connected to modern visual convnet models and can be jointly trained to simultaneously learn temporal dynamics and convolutional perceptual representations. Our results show such models have distinct advantages over state-of-the-art models for recognition or generation which are separately defined and or optimized.",
"The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost.",
"The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.9 on HMDB-51 and 98.0 on UCF-101.",
"We propose a soft attention based model for the task of action recognition in videos. We use multi-layered Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) units which are deep both spatially and temporally. Our model learns to focus selectively on parts of the video frames and classifies videos after taking a few glimpses. The model essentially learns which parts in the frames are relevant for the task at hand and attaches higher importance to them. We evaluate the model on UCF-11 (YouTube Action), HMDB-51 and Hollywood2 datasets and analyze how the model focuses its attention depending on the scene and the action being performed.",
"Convolutional neural networks (CNNs) have been extensively applied for image recognition problems giving state-of-the-art results on recognition, detection, segmentation and retrieval. In this work we propose and evaluate several deep neural network architectures to combine image information across a video over longer time periods than previously attempted. We propose two methods capable of handling full length videos. The first method explores various convolutional temporal feature pooling architectures, examining the various design choices which need to be made when adapting a CNN for this task. The second proposed method explicitly models the video as an ordered sequence of frames. For this purpose we employ a recurrent neural network that uses Long Short-Term Memory (LSTM) cells which are connected to the output of the underlying CNN. Our best networks exhibit significant performance improvements over previously published results on the Sports 1 million dataset (73.1 vs. 60.9 ) and the UCF-101 datasets with (88.6 vs. 88.0 ) and without additional optical flow information (82.6 vs. 73.0 ).",
"In this paper we compare different types of recurrent units in recurrent neural networks (RNNs). Especially, we focus on more sophisticated units that implement a gating mechanism, such as a long short-term memory (LSTM) unit and a recently proposed gated recurrent unit (GRU). We evaluate these recurrent units on the tasks of polyphonic music modeling and speech signal modeling. Our experiments revealed that these advanced recurrent units are indeed better than more traditional recurrent units such as tanh units. Also, we found GRU to be comparable to LSTM.",
"Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms."
]
} |
1905.08711 | 2945947917 | In this work we present a new efficient approach to Human Action Recognition called Video Transformer Network (VTN). It leverages the latest advances in Computer Vision and Natural Language Processing and applies them to video understanding. The proposed method allows us to create lightweight CNN models that achieve high accuracy and real-time speed using just an RGB mono camera and general purpose CPU. Furthermore, we explain how to improve accuracy by distilling from multiple models with different modalities into a single model. We conduct a comparison with state-of-the-art methods and show that our approach performs on par with most of them on famous Action Recognition datasets. We benchmark the inference time of the models using the modern inference framework and argue that our approach compares favorably with other methods in terms of speed accuracy trade-off, running at 56 FPS on CPU. The models and the training code are available. | Several viable alternative approaches to sequence modeling have been proposed recently. These approaches, for example convolutional @cite_26 or fully-attentional (e.g. Transformer @cite_23 ) networks, achieve better results on many tasks while addressing significant shortcomings of RNNs such as sequential computing or gradient vanishing. | {
"cite_N": [
"@cite_26",
"@cite_23"
],
"mid": [
"2792764867",
"2963403868"
],
"abstract": [
"For most deep learning practitioners, sequence modeling is synonymous with recurrent networks. Yet recent results indicate that convolutional architectures can outperform recurrent networks on tasks such as audio synthesis and machine translation. Given a new sequence modeling task or dataset, which architecture should one use? We conduct a systematic evaluation of generic convolutional and recurrent architectures for sequence modeling. The models are evaluated across a broad range of standard tasks that are commonly used to benchmark recurrent networks. Our results indicate that a simple convolutional architecture outperforms canonical recurrent networks such as LSTMs across a diverse range of tasks and datasets, while demonstrating longer effective memory. We conclude that the common association between sequence modeling and recurrent networks should be reconsidered, and convolutional networks should be regarded as a natural starting point for sequence modeling tasks. To assist related work, we have made code available at this http URL .",
"The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles, by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.0 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature."
]
} |
1905.08766 | 2951647648 | Image-to-image translation, which translates input images to a different domain with a learned one-to-one mapping, has achieved impressive success in recent years. The success of translation mainly relies on the network architecture to reserve the structural information while modify the appearance slightly at the pixel level through adversarial training. Although these networks are able to learn the mapping, the translated images are predictable without exclusion. It is more desirable to diversify them using image-to-image translation by introducing uncertainties, i.e., the generated images hold potential for variations in colors and textures in addition to the general similarity to the input images, and this happens in both the target and source domains. To this end, we propose a novel generative adversarial network (GAN) based model, InjectionGAN, to learn a many-to-many mapping. In this model, the input image is combined with latent variables, which comprise of domain-specific attribute and unspecific random variations. The domain-specific attribute indicates the target domain of the translation, while the unspecific random variations introduce uncertainty into the model. A unified framework is proposed to regroup these two parts and obtain diverse generations in each domain. Extensive experiments demonstrate that the diverse generations have high quality for the challenging image-to-image translation tasks where no pairing information of the training dataset exits. Both quantitative and qualitative results prove the superior performance of InjectionGAN over the state-of-the-art approaches. | Latent Space Exploring with GAN and VAE. Representation learning refers to the task of learning a representation of the data that can be easily exploited @cite_18 @cite_36 . Deep probabilistic models parameterize the learned representation by a neural network. @cite_19 proposed an interpretable model to learn two types of representations: content and style of the images. An encoder-decoder structure was employed in @cite_38 to learn both the generative and discriminative representations with strong supervision. In @cite_36 , the authors show that standard deep architectures can adversarially approximate to the latent space and explicitly represent factors of variation for image generation. @cite_1 proposed a similar framework to introduce variations to the generated images, but this approach can just learn one direction mapping with paired training dataset. | {
"cite_N": [
"@cite_38",
"@cite_18",
"@cite_36",
"@cite_1",
"@cite_19"
],
"mid": [
"2737047298",
"2777439965",
"2964219393",
"2963330667",
"2298992465"
],
"abstract": [
"The large pose discrepancy between two face images is one of the key challenges in face recognition. Conventional approaches for pose-invariant face recognition either perform face frontalization on, or learn a pose-invariant representation from, a non-frontal face image. We argue that it is more desirable to perform both tasks jointly to allow them to leverage each other. To this end, this paper proposes Disentangled Representation learning-Generative Adversarial Network (DR-GAN) with three distinct novelties. First, the encoder-decoder structure of the generator allows DR-GAN to learn a generative and discriminative representation, in addition to image synthesis. Second, this representation is explicitly disentangled from other face variations such as pose, through the pose code provided to the decoder and pose estimation in the discriminator. Third, DR-GAN can take one or multiple images as the input, and generate one unified representation along with an arbitrary number of synthetic images. Quantitative and qualitative evaluation on both controlled and in-the-wild databases demonstrate the superiority of DR-GAN over the state of the art.",
"Statistical methods are of paramount importance in discovering the modes of variation in visual data. The Principal Component Analysis (PCA) is probably the most prominent method for extracting a single mode of variation in the data. However, in practice, several factors contribute to the appearance of visual objects including pose, illumination, and deformation, to mention a few. To extract these modes of variations from visual data, several supervised methods, such as the TensorFaces relying on multilinear (tensor) decomposition have been developed. The main drawbacks of such methods is that they require both labels regarding the modes of variations and the same number of samples under all modes of variations (e.g., the same face under different expressions, poses etc.). Therefore, their applicability is limited to well-organised data, usually captured in well-controlled conditions. In this paper, we propose a novel general multilinear matrix decomposition method that discovers the multilinear structure of possibly incomplete sets of visual data in unsupervised setting (i.e., without the presence of labels). We also propose extensions of the method with sparsity and low-rank constraints in order to handle noisy data, captured in unconstrained conditions. Besides that, a graph-regularised variant of the method is also developed in order to exploit available geometric or label information for some modes of variations. We demonstrate the applicability of the proposed method in several computer vision tasks, including Shape from Shading (SfS) (in the wild and with occlusion removal), expression transfer, and estimation of surface normals from images captured in the wild.",
"Regularized autoencoders learn the latent codes, a structure with the regularization under the distribution, which enables them the capability to infer the latent codes given observations and generate new samples given the codes. However, they are sometimes ambiguous as they tend to produce reconstructions that are not necessarily a faithful reproduction of the inputs. The main reason is to enforce the learned latent code distribution to match a prior distribution while the true distribution remains unknown. To improve the reconstruction quality and learn the latent space a manifold structure, this paper presents a novel approach using the adversarially approximated autoencoder (AAAE) to investigate the latent codes with adversarial approximation. Instead of regularizing the latent codes by penalizing on the distance between the distributions of the model and the target, AAAE learns the autoencoder flexibly and approximates the latent space with a simpler generator. The ratio is estimated using a generative adversarial network to enforce the similarity of the distributions. In addition, the image space is regularized with an additional adversarial regularizer. The proposed approach unifies two deep generative models for both latent space inference and diverse generation. The learning scheme is realized without regularization on the latent codes, which also encourages faithful reconstruction. Extensive validation experiments on four real-world datasets demonstrate the superior performance of AAAE. In comparison to the state-of-the-art approaches, AAAE generates samples with better quality and shares the properties of a regularized autoencoder with a nice latent manifold structure.",
"Many image-to-image translation problems are ambiguous, as a single input image may correspond to multiple possible outputs. In this work, we aim to model a distribution of possible outputs in a conditional generative modeling setting. The ambiguity of the mapping is distilled in a low-dimensional latent vector, which can be randomly sampled at test time. A generator learns to map the given input, combined with this latent code, to the output. We explicitly encourage the connection between output and the latent code to be invertible. This helps prevent a many-to-one mapping from the latent code to the output during training, also known as the problem of mode collapse, and produces more diverse results. We explore several variants of this approach by employing different training objectives, network architectures, and methods of injecting the latent code. Our proposed method encourages bijective consistency between the latent encoding and output modes. We present a systematic comparison of our method and other variants on both perceptual realism and diversity.",
"Current generative frameworks use end-to-end learning and generate images by sampling from uniform noise distribution. However, these approaches ignore the most basic principle of image formation: images are product of: (a) Structure: the underlying 3D model; (b) Style: the texture mapped onto structure. In this paper, we factorize the image generation process and propose Style and Structure Generative Adversarial Network ( ( S ^2 )-GAN). Our ( S ^2 )-GAN has two components: the Structure-GAN generates a surface normal map; the Style-GAN takes the surface normal map as input and generates the 2D image. Apart from a real vs. generated loss function, we use an additional loss with computed surface normals from generated images. The two GANs are first trained independently, and then merged together via joint learning. We show our ( S ^2 )-GAN model is interpretable, generates more realistic images and can be used to learn unsupervised RGBD representations."
]
} |
1905.08493 | 2945965571 | Virtual Trusted Platform Modules (vTPMs) have been widely used in commercial cloud platforms (e.g. Google Cloud, VMware Cloud, and Microsoft Azure) to provide virtual root-of-trust for virtual machines. Unfortunately, current state-of-the-art vTPM implementations are suffering from confidential data leakage and high performance overhead. In this paper, we present SvTPM, a secure and efficient software-based vTPM implementation based on hardware-rooted Trusted Execution Environment (TEE), providing a whole life cycle protection of vTPMs in the cloud. SvTPM offers strong isolation protection, so that cloud tenants or even cloud administrators cannot get vTPM's private keys or any other sensitive data. In SvTPM, we identify and solve a couple of critical security challenges for vTPM protection with SGX, such as NVRAM replacement attack, rollback attacks, trust establishment, and a fine-grained trusted clock. We implement a prototype of SvTPM on both QEMU and KVM. Performance evaluation results show that SvTPM achieves orders of magnitude of performance gains comparing to the vTPMs protected with physical TPM. The launch time of SvTPM is 2600 @math faster than vTPMs built upon hardware TPM. In the micro-benchmarks evaluation, we find that the command execution latency of SvTPM is smaller than or equal to the existing schemes. | The most related works to SvTPM are eTPM @cite_16 , fTPM @cite_41 and tpmsgx @cite_6 . Their main difference are analyzed and depicted in Table 1. | {
"cite_N": [
"@cite_41",
"@cite_16",
"@cite_6"
],
"mid": [
"2499097732",
"",
"2923884942"
],
"abstract": [
"Commodity CPU architectures, such as ARM and Intel CPUs, have started to offer trusted computing features in their CPUs aimed at displacing dedicated trusted hardware. Unfortunately, these CPU architectures raise serious challenges to building trusted systems because they omit providing secure resources outside the CPU perimeter. This paper shows how to overcome these challenges to build software systems with security guarantees similar to those of dedicated trusted hardware. We present the design and implementation of a firmware-based TPM 2.0 (fTPM) leveraging ARM TrustZone. Our fTPM is the reference implementation of a TPM 2.0 used in millions of mobile devices. We also describe a set of mechanisms needed for the fTPM that can be useful for building more sophisticated trusted applications beyond just a TPM.",
"",
"With close to native performance, Linux containers are becoming the de facto platform for cloud computing. While various solutions have been proposed to secure applications and containers in the cloud environment by leveraging Intel SGX, most cloud operators do not yet offer SGX as a service. This is likely due to a number of security, scalability, and usability concerns coming from both cloud providers and users. Cloud operators worry about the security guarantees of unofficial SDKs, limited support for remote attestation within containers, limited physical memory for the Enclave Page Cache (EPC) making it difficult to support hundreds of enclaves, and potential DoS attacks against EPC by malicious users. Meanwhile, end users need to worry about careful program partitioning to reduce the TCB and adapting legacy applications to use SGX. We note that most of these concerns are the result of an incomplete infrastructure, from the OS to the application layer. We address these concerns with lxcsgx, which allows SGX applications to run inside containers while also: enabling SGX remote attestation for containerized applications, enforcing EPC memory usage control on a per-container basis, providing a general software TPM using SGX to augment legacy applications, and supporting partitioning with a GCC plugin. We then retrofit Nginx OpenSSL and Memcached using the software TPM and SGX partitioning to defend against known and potential attacks. Thanks to the small EPC footprint of each enclave, we are able to run up to 100 containerized Memcached instances without EPC swapping. Our evaluation shows the overhead introduced by lxcsgx is less than 6.9 for simple SGX applications, 9.5 for Nginx OpenSSL, and 20.9 for containerized Memcached."
]
} |
1905.08493 | 2945965571 | Virtual Trusted Platform Modules (vTPMs) have been widely used in commercial cloud platforms (e.g. Google Cloud, VMware Cloud, and Microsoft Azure) to provide virtual root-of-trust for virtual machines. Unfortunately, current state-of-the-art vTPM implementations are suffering from confidential data leakage and high performance overhead. In this paper, we present SvTPM, a secure and efficient software-based vTPM implementation based on hardware-rooted Trusted Execution Environment (TEE), providing a whole life cycle protection of vTPMs in the cloud. SvTPM offers strong isolation protection, so that cloud tenants or even cloud administrators cannot get vTPM's private keys or any other sensitive data. In SvTPM, we identify and solve a couple of critical security challenges for vTPM protection with SGX, such as NVRAM replacement attack, rollback attacks, trust establishment, and a fine-grained trusted clock. We implement a prototype of SvTPM on both QEMU and KVM. Performance evaluation results show that SvTPM achieves orders of magnitude of performance gains comparing to the vTPMs protected with physical TPM. The launch time of SvTPM is 2600 @math faster than vTPMs built upon hardware TPM. In the micro-benchmarks evaluation, we find that the command execution latency of SvTPM is smaller than or equal to the existing schemes. | fTPM @cite_41 presents a design of a firmware-based TPM, which leverages ARM TrustZone. They analyze security guarantees and shortcomings of the ARM TrustZone technology and then present three approaches to overcome the limitations of ARM TrustZone. fTPM is mainly for ARM-based mobile devices, not for vTPMs in cloud environments. Although fTPM also stores the TPM states on untrusted storage, its challenges and solutions are different compared with our solution, especially considering NVRAM replacement attacks and trust establishment between vTPM and SGX platform. | {
"cite_N": [
"@cite_41"
],
"mid": [
"2499097732"
],
"abstract": [
"Commodity CPU architectures, such as ARM and Intel CPUs, have started to offer trusted computing features in their CPUs aimed at displacing dedicated trusted hardware. Unfortunately, these CPU architectures raise serious challenges to building trusted systems because they omit providing secure resources outside the CPU perimeter. This paper shows how to overcome these challenges to build software systems with security guarantees similar to those of dedicated trusted hardware. We present the design and implementation of a firmware-based TPM 2.0 (fTPM) leveraging ARM TrustZone. Our fTPM is the reference implementation of a TPM 2.0 used in millions of mobile devices. We also describe a set of mechanisms needed for the fTPM that can be useful for building more sophisticated trusted applications beyond just a TPM."
]
} |
1905.08493 | 2945965571 | Virtual Trusted Platform Modules (vTPMs) have been widely used in commercial cloud platforms (e.g. Google Cloud, VMware Cloud, and Microsoft Azure) to provide virtual root-of-trust for virtual machines. Unfortunately, current state-of-the-art vTPM implementations are suffering from confidential data leakage and high performance overhead. In this paper, we present SvTPM, a secure and efficient software-based vTPM implementation based on hardware-rooted Trusted Execution Environment (TEE), providing a whole life cycle protection of vTPMs in the cloud. SvTPM offers strong isolation protection, so that cloud tenants or even cloud administrators cannot get vTPM's private keys or any other sensitive data. In SvTPM, we identify and solve a couple of critical security challenges for vTPM protection with SGX, such as NVRAM replacement attack, rollback attacks, trust establishment, and a fine-grained trusted clock. We implement a prototype of SvTPM on both QEMU and KVM. Performance evaluation results show that SvTPM achieves orders of magnitude of performance gains comparing to the vTPMs protected with physical TPM. The launch time of SvTPM is 2600 @math faster than vTPMs built upon hardware TPM. In the micro-benchmarks evaluation, we find that the command execution latency of SvTPM is smaller than or equal to the existing schemes. | Dave @cite_6 presents lxcsgx, which allows SGX applications to run inside containers. This solution puts a software-based TPM to an SGX enclave. However, compared with their work, SvTPM presents and resolves a number of crucial security challenges when providing secure vTPMs for cloud tenants using SGX. In addition, SvTPM can support KVM QEMU architecture, TPM 2.0 specification, and many cryptographic algorithms. | {
"cite_N": [
"@cite_6"
],
"mid": [
"2923884942"
],
"abstract": [
"With close to native performance, Linux containers are becoming the de facto platform for cloud computing. While various solutions have been proposed to secure applications and containers in the cloud environment by leveraging Intel SGX, most cloud operators do not yet offer SGX as a service. This is likely due to a number of security, scalability, and usability concerns coming from both cloud providers and users. Cloud operators worry about the security guarantees of unofficial SDKs, limited support for remote attestation within containers, limited physical memory for the Enclave Page Cache (EPC) making it difficult to support hundreds of enclaves, and potential DoS attacks against EPC by malicious users. Meanwhile, end users need to worry about careful program partitioning to reduce the TCB and adapting legacy applications to use SGX. We note that most of these concerns are the result of an incomplete infrastructure, from the OS to the application layer. We address these concerns with lxcsgx, which allows SGX applications to run inside containers while also: enabling SGX remote attestation for containerized applications, enforcing EPC memory usage control on a per-container basis, providing a general software TPM using SGX to augment legacy applications, and supporting partitioning with a GCC plugin. We then retrofit Nginx OpenSSL and Memcached using the software TPM and SGX partitioning to defend against known and potential attacks. Thanks to the small EPC footprint of each enclave, we are able to run up to 100 containerized Memcached instances without EPC swapping. Our evaluation shows the overhead introduced by lxcsgx is less than 6.9 for simple SGX applications, 9.5 for Nginx OpenSSL, and 20.9 for containerized Memcached."
]
} |
1905.08721 | 2945913957 | Many complex natural and cultural phenomena are well modelled by systems of simple interactions between particles. A number of architectures have been developed to articulate this kind of structure, both implicitly and explicitly. We consider an unsupervised explicit model, the NRI model, and make a series of representational adaptations and physically motivated changes. Most notably we factorise the inferred latent interaction graph into a multiplex graph, allowing each layer to encode for a different interaction-type. This fNRI model is smaller in size and significantly outperforms the original in both edge and trajectory prediction, establishing a new state-of-the-art. We also present a simplified variant of our model, which demonstrates the NRI's formulation as a variational auto-encoder is not necessary for good performance, and make an adaptation to the NRI's training routine, significantly improving its ability to model complex physical dynamical systems. | Many systems have been developed to model interactions implicitly with either single-layer fully connected graphs @cite_1 @cite_7 @cite_4 @cite_2 or with attention-based control mechanisms @cite_3 @cite_8 . However, to the end of developing an investigative or theorising machine assistant, modelling the interaction graph is more valuable than a high quality trajectory-reconstruction. The neural relational inference () model, introduced by Kipf2018NeuralSystemsb , is an unsupervised neural network that learns to predict the interactions and dynamics of a system of objects from observational data alone. When provided with the trajectories of a system of interacting objects, the model infers an explicit interaction graph for these objects which it uses to predict the evolution of the system. The model presents a strong foundation, answering key architectural questions and opening the door for further work dealing with explicit representations. In this work we identify two problems within the original formulation -- representational and experimental -- and in addressing these develop a model that significantly outperforms the original. We also present a variant of our model with greatly improved trajectory prediction that demonstrates the model's formulation as a variational auto-encoder () is not necessary for good performance. | {
"cite_N": [
"@cite_4",
"@cite_7",
"@cite_8",
"@cite_1",
"@cite_3",
"@cite_2"
],
"mid": [
"2624614404",
"2584280040",
"2786615588",
"2402402867",
"",
""
],
"abstract": [
"Relational reasoning is a central component of generally intelligent behavior, but has proven difficult for neural networks to learn. In this paper we describe how to use Relation Networks (RNs) as a simple plug-and-play module to solve problems that fundamentally hinge on relational reasoning. We tested RN-augmented networks on three tasks: visual question answering using a challenging dataset called CLEVR, on which we achieve state-of-the-art, super-human performance; text-based question answering using the bAbI suite of tasks; and complex reasoning about dynamic physical systems. Then, using a curated dataset called Sort-of-CLEVR we show that powerful convolutional networks do not have a general capacity to solve relational questions, but can gain this capacity when augmented with RNs. Our work shows how a deep learning architecture equipped with an RN module can implicitly discover and learn to reason about entities and their relations.",
"The introduction of convolutional layers greatly advanced the performance of neural networks on image tasks due to innately capturing a way of encoding and learning translation-invariant operations, matching one of the underlying symmetries of the image domain. In comparison, there are a number of problems in which there are a number of different inputs which are all 'of the same type' --- multiple particles, multiple agents, multiple stock prices, etc. The corresponding symmetry to this is permutation symmetry, in that the algorithm should not depend on the specific ordering of the input data. We discuss a permutation-invariant neural network layer in analogy to convolutional layers, and show the ability of this architecture to learn to predict the motion of a variable number of interacting hard discs in 2D. In the same way that convolutional layers can generalize to different image sizes, the permutation layer we describe generalizes to different numbers of objects.",
"Common-sense physical reasoning is an essential ingredient for any intelligent agent operating in the real-world. For example, it can be used to simulate the environment, or to infer the state of parts of the world that are currently unobserved. In order to match real-world conditions this causal knowledge must be learned without access to supervised data. To address this problem we present a novel method that learns to discover objects and model their physical interactions from raw visual images in a purely fashion. It incorporates prior knowledge about the compositional nature of human perception to factor interactions between object-pairs and learn efficiently. On videos of bouncing balls we show the superior modelling capabilities of our method compared to other unsupervised neural approaches that do not incorporate such prior knowledge. We demonstrate its ability to handle occlusion and show that it can extrapolate learned knowledge to scenes with different numbers of objects.",
"Many tasks in AI require the collaboration of multiple agents. Typically, the communication protocol between agents is manually specified and not altered during training. In this paper we explore a simple neural model, called CommNet, that uses continuous communication for fully cooperative tasks. The model consists of multiple agents and the communication between them is learned alongside their policy. We apply this model to a diverse set of tasks, demonstrating the ability of the agents to learn to communicate amongst themselves, yielding improved performance over non-communicative agents and baselines. In some cases, it is possible to interpret the language devised by the agents, revealing simple but effective strategies for solving the task at hand.",
"",
""
]
} |
1905.08171 | 2946215091 | In this work, we propose a simple yet effective semi-supervised learning approach called Augmented Distribution Alignment. We reveal that an essential sampling bias exists in semi-supervised learning due to the limited amount of labeled samples, which often leads to a considerable empirical distribution mismatch between labeled data and unlabeled data. To this end, we propose to align the empirical distributions of labeled and unlabeled data to alleviate the bias. On one hand, we adopt an adversarial training strategy to minimize the distribution distance between labeled and unlabeled data as inspired by domain adaptation works. On the other hand, to deal with the small sample size issue of labeled data, we also propose a simple interpolation strategy to generate pseudo training samples. Those two strategies can be easily implemented into existing deep neural networks. We demonstrate the effectiveness of our proposed approach on the benchmark SVHN and CIFAR10 datasets, on which we achieve new state-of-the-art error rates of @math and @math , respectively. Our code will be available at this https URL . | Sampling bias was usually discussed in the literature under the supervised learning and domain adaptation scenarios @cite_2 @cite_35 @cite_36 . Many works have been proposed to measure or address the sampling bias in the learning process @cite_47 @cite_22 @cite_4 @cite_9 . Recently, following the generative adversarial networks @cite_17 , the adversarial training strategy was widely used to address the empirical distribution mismatch in domain adaptation @cite_22 @cite_29 @cite_48 . Although people generally assume samples in two domains are sampled from two different distributions, while in SSL the labeled and unlabeled samples are from the identical distribution, the techniques for reducing domain distribution mismatch used in domain adaptation can be readily used to solve the empirical distribution mismatch in SSL. In this work, we employ the adversarial training strategy proposed in @cite_22 . A potential challenge as discussed in this paper is the small sample size of labeled data might lead to a lack of supports problem when aligning distribution, for which we additionally employ a sample augmentation strategy. | {
"cite_N": [
"@cite_35",
"@cite_4",
"@cite_22",
"@cite_36",
"@cite_48",
"@cite_9",
"@cite_29",
"@cite_2",
"@cite_47",
"@cite_17"
],
"mid": [
"2158754421",
"2159291411",
"2963826681",
"2811380766",
"",
"",
"",
"2129851978",
"2104068492",
"2099471712"
],
"abstract": [
"We study the problem of maximum entropy density estimation in the presence of known sample selection bias. We propose three bias correction approaches. The first one takes advantage of unbiased sufficient statistics which can be obtained from biased samples. The second one estimates the biased distribution and then factors the bias out. The third one approximates the second by only using samples from the sampling distribution. We provide guarantees for the first two approaches and evaluate the performance of all three approaches in synthetic experiments and on real data from species habitat modeling, where maxent has been successfully applied and where sample selection bias is a significant problem.",
"Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multikernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks.",
"Top-performing deep architectures are trained on massive amounts of labeled data. In the absence of labeled data for a certain task, domain adaptation often provides an attractive option given that labeled data of similar nature but from a different domain (e.g. synthetic images) are available. Here, we propose a new approach to domain adaptation in deep architectures that can be trained on large amount of labeled data from the source domain and large amount of unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of \"deep\" features that are (i) discriminative for the main learning task on the source domain and (ii) invariant with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a simple new gradient reversal layer. The resulting augmented architecture can be trained using standard back propagation. Overall, the approach can be implemented with little effort using any of the deep-learning packages. The method performs very well in a series of image classification experiments, achieving adaptation effect in the presence of big domain shifts and outperforming previous state-of-the-art on Office datasets.",
"We consider the scenario where training and test data are drawn from different distributions, commonly referred to as sample selection bias. Most algorithms for this setting try to first recover sampling distributions and then make appropriate corrections based on the distribution estimate. We present a nonparametric method which directly produces resampling weights without distribution estimation. Our method works by matching distributions between training and testing sets in feature space. Experimental results demonstrate that our method works well in practice.",
"",
"",
"",
"We consider the situation in semi-supervised learning, where the \"label sampling\" mechanism stochastically depends on the true response (as well as potentially on the features). We suggest a method of moments for estimating this stochastic dependence using the unlabeled data. This is potentially useful for two distinct purposes: a. As an input to a supervised learning procedure which can be used to \"de-bias\" its results using labeled data only and b. As a potentially interesting learning task in itself. We present several examples to illustrate the practical usefulness of our method.",
"In this paper, we introduce a new domain adaptation (DA) algorithm where the source and target domains are represented by subspaces described by eigenvectors. In this context, our method seeks a domain adaptation solution by learning a mapping function which aligns the source subspace with the target one. We show that the solution of the corresponding optimization problem can be obtained in a simple closed form, leading to an extremely fast algorithm. We use a theoretical result to tune the unique hyper parameter corresponding to the size of the subspaces. We run our method on various datasets and show that, despite its intrinsic simplicity, it outperforms state of the art DA methods.",
"We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples."
]
} |
1905.08171 | 2946215091 | In this work, we propose a simple yet effective semi-supervised learning approach called Augmented Distribution Alignment. We reveal that an essential sampling bias exists in semi-supervised learning due to the limited amount of labeled samples, which often leads to a considerable empirical distribution mismatch between labeled data and unlabeled data. To this end, we propose to align the empirical distributions of labeled and unlabeled data to alleviate the bias. On one hand, we adopt an adversarial training strategy to minimize the distribution distance between labeled and unlabeled data as inspired by domain adaptation works. On the other hand, to deal with the small sample size issue of labeled data, we also propose a simple interpolation strategy to generate pseudo training samples. Those two strategies can be easily implemented into existing deep neural networks. We demonstrate the effectiveness of our proposed approach on the benchmark SVHN and CIFAR10 datasets, on which we achieve new state-of-the-art error rates of @math and @math , respectively. Our code will be available at this https URL . | Our work is also related to the recent proposed interpolation based data augmentation methods for training neural networks @cite_30 @cite_31 @cite_1 . In particular, the method @cite_30 proposed to generate new training samples using convex combinations of pairs of training samples and their labels. In order to address the small sample size issue when aligning distributions, we generalize their approach to the semi-supervised learning by using pseudo-labels for unlabeled samples in the interpolation process. Moreover, we also show that by interpolating between labeled and unlabeled data, the empirical distribution of generate data actually gets closer to the unlabeled samples. | {
"cite_N": [
"@cite_30",
"@cite_31",
"@cite_1"
],
"mid": [
"2963399829",
"2783482415",
"2894770203"
],
"abstract": [
"Large deep neural networks are powerful, but exhibit undesirable behaviors such as memorization and sensitivity to adversarial examples. In this work, we propose mixup, a simple learning principle to alleviate these issues. In essence, mixup trains a neural network on convex combinations of pairs of examples and their labels. By doing so, mixup regularizes the neural network to favor simple linear behavior in-between training examples. Our experiments on the ImageNet-2012, CIFAR-10, CIFAR-100, Google commands and UCI datasets show that mixup improves the generalization of state-of-the-art neural network architectures. We also find that mixup reduces the memorization of corrupt labels, increases the robustness to adversarial examples, and stabilizes the training of generative adversarial networks.",
"Data augmentation is a widely used technique in many machine learning tasks, such as image classification, to virtually enlarge the training dataset size and avoid overfitting. Traditional data augmentation techniques for image classification tasks create new samples from the original training data by, for example, flipping, distorting, adding a small amount of noise to, or cropping a patch from an original image. In this paper, we introduce a simple but surprisingly effective data augmentation technique for image classification tasks. With our technique, named SamplePairing, we synthesize a new sample from one image by overlaying another image randomly chosen from the training data (i.e., taking an average of two images for each pixel). By using two images randomly selected from the training set, we can generate N^2 new samples from N training samples. This simple data augmentation technique significantly improved classification accuracy for all the tested datasets; for example, the top-1 error rate was reduced from 33.5 to 29.0 for the ILSVRC 2012 dataset with GoogLeNet and from 8.22 to 6.93 in the CIFAR-10 dataset. We also show that our SamplePairing technique largely improved accuracy when the number of samples in the training set was very small. Therefore, our technique is more valuable for tasks with a limited amount of training data, such as medical imaging tasks.",
"Deep networks often perform well on the data distribution on which they are trained, yet give incorrect (and often very confident) answers when evaluated on points from off of the training distribution. This is exemplified by the adversarial examples phenomenon but can also be seen in terms of model generalization and domain shift. Ideally, a model would assign lower confidence to points unlike those from the training distribution. We propose a regularizer which addresses this issue by training with interpolated hidden states and encouraging the classifier to be less confident at these points. Because the hidden states are learned, this has an important effect of encouraging the hidden states for a class to be concentrated in such a way so that interpolations within the same class or between two different classes do not intersect with the real data points from other classes. This has a major advantage in that it avoids the underfitting which can result from interpolating in the input space. We prove that the exact condition for this problem of underfitting to be avoided by Manifold Mixup is that the dimensionality of the hidden states exceeds the number of classes, which is often the case in practice. Additionally, this concentration can be seen as making the features in earlier layers more discriminative. We show that despite requiring no significant additional computation, Manifold Mixup achieves large improvements over strong baselines in supervised learning, robustness to single-step adversarial attacks, semi-supervised learning, and Negative Log-Likelihood on held out samples."
]
} |
1905.08027 | 2950307844 | Heterogeneous information network (HIN) embedding aims to embed multiple types of nodes into a low-dimensional space. Although most existing HIN embedding methods consider heterogeneous relations in HINs, they usually employ one single model for all relations without distinction, which inevitably restricts the capability of network embedding. In this paper, we take the structural characteristics of heterogeneous relations into consideration and propose a novel Relation structure-aware Heterogeneous Information Network Embedding model (RHINE). By exploring the real-world networks with thorough mathematical analysis, we present two structure-related measures which can consistently distinguish heterogeneous relations into two categories: Affiliation Relations (ARs) and Interaction Relations (IRs). To respect the distinctive characteristics of relations, in our RHINE, we propose different models specifically tailored to handle ARs and IRs, which can better capture the structures and semantics of the networks. At last, we combine and optimize these models in a unified and elegant manner. Extensive experiments on three real-world datasets demonstrate that our model significantly outperforms the state-of-the-art methods in various tasks, including node clustering, link prediction, and node classification. | Recently, network embedding has attracted considerable attention. Inspired by word2vec @cite_8 , random walk based methods @cite_0 @cite_9 have been proposed to learn representations of networks by the skip-gram model. After that, several models are designed to better preserve network properties @cite_29 @cite_15 @cite_16 . Besides, there are some deep neural network based models for network embedding @cite_28 @cite_1 . However, all the aforementioned methods focus only on learning the representations of homogeneous networks. | {
"cite_N": [
"@cite_8",
"@cite_28",
"@cite_9",
"@cite_29",
"@cite_1",
"@cite_0",
"@cite_15",
"@cite_16"
],
"mid": [
"2153579005",
"2393319904",
"2962756421",
"1888005072",
"2415243320",
"2154851992",
"2387462954",
"2607500032"
],
"abstract": [
"The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling. An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of \"Canada\" and \"Air\" cannot be easily combined to obtain \"Air Canada\". Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible.",
"Network embedding is an important method to learn low-dimensional representations of vertexes in networks, aiming to capture and preserve the network structure. Almost all the existing network embedding methods adopt shallow models. However, since the underlying network structure is complex, shallow models cannot capture the highly non-linear network structure, resulting in sub-optimal network representations. Therefore, how to find a method that is able to effectively capture the highly non-linear network structure and preserve the global and local structure is an open yet important problem. To solve this problem, in this paper we propose a Structural Deep Network Embedding method, namely SDNE. More specifically, we first propose a semi-supervised deep model, which has multiple layers of non-linear functions, thereby being able to capture the highly non-linear network structure. Then we propose to exploit the first-order and second-order proximity jointly to preserve the network structure. The second-order proximity is used by the unsupervised component to capture the global network structure. While the first-order proximity is used as the supervised information in the supervised component to preserve the local network structure. By jointly optimizing them in the semi-supervised deep model, our method can preserve both the local and global network structure and is robust to sparse networks. Empirically, we conduct the experiments on five real-world networks, including a language network, a citation network and three social networks. The results show that compared to the baselines, our method can reconstruct the original network significantly better and achieves substantial gains in three applications, i.e. multi-label classification, link prediction and visualization.",
"Prediction tasks over nodes and edges in networks require careful effort in engineering features used by learning algorithms. Recent research in the broader field of representation learning has led to significant progress in automating prediction by learning the features themselves. However, present feature learning approaches are not expressive enough to capture the diversity of connectivity patterns observed in networks. Here we propose node2vec, an algorithmic framework for learning continuous feature representations for nodes in networks. In node2vec, we learn a mapping of nodes to a low-dimensional space of features that maximizes the likelihood of preserving network neighborhoods of nodes. We define a flexible notion of a node's network neighborhood and design a biased random walk procedure, which efficiently explores diverse neighborhoods. Our algorithm generalizes prior work which is based on rigid notions of network neighborhoods, and we argue that the added flexibility in exploring neighborhoods is the key to learning richer representations. We demonstrate the efficacy of node2vec over existing state-of-the-art techniques on multi-label classification and link prediction in several real-world networks from diverse domains. Taken together, our work represents a new way for efficiently learning state-of-the-art task-independent representations in complex networks.",
"This paper studies the problem of embedding very large information networks into low-dimensional vector spaces, which is useful in many tasks such as visualization, node classification, and link prediction. Most existing graph embedding methods do not scale for real world information networks which usually contain millions of nodes. In this paper, we propose a novel network embedding method called the LINE,'' which is suitable for arbitrary types of information networks: undirected, directed, and or weighted. The method optimizes a carefully designed objective function that preserves both the local and global network structures. An edge-sampling algorithm is proposed that addresses the limitation of the classical stochastic gradient descent and improves both the effectiveness and the efficiency of the inference. Empirical experiments prove the effectiveness of the LINE on a variety of real-world information networks, including language networks, social networks, and citation networks. The algorithm is very efficient, which is able to learn the embedding of a network with millions of vertices and billions of edges in a few hours on a typical single machine. The source code of the LINE is available online https: github.com tangjianpku LINE .",
"In this paper, we propose a novel model for learning graph representations, which generates a low-dimensional vector representation for each vertex by capturing the graph structural information. Different from other previous research efforts, we adopt a random surfing model to capture graph structural information directly, instead of using the sampling-based method for generating linear sequences proposed by (2014). The advantages of our approach will be illustrated from both theorical and empirical perspectives. We also give a new perspective for the matrix factorization method proposed by Levy and Goldberg (2014), in which the pointwise mutual information (PMI) matrix is considered as an analytical solution to the objective function of the skip-gram model with negative sampling proposed by (2013). Unlike their approach which involves the use of the SVD for finding the low-dimensitonal projections from the PMI matrix, however, the stacked denoising autoencoder is introduced in our model to extract complex features and model non-linearities. To demonstrate the effectiveness of our model, we conduct experiments on clustering and visualization tasks, employing the learned vertex representations as features. Empirical results on datasets of varying sizes show that our model outperforms other stat-of-the-art models in such tasks.",
"We present DeepWalk, a novel approach for learning latent representations of vertices in a network. These latent representations encode social relations in a continuous vector space, which is easily exploited by statistical models. DeepWalk generalizes recent advancements in language modeling and unsupervised feature learning (or deep learning) from sequences of words to graphs. DeepWalk uses local information obtained from truncated random walks to learn latent representations by treating walks as the equivalent of sentences. We demonstrate DeepWalk's latent representations on several multi-label network classification tasks for social networks such as BlogCatalog, Flickr, and YouTube. Our results show that DeepWalk outperforms challenging baselines which are allowed a global view of the network, especially in the presence of missing information. DeepWalk's representations can provide F1 scores up to 10 higher than competing methods when labeled data is sparse. In some experiments, DeepWalk's representations are able to outperform all baseline methods while using 60 less training data. DeepWalk is also scalable. It is an online learning algorithm which builds useful incremental results, and is trivially parallelizable. These qualities make it suitable for a broad class of real world applications such as network classification, and anomaly detection.",
"Graph embedding algorithms embed a graph into a vector space where the structure and the inherent properties of the graph are preserved. The existing graph embedding methods cannot preserve the asymmetric transitivity well, which is a critical property of directed graphs. Asymmetric transitivity depicts the correlation among directed edges, that is, if there is a directed path from u to v, then there is likely a directed edge from u to v. Asymmetric transitivity can help in capturing structures of graphs and recovering from partially observed graphs. To tackle this challenge, we propose the idea of preserving asymmetric transitivity by approximating high-order proximity which are based on asymmetric transitivity. In particular, we develop a novel graph embedding algorithm, High-Order Proximity preserved Embedding (HOPE for short), which is scalable to preserve high-order proximities of large scale graphs and capable of capturing the asymmetric transitivity. More specifically, we first derive a general formulation that cover multiple popular high-order proximity measurements, then propose a scalable embedding algorithm to approximate the high-order proximity measurements based on their general formulation. Moreover, we provide a theoretical upper bound on the RMSE (Root Mean Squared Error) of the approximation. Our empirical experiments on a synthetic dataset and three real-world datasets demonstrate that HOPE can approximate the high-order proximities significantly better than the state-of-art algorithms and outperform the state-of-art algorithms in tasks of reconstruction, link prediction and vertex recommendation.",
"Structural identity is a concept of symmetry in which network nodes are identified according to the network structure and their relationship to other nodes. Structural identity has been studied in theory and practice over the past decades, but only recently has it been addressed with representational learning techniques. This work presents struc2vec, a novel and flexible framework for learning latent representations for the structural identity of nodes. struc2vec uses a hierarchy to measure node similarity at different scales, and constructs a multilayer graph to encode structural similarities and generate structural context for nodes. Numerical experiments indicate that state-of-the-art techniques for learning node representations fail in capturing stronger notions of structural identity, while struc2vec exhibits much superior performance in this task, as it overcomes limitations of prior approaches. As a consequence, numerical experiments indicate that struc2vec improves performance on classification tasks that depend more on structural identity."
]
} |
1905.08027 | 2950307844 | Heterogeneous information network (HIN) embedding aims to embed multiple types of nodes into a low-dimensional space. Although most existing HIN embedding methods consider heterogeneous relations in HINs, they usually employ one single model for all relations without distinction, which inevitably restricts the capability of network embedding. In this paper, we take the structural characteristics of heterogeneous relations into consideration and propose a novel Relation structure-aware Heterogeneous Information Network Embedding model (RHINE). By exploring the real-world networks with thorough mathematical analysis, we present two structure-related measures which can consistently distinguish heterogeneous relations into two categories: Affiliation Relations (ARs) and Interaction Relations (IRs). To respect the distinctive characteristics of relations, in our RHINE, we propose different models specifically tailored to handle ARs and IRs, which can better capture the structures and semantics of the networks. At last, we combine and optimize these models in a unified and elegant manner. Extensive experiments on three real-world datasets demonstrate that our model significantly outperforms the state-of-the-art methods in various tasks, including node clustering, link prediction, and node classification. | Different from homogeneous networks, HINs consist of multiple types of nodes and edges. Several attempts have been done on HIN embedding and achieved promising performance in various tasks @cite_3 @cite_7 @cite_25 @cite_23 @cite_24 . PTE @cite_3 decomposes an HIN to a set of bipartite networks and then performs network embedding individually. ESim @cite_7 utilizes user-defined meta-paths as guidance to learn node embeddings. Metapath2vec @cite_17 combines meta-path based random walks and skip-gram model for HIN embedding. HIN2Vec @cite_25 learns the embeddings of HINs by conducting multiple prediction training tasks jointly. HERec @cite_24 filters the node sequences with type constraints and thus captures the semantics of HINs. | {
"cite_N": [
"@cite_7",
"@cite_3",
"@cite_24",
"@cite_23",
"@cite_25",
"@cite_17"
],
"mid": [
"2546547051",
"2145658888",
"2963707260",
"2772021946",
"2767774008",
"2743104969"
],
"abstract": [
"Most real-world data can be modeled as heterogeneous information networks (HINs) consisting of vertices of multiple types and their relationships. Search for similar vertices of the same type in large HINs, such as bibliographic networks and business-review networks, is a fundamental problem with broad applications. Although similarity search in HINs has been studied previously, most existing approaches neither explore rich semantic information embedded in the network structures nor take user's preference as a guidance. In this paper, we re-examine similarity search in HINs and propose a novel embedding-based framework. It models vertices as low-dimensional vectors to explore network structure-embedded similarity. To accommodate user preferences at defining similarity semantics, our proposed framework, ESim, accepts user-defined meta-paths as guidance to learn vertex vectors in a user-preferred embedding space. Moreover, an efficient and parallel sampling-based optimization algorithm has been developed to learn embeddings in large-scale HINs. Extensive experiments on real-world large-scale HINs demonstrate a significant improvement on the effectiveness of ESim over several state-of-the-art algorithms as well as its scalability.",
"Unsupervised text embedding methods, such as Skip-gram and Paragraph Vector, have been attracting increasing attention due to their simplicity, scalability, and effectiveness. However, comparing to sophisticated deep learning architectures such as convolutional neural networks, these methods usually yield inferior results when applied to particular machine learning tasks. One possible reason is that these text embedding methods learn the representation of text in a fully unsupervised way, without leveraging the labeled information available for the task. Although the low dimensional representations learned are applicable to many different tasks, they are not particularly tuned for any task. In this paper, we fill this gap by proposing a semi-supervised representation learning method for text data, which we call the predictive text embedding (PTE). Predictive text embedding utilizes both labeled and unlabeled data to learn the embedding of text. The labeled information and different levels of word co-occurrence information are first represented as a large-scale heterogeneous text network, which is then embedded into a low dimensional space through a principled and efficient algorithm. This low dimensional embedding not only preserves the semantic closeness of words and documents, but also has a strong predictive power for the particular task. Compared to recent supervised approaches based on convolutional neural networks, predictive text embedding is comparable or more effective, much more efficient, and has fewer parameters to tune.",
"Due to the flexibility in modelling data heterogeneity, heterogeneous information network (HIN) has been adopted to characterize complex and heterogeneous auxiliary data in recommender systems, called HIN based recommendation . It is challenging to develop effective methods for HIN based recommendation in both extraction and exploitation of the information from HINs. Most of HIN based recommendation methods rely on path based similarity, which cannot fully mine latent structure features of users and items. In this paper, we propose a novel heterogeneous network embedding based approach for HIN based recommendation, called HERec. To embed HINs, we design a meta-path based random walk strategy to generate meaningful node sequences for network embedding. The learned node embeddings are first transformed by a set of fusion functions, and subsequently integrated into an extended matrix factorization (MF) model. The extended MF model together with fusion functions are jointly optimized for the rating prediction task. Extensive experiments on three real-world datasets demonstrate the effectiveness of the HERec model. Moreover, we show the capability of the HERec model for the cold-start problem, and reveal that the transformed embedding information from HINs can improve the recommendation performance.",
"In online social networks people often express attitudes towards others, which forms massive sentiment links among users. Predicting the sign of sentiment links is a fundamental task in many areas such as personal advertising and public opinion analysis. Previous works mainly focus on textual sentiment classification, however, text information can only disclose the \"tip of the iceberg»» about users» true opinions, of which the most are unobserved but implied by other sources of information such as social relation and users» profile. To address this problem, in this paper we investigate how to predict possibly existing sentiment links in the presence of heterogeneous information. First, due to the lack of explicit sentiment links in mainstream social networks, we establish a labeled heterogeneous sentiment dataset which consists of users» sentiment relation, social relation and profile knowledge by entity-level sentiment extraction method. Then we propose a novel and flexible end-to-end Signed Heterogeneous Information Network Embedding (SHINE) framework to extract users» latent representations from heterogeneous networks and predict the sign of unobserved sentiment links. SHINE utilizes multiple deep autoencoders to map each user into a low-dimension feature space while preserving the network structure. We demonstrate the superiority of SHINE over state-of-the-art baselines on link prediction and node recommendation in two real-world datasets. The experimental results also prove the efficacy of SHINE in cold start scenario.",
"In this paper, we propose a novel representation learning framework, namely HIN2Vec, for heterogeneous information networks (HINs). The core of the proposed framework is a neural network model, also called HIN2Vec, designed to capture the rich semantics embedded in HINs by exploiting different types of relationships among nodes. Given a set of relationships specified in forms of meta-paths in an HIN, HIN2Vec carries out multiple prediction training tasks jointly based on a target set of relationships to learn latent vectors of nodes and meta-paths in the HIN. In addition to model design, several issues unique to HIN2Vec, including regularization of meta-path vectors, node type selection in negative sampling, and cycles in random walks, are examined. To validate our ideas, we learn latent vectors of nodes using four large-scale real HIN datasets, including Blogcatalog, Yelp, DBLP and U.S. Patents, and use them as features for multi-label node classification and link prediction applications on those networks. Empirical results show that HIN2Vec soundly outperforms the state-of-the-art representation learning models for network data, including DeepWalk, LINE, node2vec, PTE, HINE and ESim, by 6.6 to 23.8 of @math - @math in multi-label node classification and 5 to 70.8 of @math in link prediction.",
"We study the problem of representation learning in heterogeneous networks. Its unique challenges come from the existence of multiple types of nodes and links, which limit the feasibility of the conventional network embedding techniques. We develop two scalable representation learning models, namely metapath2vec and metapath2vec++. The metapath2vec model formalizes meta-path-based random walks to construct the heterogeneous neighborhood of a node and then leverages a heterogeneous skip-gram model to perform node embeddings. The metapath2vec++ model further enables the simultaneous modeling of structural and semantic correlations in heterogeneous networks. Extensive experiments show that metapath2vec and metapath2vec++ are able to not only outperform state-of-the-art embedding models in various heterogeneous network mining tasks, such as node classification, clustering, and similarity search, but also discern the structural and semantic correlations between diverse network objects."
]
} |
1905.08152 | 2951368390 | Recent advances in deep reinforcement learning have achieved human-level performance on a variety of real-world applications. However, the current algorithms still suffer from poor gradient estimation with excessive variance, resulting in unstable training and poor sample efficiency. In our paper, we proposed an innovative optimization strategy by utilizing stochastic variance reduced gradient (SVRG) techniques. With extensive experiments on Atari domain, our method outperforms the deep q-learning baselines on 18 out of 20 games. | In recent years, numerous techniques have been proposed to improve the convergence and stability of deep reinforcement learning and optimization method plays a critical role. The well-known @cite_25 uses SGD method. To accelerate the convergence rate and solve the challenges aforementioned, some important improvements are explored. AdaGrad @cite_18 adapts learning rate with respect to the frequency of parameters, and is well-suited for dealing with sparse data. RMSprop @cite_21 improves AdaGrad by resolving its radically diminishing learning rates. Adaptive Moment Estimate (Adam) @cite_8 combines the advantage of both AdaGrad and RMSprop while keeping momentum technique, empirically outperforming other adaptive learning-method algorithms. Under the mechanics of variance control, representative methods such as SAG @cite_13 and SDCA @cite_1 are proposed,.In terms of convergence rate with dramatic variance reduction and not requiring large gradient storage, Stochastic Variance Reduction @cite_20 significantly outperforms SAG and SDCA. Recently, second-order statistics optimization algorithms are adopted @cite_17 @cite_0 . However, second-order methods are infeasible in practice for high-dimensional training, such as neural network. | {
"cite_N": [
"@cite_18",
"@cite_8",
"@cite_21",
"@cite_1",
"@cite_0",
"@cite_13",
"@cite_25",
"@cite_20",
"@cite_17"
],
"mid": [
"2146502635",
"1522301498",
"",
"2952594493",
"2722088290",
"2105875671",
"2119717200",
"2107438106",
"2006544565"
],
"abstract": [
"We present a new family of subgradient methods that dynamically incorporate knowledge of the geometry of the data observed in earlier iterations to perform more informative gradient-based learning. Metaphorically, the adaptation allows us to find needles in haystacks in the form of very predictive but rarely seen features. Our paradigm stems from recent advances in stochastic optimization and online learning which employ proximal functions to control the gradient steps of the algorithm. We describe and analyze an apparatus for adaptively modifying the proximal function, which significantly simplifies setting a learning rate and results in regret guarantees that are provably as good as the best proximal function that can be chosen in hindsight. We give several efficient algorithms for empirical risk minimization problems with common and important regularization functions and domain constraints. We experimentally study our theoretical analysis and show that adaptive subgradient methods outperform state-of-the-art, yet non-adaptive, subgradient algorithms.",
"We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.",
"",
"Stochastic Gradient Descent (SGD) has become popular for solving large scale supervised machine learning optimization problems such as SVM, due to their strong theoretical guarantees. While the closely related Dual Coordinate Ascent (DCA) method has been implemented in various software packages, it has so far lacked good convergence analysis. This paper presents a new analysis of Stochastic Dual Coordinate Ascent (SDCA) showing that this class of methods enjoy strong theoretical guarantees that are comparable or better than SGD. This analysis justifies the effectiveness of SDCA for practical applications.",
"We present novel minibatch stochastic optimization methods for empirical risk minimization problems, the methods efficiently leverage variance reduced first-order and sub-sampled higher-order information to accelerate the convergence speed. For quadratic objectives, we prove improved iteration complexity over state-of-the-art under reasonable assumptions. We also provide empirical evidence of the advantages of our method compared to existing approaches in the literature.",
"We propose a new stochastic gradient method for optimizing the sum of a finite set of smooth functions, where the sum is strongly convex. While standard stochastic gradient methods converge at sublinear rates for this problem, the proposed method incorporates a memory of previous gradient values in order to achieve a linear convergence rate. In a machine learning context, numerical experiments indicate that the new algorithm can dramatically outperform standard algorithms, both in terms of optimizing the training error and reducing the test error quickly.",
"This article presents a general class of associative reinforcement learning algorithms for connectionist networks containing stochastic units. These algorithms, called REINFORCE algorithms, are shown to make weight adjustments in a direction that lies along the gradient of expected reinforcement in both immediate-reinforcement tasks and certain limited forms of delayed-reinforcement tasks, and they do this without explicitly computing gradient estimates or even storing information from which such estimates could be computed. Specific examples of such algorithms are presented, some of which bear a close relationship to certain existing algorithms while others are novel but potentially interesting in their own right. Also given are results that show how such algorithms can be naturally integrated with backpropagation. We close with a brief discussion of a number of additional issues surrounding the use of such algorithms, including what is known about their limiting behaviors as well as further considerations that might be used to help develop similar but potentially more powerful reinforcement learning algorithms.",
"Stochastic gradient descent is popular for large scale optimization but has slow convergence asymptotically due to the inherent variance. To remedy this problem, we introduce an explicit variance reduction method for stochastic gradient descent which we call stochastic variance reduced gradient (SVRG). For smooth and strongly convex functions, we prove that this method enjoys the same fast convergence rate as those of stochastic dual coordinate ascent (SDCA) and Stochastic Average Gradient (SAG). However, our analysis is significantly simpler and more intuitive. Moreover, unlike SDCA or SAG, our method does not require the storage of gradients, and thus is more easily applicable to complex problems such as some structured prediction problems and neural network learning.",
"On-line first-order backpropagation is sufficiently fast and effective for many large-scale classification problems but for very high precision mappings, batch processing may be the method of choice. This paper reviews first- and second-order optimization methods for learning in feedforward neural networks. The viewpoint is that of optimization: many methods can be cast in the language of optimization techniques, allowing the transfer to neural nets of detailed results about computational complexity and safety procedures to ensure convergence and to avoid numerical problems. The review is not intended to deliver detailed prescriptions for the most appropriate methods in specific applications, but to illustrate the main characteristics of the different methods and their mutual relations."
]
} |
1905.08152 | 2951368390 | Recent advances in deep reinforcement learning have achieved human-level performance on a variety of real-world applications. However, the current algorithms still suffer from poor gradient estimation with excessive variance, resulting in unstable training and poor sample efficiency. In our paper, we proposed an innovative optimization strategy by utilizing stochastic variance reduced gradient (SVRG) techniques. With extensive experiments on Atari domain, our method outperforms the deep q-learning baselines on 18 out of 20 games. | By far, stochastic gradient descent is a common method for neural networks optimization @cite_8 . Many optimization problems can be summarized as finding the minima or maxima of scalar objective function @math . Gradient descent updates the parameters in the opposite direction of the gradient of @math until reaching a minimum. However, objective functions are often stochastic as they are composed of different subfunctions @cite_8 . In such cases, stochastic gradient descent (SGD) improves gradient descent by computing gradients with a single or a few training examples and takes gradient steps through individual subfunctions @cite_16 . Although SGD exhibited its efficiency in many machine learning successful stories, there are still some key challenges wait to be solved, including choosing a proper learning rate schedule and avoiding to get trapped in numerous suboptimal local minima in non-convex neural networks training @cite_2 . Therefore, efficient stochastic optimization techniques are required. | {
"cite_N": [
"@cite_16",
"@cite_2",
"@cite_8"
],
"mid": [
"114517082",
"1899249567",
"1522301498"
],
"abstract": [
"During the last decade, the data sizes have grown faster than the speed of processors. In this context, the capabilities of statistical machine learning methods is limited by the computing time rather than the sample size. A more precise analysis uncovers qualitatively different tradeoffs for the case of small-scale and large-scale learning problems. The large-scale case involves the computational complexity of the underlying optimization algorithm in non-trivial ways. Unlikely optimization algorithms such as stochastic gradient descent show amazing performance for large-scale problems. In particular, second order stochastic gradient and averaged stochastic gradient are asymptotically efficient after a single pass on the training set.",
"We study the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model under the assumptions of: i) variable independence, ii) redundancy in network parametrization, and iii) uniformity. These assumptions enable us to explain the complexity of the fully decoupled neural network through the prism of the results from random matrix theory. We show that for large-size decoupled networks the lowest critical values of the random loss function form a layered structure and they are located in a well-defined band lower-bounded by the global minimum. The number of local minima outside that band diminishes exponentially with the size of the network. We empirically verify that the mathematical model exhibits similar behavior as the computer simulations, despite the presence of high dependencies in real networks. We conjecture that both simulated annealing and SGD converge to the band of low critical points, and that all critical points found there are local minima of high quality measured by the test error. This emphasizes a major difference between largeand small-size networks where for the latter poor quality local minima have nonzero probability of being recovered. Finally, we prove that recovering the global minimum becomes harder as the network size increases and that it is in practice irrelevant as global minimum often leads to overfitting.",
"We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm."
]
} |
1905.08152 | 2951368390 | Recent advances in deep reinforcement learning have achieved human-level performance on a variety of real-world applications. However, the current algorithms still suffer from poor gradient estimation with excessive variance, resulting in unstable training and poor sample efficiency. In our paper, we proposed an innovative optimization strategy by utilizing stochastic variance reduced gradient (SVRG) techniques. With extensive experiments on Atari domain, our method outperforms the deep q-learning baselines on 18 out of 20 games. | Another variance in DQN is the Target Approximation Error (TAE), which is investigated by @cite_23 where TAE is result of sub-optimality of Q function parameter @math due to inexact minimization and limited representation power of DQN. A efficient method to reduce TAE variance is Average DQN @cite_23 , the key idea is to use the @math previously calculated Q-values to estimate the current action-value estimate. Another recent explored variance is caused by reward signals noise, which is investigated by @cite_24 . In order to reduce reward signal variance, a direct reward estimator @math is proposed to update the discounted value function instead of sampled reward. Our stochastic variance reduction for deep Q learning method differs from all of the aforementioned approaches. The key idea of our method is to reduce the variance caused by approximate gradient estimation, and thus greatly improve the efficiency and performance. | {
"cite_N": [
"@cite_24",
"@cite_23"
],
"mid": [
"2800191438",
"2953364219"
],
"abstract": [
"Reinforcement Learning (RL) agents require the specification of a reward signal for learning behaviours. However, introduction of corrupt or stochastic rewards can yield high variance in learning. Such corruption may be a direct result of goal misspecification, randomness in the reward signal, or correlation of the reward with external factors that are not known to the agent. Corruption or stochasticity of the reward signal can be especially problematic in robotics, where goal specification can be particularly difficult for complex tasks. While many variance reduction techniques have been studied to improve the robustness of the RL process, handling such stochastic or corrupted reward structures remains difficult. As an alternative for handling this scenario in model-free RL methods, we suggest using an estimator for both rewards and value functions. We demonstrate that this improves performance under corrupted stochastic rewards in both the tabular and non-linear function approximation settings for a variety of noise types and environments. The use of reward estimation is a robust and easy-to-implement improvement for handling corrupted reward signals in model-free RL.",
"Instability and variability of Deep Reinforcement Learning (DRL) algorithms tend to adversely affect their performance. Averaged-DQN is a simple extension to the DQN algorithm, based on averaging previously learned Q-values estimates, which leads to a more stable training procedure and improved performance by reducing approximation error variance in the target values. To understand the effect of the algorithm, we examine the source of value function estimation errors and provide an analytical comparison within a simplified model. We further present experiments on the Arcade Learning Environment benchmark that demonstrate significantly improved stability and performance due to the proposed extension."
]
} |
1905.08138 | 2945657238 | During the last decades, we have witnessed a surge of interests of learning a low-dimensional space with discriminative information from one single view. Even though most of them can achieve satisfactory performance in some certain situations, they fail to fully consider the information from multiple views which are highly relevant but sometimes look different from each other. Besides, correlations between features from multiple views always vary greatly, which challenges multi-view subspace learning. Therefore, how to learn an appropriate subspace which can maintain valuable information from multi-view features is of vital importance but challenging. To tackle this problem, this paper proposes a novel multi-view dimension reduction method named Multi-view Locality Low-rank Embedding for Dimension Reduction (MvL2E). MvL2E makes full use of correlations between multi-view features by adopting low-rank representations. Meanwhile, it aims to maintain the correlations and construct a suitable manifold space to capture the low-dimensional embedding for multi-view features. A centroid based scheme is designed to force multiple views to learn from each other. And an iterative alternating strategy is developed to obtain the optimal solution of MvL2E. The proposed method is evaluated on 5 benchmark datasets. Comprehensive experiments show that our proposed MvL2E can achieve comparable performance with previous approaches proposed in recent literatures. | Let @math denote the features set and @math denote the similarity matrix of the features set @math . In addition, let @math denote a diagonal matrix and @math . We write the normalized graph Laplacian matrix for the view as @math . The spectral clustering method @cite_36 solves the following optimization problem for the normalized graph Laplacian matrix @math : | {
"cite_N": [
"@cite_36"
],
"mid": [
"2165874743"
],
"abstract": [
"Despite many empirical successes of spectral clustering methods— algorithms that cluster points using eigenvectors of matrices derived from the data—there are several unresolved issues. First. there are a wide variety of algorithms that use the eigenvectors in slightly different ways. Second, many of these algorithms have no proof that they will actually compute a reasonable clustering. In this paper, we present a simple spectral clustering algorithm that can be implemented using a few lines of Matlab. Using tools from matrix perturbation theory, we analyze the algorithm, and give conditions under which it can be expected to do well. We also show surprisingly good experimental results on a number of challenging clustering problems."
]
} |
1905.08090 | 2964210793 | Cross-domain synthesizing realistic faces to learn deep models has attracted increasing attention for facial expression analysis as it helps to improve the performance of expression recognition accuracy despite having small number of real training images. However, learning from synthetic face images can be problematic due to the distribution discrepancy between low-quality synthetic images and real face images and may not achieve the desired performance when the learned model applies to real world scenarios. To this end, we propose a new attribute guided face image synthesis to perform a translation between multiple image domains using a single model. In addition, we adopt the proposed model to learn from synthetic faces by matching the feature distributions between different domains while preserving each domain’s characteristics. We evaluate the effectiveness of the proposed approach on several face datasets on generating realistic face images. We demonstrate that the expression recognition performance can be enhanced by benefiting from our face synthesis model. Moreover, we also conduct experiments on a near-infrared dataset containing facial expression videos of drivers to assess the performance using in-the-wild data for driver emotion recognition. | @cite_8 proposed a Deep convolutional network model for Identity-Aware Transfer (DIAT) of the facial attributes. The work @cite_18 and @cite_21 proposed to edit only single facial attribute. @cite_9 proposed attribute-guided face generation to translate low-resolution face images to high-resolution face images. @cite_22 introduced the spatial attention mechanism into GAN framework, to only alter the attribute-specific face region and keep the rest unchanged. @cite_27 proposed a Two-Pathway Generative Adversarial Network (TP-GAN) for photorealistic face synthesis by simultaneously considering local face details and global structures. | {
"cite_N": [
"@cite_18",
"@cite_22",
"@cite_8",
"@cite_9",
"@cite_21",
"@cite_27"
],
"mid": [
"2579578355",
"2895743108",
"2536132686",
"2895749211",
"2744091666",
"2964118024"
],
"abstract": [
"Face attributes are interesting due to their detailed description of human faces. Unlike prior researches working on attribute prediction, we address an inverse and more challenging problem called face attribute manipulation which aims at modifying a face image according to a given attribute value. Instead of manipulating the whole image, we propose to learn the corresponding residual image defined as the difference between images before and after the manipulation. In this way, the manipulation can be operated efficiently with modest pixel modification. The framework of our approach is based on the Generative Adversarial Network. It consists of two image transformation networks and a discriminative network. The transformation networks are responsible for the attribute manipulation and its dual operation and the discriminative network is used to distinguish the generated images from real images. We also apply dual learning to allow transformation networks to learn from each other. Experiments show that residual images can be effectively learned and used for attribute manipulations. The generated images remain most of the details in attribute-irrelevant areas.",
"Face attribute editing aims at editing the face image with the given attribute. Most existing works employ Generative Adversarial Network (GAN) to operate face attribute editing. However, these methods inevitably change the attribute-irrelevant regions, as shown in Fig. 1. Therefore, we introduce the spatial attention mechanism into GAN framework (referred to as SaGAN), to only alter the attribute-specific region and keep the rest unchanged. Our approach SaGAN consists of a generator and a discriminator. The generator contains an attribute manipulation network (AMN) to edit the face image, and a spatial attention network (SAN) to localize the attribute-specific region which restricts the alternation of AMN within this region. The discriminator endeavors to distinguish the generated images from the real ones, and classify the face attribute. Experiments demonstrate that our approach can achieve promising visual results, and keep those attribute-irrelevant regions unchanged. Besides, our approach can benefit the face recognition by data augmentation.",
"This paper presents a Deep convolutional network model for Identity-Aware Transfer (DIAT) of facial attributes. Given the source input image and the reference attribute, DIAT aims to generate a facial image that owns the reference attribute as well as keeps the same or similar identity to the input image. In general, our model consists of a mask network and an attribute transform network which work in synergy to generate a photo-realistic facial image with the reference attribute. Considering that the reference attribute may be only related to some parts of the image, the mask network is introduced to avoid the incorrect editing on attribute irrelevant region. Then the estimated mask is adopted to combine the input and transformed image for producing the transfer result. For joint training of transform network and mask network, we incorporate the adversarial attribute loss, identity-aware adaptive perceptual loss, and VGG-FACE based identity loss. Furthermore, a denoising network is presented to serve for perceptual regularization to suppress the artifacts in transfer result, while an attribute ratio regularization is introduced to constrain the size of attribute relevant region. Our DIAT can provide a unified solution for several representative facial attribute transfer tasks, e.g., expression transfer, accessory removal, age progression, and gender transfer, and can be extended for other face enhancement tasks such as face hallucination. The experimental results validate the effectiveness of the proposed method. Even for the identity-related attribute (e.g., gender), our DIAT can obtain visually impressive results by changing the attribute while retaining most identity-aware features.",
"We are interested in attribute-guided face generation: given a low-res face input image, an attribute vector that can be extracted from a high-res image (attribute image), our new method generates a high-res face image for the low-res input that satisfies the given attributes. To address this problem, we condition the CycleGAN and propose conditional CycleGAN, which is designed to (1) handle unpaired training data because the training low high-res and high-res attribute images may not necessarily align with each other, and to (2) allow easy control of the appearance of the generated face via the input attributes. We demonstrate high-quality results on the attribute-guided conditional CycleGAN, which can synthesize realistic face images with appearance easily controlled by user-supplied attributes (e.g., gender, makeup, hair color, eyeglasses). Using the attribute image as identity to produce the corresponding conditional vector and by incorporating a face verification network, the attribute-guided network becomes the identity-guided conditional CycleGAN which produces high-quality and interesting results on identity transfer. We demonstrate three applications on identity-guided conditional CycleGAN: identity-preserving face superresolution, face swapping, and frontal face generation, which consistently show the advantage of our new method.",
"We present a generative attribute controller (GAC), a novel functionality for generating or editing an image while intuitively controlling large variations of an attribute. This controller is based on a novel generative model called the conditional filtered generative adversarial network (CFGAN), which is an extension of the conventional conditional GAN (CGAN) that incorporates a filtering architecture into the generator input. Unlike the conventional CGAN, which represents an attribute directly using an observable variable (e.g., the binary indicator of attribute presence) so its controllability is restricted to attribute labeling (e.g., restricted to an ON or OFF control), the CFGAN has a filtering architecture that associates an attribute with a multi-dimensional latent variable, enabling latent variations of the attribute to be represented. We also define the filtering architecture and training scheme considering controllability, enabling the variations of the attribute to be intuitively controlled using typical controllers (radio buttons and slide bars). We evaluated our CFGAN on MNIST, CUB, and CelebA datasets and show that it enables large variations of an attribute to be not only represented but also intuitively controlled while retaining identity. We also show that the learned latent space has enough expressive power to conduct attribute transfer and attribute-based image retrieval.",
"Recent studies on face attribute transfer have achieved great success. A lot of models are able to transfer face attributes with an input image. However, they suffer from three limitations: (1) incapability of generating image by exemplars; (2) being unable to transfer multiple face attributes simultaneously; (3) low quality of generated images, such as low-resolution or artifacts. To address these limitations, we propose a novel model which receives two images of opposite attributes as inputs. Our model can transfer exactly the same type of attributes from one image to another by exchanging certain part of their encodings. All the attributes are encoded in a disentangled manner in the latent space, which enables us to manipulate several attributes simultaneously. Besides, our model learns the residual images so as to facilitate training on higher resolution images. With the help of multi-scale discriminators for adversarial training, it can even generate high-quality images with finer details and less artifacts. We demonstrate the effectiveness of our model on overcoming the above three limitations by comparing with other methods on the CelebA face database. A pytorch implementation is available at https: github.com Prinsphield ELEGANT."
]
} |
1905.08090 | 2964210793 | Cross-domain synthesizing realistic faces to learn deep models has attracted increasing attention for facial expression analysis as it helps to improve the performance of expression recognition accuracy despite having small number of real training images. However, learning from synthetic face images can be problematic due to the distribution discrepancy between low-quality synthetic images and real face images and may not achieve the desired performance when the learned model applies to real world scenarios. To this end, we propose a new attribute guided face image synthesis to perform a translation between multiple image domains using a single model. In addition, we adopt the proposed model to learn from synthetic faces by matching the feature distributions between different domains while preserving each domain’s characteristics. We evaluate the effectiveness of the proposed approach on several face datasets on generating realistic face images. We demonstrate that the expression recognition performance can be enhanced by benefiting from our face synthesis model. Moreover, we also conduct experiments on a near-infrared dataset containing facial expression videos of drivers to assess the performance using in-the-wild data for driver emotion recognition. | @cite_16 proposed a method by disentangling the attributes (expression and pose) for simultaneous pose-invariant facial expression recognition and face images synthesis. Instead, we seek to learn attribute-invariant information in the latent space by imposing auxiliary classifier to classify the generated images. @cite_13 introduced a conditional difference adversarial autoencoder (CDAAE) to use emotion states as a conditional attribute for face generation. @cite_11 proposed a multi-task GAN-based network that learns to synthesize the frontal face images from profile face images. However, they require paired training data of frontal and profile faces. Instead, we seek to add realism to the synthetic frontal face images without requiring real frontal face images during training. Our method could generate realistic frontal face images using synthetic faces and real faces with arbitrary poses as input. | {
"cite_N": [
"@cite_16",
"@cite_13",
"@cite_11"
],
"mid": [
"2798553619",
"2963520896",
"2805613995"
],
"abstract": [
"Facial expression recognition (FER) is a challenging task due to different expressions under arbitrary poses. Most conventional approaches either perform face frontalization on a non-frontal facial image or learn separate classifiers for each pose. Different from existing methods, in this paper, we propose an end-to-end deep learning model by exploiting different poses and expressions jointly for simultaneous facial image synthesis and pose-invariant facial expression recognition. The proposed model is based on generative adversarial network (GAN) and enjoys several merits. First, the encoder-decoder structure of the generator can learn a generative and discriminative identity representation for face images. Second, the identity representation is explicitly disentangled from both expression and pose variations through the expression and pose codes. Third, our model can automatically generate face images with different expressions under arbitrary poses to enlarge and enrich the training set for FER. Quantitative and qualitative evaluations on both controlled and in-the-wild datasets demonstrate that the proposed algorithm performs favorably against state-of-the-art methods.",
"Photorealistic facial expression synthesis from single face image can be widely applied to face recognition, data augmentation for emotion recognition or entertainment. This problem is challenging, in part due to a paucity of labeled facial expression data, making it difficult for algorithms to disambiguate changes due to identity and changes due to expression. In this paper, we propose the conditional difference adversarial autoencoder (CDAAE) for facial expression synthesis. The CDAAE takes a facial image of a previously unseen person and generates an image of that person's face with a target emotion or facial action unit (AU) label. The CDAAE adds a feedforward path to an autoencoder structure connecting low level features at the encoder to features at the corresponding level at the decoder. It handles the problem of disambiguating changes due to identity and changes due to facial expression by learning to generate the difference between low-level features of images of the same person but with different facial expressions. The CDAAE structure can be used to generate novel expressions by combining and interpolating between facial expressions action units within the training set. Our experimental results demonstrate that the CDAAE can preserve identity information when generating facial expression for unseen subjects more faithfully than previous approaches. This is especially advantageous when training with small databases.",
"Face frontalization is one way to overcome the pose variation problem, which simplifies multi-view recognition into one canonical-view recognition. This paper presents a multi-task learning approach based on the generative adversarial network (GAN) that learns the emotion-preserving representations in the face frontalization framework. Taking advantage of adversarial relationship between the generator and the discriminator in GAN, the generator can frontalize input non-frontal face images into frontal face images while preserving the identity and expression characteristics; in the meantime, it can employ the learnt emotion-preserving representations to predict the expression class label from the input face. The proposed network is optimized by combining both synthesis and classification objective functions to make the learnt representations generative and discriminative simultaneously. Experimental results demonstrate that the proposed face frontalization system is very effective for expression recognition with large head pose variations."
]
} |
1905.08339 | 2945720393 | This paper studies the structure of several real-world traces (including Facebook, High-Performance Computing, Machine Learning, and simulation generated traces) and presents a systematic approach to quantify and compare the structure of packet traces based on the entropy contained in the trace file. Insights into the structure of packet traces can lead to improved network algorithms that are optimized toward specific traffic patterns. We then present a methodology to quantify the temporal and non-temporal components of entropy contained in a packet trace, called the trace complexity, using randomization and compression. We show that trace complexity provides unique insights into the characteristics of various applications and argue that there is a need for traffic generation models that preserve the intrinsic structure of empirically measured application traces. We then propose a traffic generator model that is able to produce a synthetic trace that matches the complexity level of its corresponding real-world trace. | Information-theoretic approaches and compression methodologies have already been proven successful in capturing entropy in other domains such as email @cite_10 , or comment @cite_28 spam filtering, or estimating neural discharges @cite_11 . The study of traffic patterns and the design of models is an evergreen topic of high relevance in the networking literature, and examples where measurement studies spurred much research into traffic modeling dates back to the 1990s @cite_38 @cite_42 @cite_33 . Since then, a large number of methodologies have been developed @cite_20 @cite_12 @cite_32 , based on temporal statistics @cite_17 @cite_3 @cite_31 , spatial statistics @cite_43 @cite_36 , and physical @cite_13 and information-theoretic @cite_7 @cite_39 models. In contrast to prior work, we are primarily interested in the communication pattern itself, rather than in the volume or headers of the exchanged data. While our work builds upon many significant results developed over the last decades @cite_34 , we are not aware of any work which allows to systematically differentiate between temporal and non-temporal components of traffic traces. | {
"cite_N": [
"@cite_13",
"@cite_38",
"@cite_33",
"@cite_7",
"@cite_28",
"@cite_36",
"@cite_42",
"@cite_32",
"@cite_17",
"@cite_3",
"@cite_39",
"@cite_43",
"@cite_31",
"@cite_34",
"@cite_10",
"@cite_12",
"@cite_20",
"@cite_11"
],
"mid": [
"2133985894",
"2152536325",
"2105818147",
"2074373872",
"2123449924",
"",
"2138543759",
"2010488817",
"2158712333",
"2155669300",
"2622770015",
"2143022286",
"2146603609",
"2122962290",
"2104290684",
"2165773639",
"2146320425",
"2116393125"
],
"abstract": [
"One role for workload generation is as a means for understanding how servers and networks respond to variation in load. This enables management and capacity planning based on current and projected usage. This paper applies a number of observations of Web server usage to create a realistic Web workload generation tool which mimics a set of real users accessing a server. The tool, called Surge (Scalable URL Reference Generator) generates references matching empirical measurements of 1) server file size distribution; 2) request size distribution; 3) relative file popularity; 4) embedded file references; 5) temporal locality of reference; and 6) idle periods of individual users. This paper reviews the essential elements required in the generation of a representative Web workload. It also addresses the technical challenges to satisfying this large set of simultaneous constraints on the properties of the reference stream, the solutions we adopted, and their associated accuracy. Finally, we present evidence that Surge exercises servers in a manner significantly different from other Web server benchmarks.",
"As ATM high-speed, cell-relay networks will most likely first make their impact as backbones interconnecting enterprise networks consisting of Ethernet and other LANs, their proper design and control is crucial. Recent studies of high quality, high resolution traffic measurements in Bellcore Ethernets have revealed that this aggregate Ethernet traffic is self-similar (\"fractal\") in nature, quite different in \"burstiness\" features from traffic considered and studied up to now. This paper presents an analytical study of an ATM buffer driven with self-similar traffic. The probability of buffer occupancy is obtained. It is shown that this probability decreases with the buffer size not exponentially, as in traditionally Markovian traffic models, but algebraically.",
"Demonstrates that Ethernet LAN traffic is statistically self-similar, that none of the commonly used traffic models is able to capture this fractal-like behavior, that such behavior has serious implications for the design, control, and analysis of high-speed, cell-based networks, and that aggregating streams of such traffic typically intensifies the self-similarity (\"burstiness\") instead of smoothing it. These conclusions are supported by a rigorous statistical analysis of hundreds of millions of high quality Ethernet traffic measurements collected between 1989 and 1992, coupled with a discussion of the underlying mathematical and statistical properties of self-similarity and their relationship with actual network behavior. The authors also present traffic models based on self-similar stochastic processes that provide simple, accurate, and realistic descriptions of traffic scenarios expected during B-ISDN deployment. >",
"Network engineers and operators are faced with a number of challenges that arise in the context of network monitoring and measurement. These include: i) how much information is included in measurement traces and by how much can we compress those traces?, ii) how much information is captured by different monitoring paradigms and tools ranging from full packet header captures to flow-level captures (such as with NetFlow) to packet and byte counts (such as with SNMP)? and iii) how much joint information is included in traces collected at different points and can we take advantage of this joint information? In this paper we develop a network model and an information theoretic framework within which to address these questions. We use the model and the framework to first determine the benefits of compressing traces captured at a single monitoring point, and we outline approaches to achieve those benefits. We next consider the benefits of joint coding, or equivalently of joint compression of traces captured a different monitoring points. Finally, we examine the difference in information content when measurements are made at either the flow level or the packet byte count level. In all of these cases, the effect of temporal and spatial correlation on the answers to the above questions is examined. Both our model and its predictions are validated against measurements taken from a large operational network.",
"In this work, we design a method for blog comment spam detection using the assumption that spam is any kind of uninformative content. To measure the \"informativeness\" of a set of blog comments, we construct a language and tokenization independent metric which we call content complexity, providing a normalized answer to the informal question \"how much information does this text contain?\" We leverage this metric to create a small set of features well-adjusted to comment spam detection by computing the content complexity over groupings of messages sharing the same author, the same sender IP, the same included links, etc. We evaluate our method against an exact set of tens of millions of comments collected over a four months period and containing a variety of websites, including blogs and news sites. The data was provided to us with an initial spam labeling from an industry competitive source. Nevertheless the initial spam labeling had unknown performance characteristics. To train a logistic regression on this dataset using our features, we derive a simple mislabeling tolerant logistic regression algorithm based on expectation-maximization, which we show generally outperforms the plain version in precision-recall space. By using a parsimonious hand-labeling strategy, we show that our method can operate at an arbitrary high precision level, and that it significantly dominates, both in terms of precision and recall, the original labeling, despite being trained on it alone. The content complexity metric, the use of a noise-tolerant logistic regression and the evaluation methodology are thus the three central contributions with this work.",
"",
"The notion of self-similarity has been shown to apply to wide-area and local-area network traffic. We show evidence that the subset of network traffic that is due to World Wide Web (WWW) transfers can show characteristics that are consistent with self-similarity, and we present a hypothesized explanation for that self-similarity. Using a set of traces of actual user executions of NCSA Mosaic, we examine the dependence structure of WWW traffic. First, we show evidence that WWW traffic exhibits behavior that is consistent with self-similar traffic models. Then we show that the self-similarity in such traffic can be explained based on the underlying distributions of WWW document sizes, the effects of caching and user preference in file transfer, the effect of user \"think time\", and the superimposition of many such transfers in a local-area network. To do this, we rely on empirically measured distributions both from client traces and from data independently collected at WWW servers.",
"Modern disk drives read-ahead data and reorder incoming requests in a workload-dependent fashion. This improves their performance, but makes simple analytical models of them inadequate for performance prediction, capacity planning, workload balancing, and so on. To address this problem we have developed a new analytic model for disk drives that do readahead and request reordering. We did so by developing performance models of the disk drive components (queues, caches, and the disk mechanism) and a workload transformation technique for composing them. Our model includes the effects of workload-specific parameters such as request size and spatial locality. The result is capable of predicting the behavior of a variety of real-world devices to within 17 across a variety of workloads and disk drives.",
"We present a detailed statistical analysis of a 2-hour long empirical sample of VBR video. The sample was obtained by applying a simple intraframe video compression code to an action movie. The main findings of our analysis are (1) the tail behavior of the marginal bandwidth distribution can be accurately described using “heavy-tailed” distributions (e.g., Pareto); (2) the autocorrelation of the VBR video sequence decays hyperbolically (equivalent to long-range dependence ) and can be modeled using self-similar processes. We combine our findings in a new (non-Markovian) source model for VBR video and present an algorithm for generating synthetic traffic. Trace-driven simulations show that statistical multiplexing results in significant bandwidth efficiency even when long-range dependence is present. Simulations of our source model show long-range dependence and heavy-tailed marginals to be important components which are not accounted for in currently used VBR video traffic models.",
"We develop a new multiscale modeling framework for characterizing positive-valued data with long-range-dependent correlations (1 f noise). Using the Haar wavelet transform and a special multiplicative structure on the wavelet and scaling coefficients to ensure positive results, the model provides a rapid O(N) cascade algorithm for synthesizing N-point data sets. We study both the second-order and multifractal properties of the model, the latter after a tutorial overview of multifractal analysis. We derive a scheme for matching the model to real data observations and, to demonstrate its effectiveness, apply the model to network traffic synthesis. The flexibility and accuracy of the model and fitting procedure result in a close fit to the real data statistics (variance-time plots and moment scaling) and queuing behavior. Although for illustrative purposes we focus on applications in network traffic modeling, the multifractal wavelet model could be useful in a number of other areas involving positive data, including image processing, finance, and geophysics.",
"In this paper, we propose an information theoretic framework within which to study the redundancy present in packet header traces. Packet level and flow level models are developed that capture both temporal and spatial correlation present in packet headers, which can be exploited for packet trace compression. Information theoretic bounds are established for lossless packet header compression. Dependencies between the potential compression ratio and network parameters, such as the average route length and average flow size are derived. Valuable insights are obtained to guide the design of efficient packet trace compression algorithms.",
"Statistics for Spatial Data GEOSTATISTICAL DATA Geostatistics Spatial Prediction and Kriging Applications of Geostatistics Special Topics in Statistics for Spatial Data LATTICE DATA Spatial Models on Lattices Inference for Lattice Models SPATIAL PATTERNS Spatial Point Patterns Modeling Objects References Author Index Subject Index.",
"Network, Web, and disk I O traffic are usually bursty and self-similar and therefore cannot be modeled adequately with Poisson arrivals. However, we wish to model these types of traffic and generate realistic traces, because of obvious applications for disk scheduling, network management, and Web server design. Previous models (like fractional Brownian motion and FARIMA, etc.) tried to capture the 'burstiness'. However, the proposed models either require too many parameters to fit and or require prohibitively large (quadratic) time to generate large traces. We propose a simple, parsimonious method, the b-model, which solves both problems: it requires just one parameter, and can easily generate large traces. In addition, it has many more attractive properties: (a) with our proposed estimation algorithm, it requires just a single pass over the actual trace to estimate b. For example, a one-day-long disk trace in milliseconds contains about 86 Mb data points and requires about 3 minutes for model fitting and 5 minutes for generation. (b) The resulting synthetic traces are very realistic: our experiments on real disk and Web traces show that our synthetic traces match the real ones very well in terms of queuing behavior.",
"Compressibility of individual sequences by the class of generalized finite-state information-lossless encoders is investigated. These encoders can operate in a variable-rate mode as well as a fixed-rate one, and they allow for any finite-state scheme of variable-length-to-variable-length coding. For every individual infinite sequence x a quantity (x) is defined, called the compressibility of x , which is shown to be the asymptotically attainable lower bound on the compression ratio that can be achieved for x by any finite-state encoder. This is demonstrated by means of a constructive coding theorem and its converse that, apart from their asymptotic significance, also provide useful performance criteria for finite and practical data-compression tasks. The proposed concept of compressibility is also shown to play a role analogous to that of entropy in classical information theory where one deals with probabilistic ensembles of sequences rather than with individual sequences. While the definition of (x) allows a different machine for each different sequence to be compressed, the constructive coding theorem leads to a universal algorithm that is asymptotically optimal for all sequences.",
"Spam filtering poses a special problem in text categorization, of which the defining characteristic is that filters face an active adversary, which constantly attempts to evade filtering. Since spam evolves continuously and most practical applications are based on online user feedback, the task calls for fast, incremental and robust learning algorithms. In this paper, we investigate a novel approach to spam filtering based on adaptive statistical data compression models. The nature of these models allows them to be employed as probabilistic text classifiers based on character-level or binary sequences. By modeling messages as sequences, tokenization and other error-prone preprocessing steps are omitted altogether, resulting in a method that is very robust. The models are also fast to construct and incrementally updateable. We evaluate the filtering performance of two different compression algorithms; dynamic Markov compression and prediction by partial matching. The results of our empirical evaluation indicate that compression models outperform currently established spam filters, as well as a number of methods proposed in previous studies.",
"Various methods for estimating the self-similarity parameter and or the intensity of long-range dependence in a time series are available. Some are more reliable than others. To discover the ones that work best, we apply the different methods to simulated sequences of fractional Gaussian noise and fractional ARIMA (0, d, 0). We also provide here a theoretical justification for the method of residuals of regression.",
"Measurements of LAN and WAN traffic show that network traffic exhibits variability on different scales. We examine a mechanism that gives rise to self-similar network traffic and discuss performance. The mechanism we study is the transfer of files or messages whose size is drawn from a heavy-tailed distribution. In a realistic client server network the degree to which file sizes are heavy-tailed can directly determine the degree of traffic self-similarity at the link level. This causal relationship is robust relative to changes in network resources, network topology, the influence of cross-traffic, and the distribution of interarrival times. Properties of the transport layer play an important role in preserving and modulating this relationship. The reliable transmission and flow control mechanisms of TCP serve to maintain the long-range dependency structure induced by heavy-tailed file size distributions. In contrast, if a non-flow-controlled and unreliable (UDP-based) transport protocol is used, the resulting traffic shows little self-similarity: although still bursty at short time scales, it has little long-range dependence. Performance implications of self-similarity are discussed as represented by various performance measures. Increased self-similarity as expected, results in degradation of performance. Queueing delay, in particular is discussed. Throughput-related measures such as packet loss and retransmission rate, however increase only gradually with increasing traffic self-similarity as long as reliable, flow-controlled transport protocol is used.",
"Normalized Lempel-Ziv complexity, which measures the generation rate of new patterns along a digital sequence, is closely related to such important source properties as entropy and compression ratio, but, in contrast to these, it is a property of individual sequences. In this article, we propose to exploit this concept to estimate (or, at least, to bound from below) the entropy of neural discharges (spike trains). The main advantages of this method include fast convergence of the estimator (as supported by numerical simulation) and the fact that there is no need to know the probability law of the process generating the signal. Furthermore, we present numerical and experimental comparisons of the new method against the standard method based on word frequencies, providing evidence that this new approach is an alternative entropy estimator for binned spike trains."
]
} |
1905.08377 | 2946205552 | Usage similarity estimation addresses the semantic proximity of word instances in different contexts. We apply contextualized (ELMo and BERT) word and sentence embeddings to this task, and propose supervised models that leverage these representations for prediction. Our models are further assisted by lexical substitute annotations automatically assigned to word instances by context2vec, a neural model that relies on a bidirectional LSTM. We perform an extensive comparison of existing word and sentence representations on benchmark datasets addressing both graded and binary similarity. The best performing models outperform previous methods in both settings. | Usage similarity is a means for representing word meaning which involves assessing in-context semantic similarity, rather than mapping to word senses from external inventories @cite_0 @cite_10 . This methodology followed from the gradual shift from word sense disambiguation models that would select the best sense in context from a dictionary, to models that reason about meaning by solely relying on distributional similarity @cite_5 @cite_2 , or allow multiple sense interpretations @cite_11 . In , the idea is to model meaning in context in a way that captures different degrees of similarity to a word sense, or between word instances. | {
"cite_N": [
"@cite_0",
"@cite_2",
"@cite_5",
"@cite_10",
"@cite_11"
],
"mid": [
"2147809840",
"2137607259",
"1973942085",
"2096819730",
"2250770534"
],
"abstract": [
"The vast majority of work on word senses has relied on predefined sense inventories and an annotation schema where each word instance is tagged with the best fitting sense. This paper examines the case for a graded notion of word meaning in two experiments, one which uses WordNet senses in a graded fashion, contrasted with the \"winner takes all\" annotation, and one which asks annotators to judge the similarity of two usages. We find that the graded responses correlate with annotations from previous datasets, but sense assignments are used in a way that weakens the case for clear cut sense boundaries. The responses from both experiments correlate with the overlap of paraphrases from the English lexical substitution task which bodes well for the use of substitutes as a proxy for word sense. This paper also provides two novel datasets which can be used for evaluating computational systems.",
"This paper proposes a framework for representing the meaning of phrases and sentences in vector space. Central to our approach is vector composition which we operationalize in terms of additive and multiplicative functions. Under this framework, we introduce a wide range of composition models which we evaluate empirically on a sentence similarity task. Experimental results demonstrate that the multiplicative models are superior to the additive alternatives when compared against human judgments.",
"We address the task of computing vector space representations for the meaning of word occurrences, which can vary widely according to context. This task is a crucial step towards a robust, vector-based compositional account of sentence meaning. We argue that existing models for this task do not take syntactic structure sufficiently into account. We present a novel structured vector space model that addresses these issues by incorporating the selectional preferences for words' argument positions. This makes it possible to integrate syntax into the computation of word meaning in context. In addition, the model performs at and above the state of the art for modeling the contextual adequacy of paraphrases.",
"Word sense disambiguation (WSD) is an old and important task in computational linguistics that still remains challenging, to machines as well as to human annotators. Recently there have been several proposals for representing word meaning in context that diverge from the traditional use of a single best sense for each occurrence. They represent word meaning in context through multiple paraphrases, as points in vector space, or as distributions over latent senses. New methods of evaluating and comparing these different representations are needed.In this paper we propose two novel annotation schemes that characterize word meaning in context in a graded fashion. In WSsim annotation, the applicability of each dictionary sense is rated on an ordinal scale. Usim annotation directly rates the similarity of pairs of usages of the same lemma, again on a scale. We find that the novel annotation schemes show good inter-annotator agreement, as well as a strong correlation with traditional single-sense annotation and ...",
"Word sense annotation is a challenging task where annotators distinguish which meaning of a word is present in a given context. In some contexts, a word usage may elicit multiple interpretations, resulting either in annotators disagreeing or in allowing the usage to be annotated with multiple senses. While some works have allowed the latter, the extent to which multiple sense annotations are needed has not been assessed. The present work analyzes a dataset of instances annotated with multiple WordNet senses to assess the causes of the multiple interpretations and their relative frequencies, along with the effect of the multiple senses on the contextual interpretation. We show that contextual underspecification is the primary cause of multiple interpretations but that syllepsis still accounts for more than a third of the cases. In addition, we show that sense coarsening can only partially remove the need for labeling instances with multiple senses and we provide suggestions for how future sense annotation guidelines might be developed to account for this need."
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.