April 2, 2020

3526 words 17 mins read

Paper Group ANR 104

Paper Group ANR 104

VAE/WGAN-Based Image Representation Learning For Pose-Preserving Seamless Identity Replacement In Facial Images. Preventing Imitation Learning with Adversarial Policy Ensembles. Ethics in the digital era. Online Meta-Critic Learning for Off-Policy Actor-Critic Methods. Topologically Densified Distributions. Optimization of Graph Total Variation via …

VAE/WGAN-Based Image Representation Learning For Pose-Preserving Seamless Identity Replacement In Facial Images

Title VAE/WGAN-Based Image Representation Learning For Pose-Preserving Seamless Identity Replacement In Facial Images
Authors Hiroki Kawai, Jiawei Chen, Prakash Ishwar, Janusz Konrad
Abstract We present a novel variational generative adversarial network (VGAN) based on Wasserstein loss to learn a latent representation from a face image that is invariant to identity but preserves head-pose information. This facilitates synthesis of a realistic face image with the same head pose as a given input image, but with a different identity. One application of this network is in privacy-sensitive scenarios; after identity replacement in an image, utility, such as head pose, can still be recovered. Extensive experimental validation on synthetic and real human-face image datasets performed under 3 threat scenarios confirms the ability of the proposed network to preserve head pose of the input image, mask the input identity, and synthesize a good-quality realistic face image of a desired identity. We also show that our network can be used to perform pose-preserving identity morphing and identity-preserving pose morphing. The proposed method improves over a recent state-of-the-art method in terms of quantitative metrics as well as synthesized image quality.
Tasks Representation Learning
Published 2020-03-02
URL https://arxiv.org/abs/2003.00641v1
PDF https://arxiv.org/pdf/2003.00641v1.pdf
PWC https://paperswithcode.com/paper/vaewgan-based-image-representation-learning

Preventing Imitation Learning with Adversarial Policy Ensembles

Title Preventing Imitation Learning with Adversarial Policy Ensembles
Authors Albert Zhan, Stas Tiomkin, Pieter Abbeel
Abstract Imitation learning can reproduce policies by observing experts, which poses a problem regarding policy privacy. Policies, such as human, or policies on deployed robots, can all be cloned without consent from the owners. How can we protect against external observers cloning our proprietary policies? To answer this question we introduce a new reinforcement learning framework, where we train an ensemble of near-optimal policies, whose demonstrations are guaranteed to be useless for an external observer. We formulate this idea by a constrained optimization problem, where the objective is to improve proprietary policies, and at the same time deteriorate the virtual policy of an eventual external observer. We design a tractable algorithm to solve this new optimization problem by modifying the standard policy gradient algorithm. Our formulation can be interpreted in lenses of confidentiality and adversarial behaviour, which enables a broader perspective of this work. We demonstrate the existence of “non-clonable” ensembles, providing a solution to the above optimization problem, which is calculated by our modified policy gradient algorithm. To our knowledge, this is the first work regarding the protection of policies in Reinforcement Learning.
Tasks Imitation Learning
Published 2020-01-31
URL https://arxiv.org/abs/2002.01059v1
PDF https://arxiv.org/pdf/2002.01059v1.pdf
PWC https://paperswithcode.com/paper/preventing-imitation-learning-with-1

Ethics in the digital era

Title Ethics in the digital era
Authors David Pastor-Escuredo
Abstract Ethics is an ancient matter for human kind, from the origin of civilizations ethics have been related with the most relevant human concerns and determined human behavior. Ethics was initially related to religion, politics and philosophy to then be fragmented into specific disciplines and communities of practice. The undergoing digital revolution enabled by Artificial Intelligence and Big Data are bringing ethical wicked problems in the social application of these technologies. However, a broader perspective is also necessary. We now face global challenges that affect groups and individuals, specially those that are most vulnerable, but cannot reduced only to individual-oriented solutions. Thus, ethics has to consider the several scales in which the current complex society is organized and the interconnections between different systems. Ethics should also give a response to the systemic changes in individual to collective behavior produced by external factors and threats. Furthermore, Artificial Intelligence and digital technologies are global and make humans more connected and smart but also more homogeneous, predictable and ultimately controllable. Ethics must take a stand to preserve and keep promoting individuals rights and uniqueness and cultural heterogeneity. The digital revolution has been so far an industry-driven movement, so it is necessary to establish mechanisms to ensure that the society becomes conscious about its own future. Finally, Artificial Intelligence has advanced through the ambition to humanize matter, so we should expect ethics to give a response to the future status of machines and their interactions with humans.
Published 2020-03-14
URL https://arxiv.org/abs/2003.06530v2
PDF https://arxiv.org/pdf/2003.06530v2.pdf
PWC https://paperswithcode.com/paper/ethics-in-the-digital-era

Online Meta-Critic Learning for Off-Policy Actor-Critic Methods

Title Online Meta-Critic Learning for Off-Policy Actor-Critic Methods
Authors Wei Zhou, Yiying Li, Yongxin Yang, Huaimin Wang, Timothy M. Hospedales
Abstract Off-Policy Actor-Critic (Off-PAC) methods have proven successful in a variety of continuous control tasks. Normally, the critic’s action-value function is updated using temporal-difference, and the critic in turn provides a loss for the actor that trains it to take actions with higher expected return. In this paper, we introduce a novel and flexible meta-critic that observes the learning process and meta-learns an additional loss for the actor that accelerates and improves actor-critic learning. Compared to the vanilla critic, the meta-critic network is explicitly trained to accelerate the learning process; and compared to existing meta-learning algorithms, meta-critic is rapidly learned online for a single task, rather than slowly over a family of tasks. Crucially, our meta-critic framework is designed for off-policy based learners, which currently provide state-of-the-art reinforcement learning sample efficiency. We demonstrate that online meta-critic learning leads to improvements in avariety of continuous control environments when combined with contemporary Off-PAC methods DDPG, TD3 and the state-of-the-art SAC.
Tasks Continuous Control, Meta-Learning
Published 2020-03-11
URL https://arxiv.org/abs/2003.05334v1
PDF https://arxiv.org/pdf/2003.05334v1.pdf
PWC https://paperswithcode.com/paper/online-meta-critic-learning-for-off-policy-1

Topologically Densified Distributions

Title Topologically Densified Distributions
Authors Christoph D. Hofer, Florian Graf, Marc Niethammer, Roland Kwitt
Abstract We study regularization in the context of small sample-size learning with over-parameterized neural networks. Specifically, we shift focus from architectural properties, such as norms on the network weights, to properties of the internal representations before a linear classifier. Specifically, we impose a topological constraint on samples drawn from the probability measure induced in that space. This provably leads to mass concentration effects around the representations of training instances, i.e., a property beneficial for generalization. By leveraging previous work to impose topological constraints in a neural network setting, we provide empirical evidence (across various vision benchmarks) to support our claim for better generalization.
Published 2020-02-12
URL https://arxiv.org/abs/2002.04805v1
PDF https://arxiv.org/pdf/2002.04805v1.pdf
PWC https://paperswithcode.com/paper/topologically-densified-distributions

Optimization of Graph Total Variation via Active-Set-based Combinatorial Reconditioning

Title Optimization of Graph Total Variation via Active-Set-based Combinatorial Reconditioning
Authors Zhenzhang Ye, Thomas Möllenhoff, Tao Wu, Daniel Cremers
Abstract Structured convex optimization on weighted graphs finds numerous applications in machine learning and computer vision. In this work, we propose a novel adaptive preconditioning strategy for proximal algorithms on this problem class. Our preconditioner is driven by a sharp analysis of the local linear convergence rate depending on the “active set” at the current iterate. We show that nested-forest decomposition of the inactive edges yields a guaranteed local linear convergence rate. Further, we propose a practical greedy heuristic which realizes such nested decompositions and show in several numerical experiments that our reconditioning strategy, when applied to proximal gradient or primal-dual hybrid gradient algorithm, achieves competitive performances. Our results suggest that local convergence analysis can serve as a guideline for selecting variable metrics in proximal algorithms.
Published 2020-02-27
URL https://arxiv.org/abs/2002.12236v1
PDF https://arxiv.org/pdf/2002.12236v1.pdf
PWC https://paperswithcode.com/paper/optimization-of-graph-total-variation-via

Transfer Learning for Information Extraction with Limited Data

Title Transfer Learning for Information Extraction with Limited Data
Authors Minh-Tien Nguyen, Viet-Anh Phan, Le Thai Linh, Nguyen Hong Son, Le Tien Dung, Miku Hirano, Hajime Hotta
Abstract This paper presents a practical approach to fine-grained information extraction. Through plenty of experiences of authors in practically applying information extraction to business process automation, there can be found a couple of fundamental technical challenges: (i) the availability of labeled data is usually limited and (ii) highly detailed classification is required. The main idea of our proposal is to leverage the concept of transfer learning, which is to reuse the pre-trained model of deep neural networks, with a combination of common statistical classifiers to determine the class of each extracted term. To do that, we first exploit BERT to deal with the limitation of training data in real scenarios, then stack BERT with Convolutional Neural Networks to learn hidden representation for classification. To validate our approach, we applied our model to an actual case of document processing, which is a process of competitive bids for government projects in Japan. We used 100 documents for training and testing and confirmed that the model enables to extract fine-grained named entities with a detailed level of information preciseness specialized in the targeted business process, such as a department name of application receivers.
Tasks Transfer Learning
Published 2020-03-06
URL https://arxiv.org/abs/2003.03064v1
PDF https://arxiv.org/pdf/2003.03064v1.pdf
PWC https://paperswithcode.com/paper/transfer-learning-for-information-extraction

Phase Transitions for the Information Bottleneck in Representation Learning

Title Phase Transitions for the Information Bottleneck in Representation Learning
Authors Tailin Wu, Ian Fischer
Abstract In the Information Bottleneck (IB), when tuning the relative strength between compression and prediction terms, how do the two terms behave, and what’s their relationship with the dataset and the learned representation? In this paper, we set out to answer these questions by studying multiple phase transitions in the IB objective: $\text{IB}_\beta[p(zx)] = I(X; Z) - \beta I(Y; Z)$ defined on the encoding distribution p(zx) for input $X$, target $Y$ and representation $Z$, where sudden jumps of $dI(Y; Z)/d \beta$ and prediction accuracy are observed with increasing $\beta$. We introduce a definition for IB phase transitions as a qualitative change of the IB loss landscape, and show that the transitions correspond to the onset of learning new classes. Using second-order calculus of variations, we derive a formula that provides a practical condition for IB phase transitions, and draw its connection with the Fisher information matrix for parameterized models. We provide two perspectives to understand the formula, revealing that each IB phase transition is finding a component of maximum (nonlinear) correlation between $X$ and $Y$ orthogonal to the learned representation, in close analogy with canonical-correlation analysis (CCA) in linear settings. Based on the theory, we present an algorithm for discovering phase transition points. Finally, we verify that our theory and algorithm accurately predict phase transitions in categorical datasets, predict the onset of learning new classes and class difficulty in MNIST, and predict prominent phase transitions in CIFAR10.
Tasks Representation Learning
Published 2020-01-07
URL https://arxiv.org/abs/2001.01878v1
PDF https://arxiv.org/pdf/2001.01878v1.pdf
PWC https://paperswithcode.com/paper/phase-transitions-for-the-information-1

CASIA-SURF CeFA: A Benchmark for Multi-modal Cross-ethnicity Face Anti-spoofing

Title CASIA-SURF CeFA: A Benchmark for Multi-modal Cross-ethnicity Face Anti-spoofing
Authors Ajian Li, Zichang Tan, Xuan Li, Jun Wan, Sergio Escalera, Guodong Guo, Stan Z. Li
Abstract Ethnic bias has proven to negatively affect the performance of face recognition systems, and it remains an open research problem in face anti-spoofing. In order to study the ethnic bias for face anti-spoofing, we introduce the largest up to date CASIA-SURF Cross-ethnicity Face Anti-spoofing (CeFA) dataset (briefly named CeFA), covering $3$ ethnicities, $3$ modalities, $1,607$ subjects, and 2D plus 3D attack types. Four protocols are introduced to measure the affect under varied evaluation conditions, such as cross-ethnicity, unknown spoofs or both of them. To the best of our knowledge, CeFA is the first dataset including explicit ethnic labels in current published/released datasets for face anti-spoofing. Then, we propose a novel multi-modal fusion method as a strong baseline to alleviate these bias, namely, the static-dynamic fusion mechanism applied in each modality (i.e., RGB, Depth and infrared image). Later, a partially shared fusion strategy is proposed to learn complementary information from multiple modalities. Extensive experiments demonstrate that the proposed method achieves state-of-the-art results on the CASIA-SURF, OULU-NPU, SiW and the CeFA dataset.
Tasks Face Anti-Spoofing, Face Recognition
Published 2020-03-11
URL https://arxiv.org/abs/2003.05136v1
PDF https://arxiv.org/pdf/2003.05136v1.pdf
PWC https://paperswithcode.com/paper/casia-surf-cefa-a-benchmark-for-multi-modal

Attack of the Genes: Finding Keys and Parameters of Locked Analog ICs Using Genetic Algorithm

Title Attack of the Genes: Finding Keys and Parameters of Locked Analog ICs Using Genetic Algorithm
Authors Rabin Yu Acharya, Sreeja Chowdhury, Fatemeh Ganji, Domenic Forte
Abstract Hardware intellectual property (IP) theft is a major issue in today’s globalized supply chain. To address it, numerous logic locking and obfuscation techniques have been proposed. While locking initially focused on digital integrated circuits (ICs), there have been recent attempts to extend it to analog ICs, which are easier to reverse engineer and to copy than digital ICs. In this paper, we use algorithms based on evolutionary strategies to investigate the security of analog obfuscation/locking techniques. We present a genetic algorithm (GA) approach which is capable of completely breaking a locked analog circuit by finding either its obfuscation key or its obfuscated parameters. We implement both the GA attack as well as a more naive satisfiability modulo theory (SMT)-based attack on common analog benchmark circuits obfuscated by combinational locking and parameter biasing. We find that GA attack can unlock all the circuits using only the locked netlist and an unlocked chip in minutes. On the other hand, while the SMT attack converges faster, it requires circuit specification to execute and it also returns multiple keys that need to be brute-forced by a post-processing step. We also discuss how the GA attack can generalize to other recent analog locking techniques not tested in the paper
Published 2020-03-31
URL https://arxiv.org/abs/2003.13904v1
PDF https://arxiv.org/pdf/2003.13904v1.pdf
PWC https://paperswithcode.com/paper/attack-of-the-genes-finding-keys-and

Deep learning for brake squeal: vibration detection, characterization and prediction

Title Deep learning for brake squeal: vibration detection, characterization and prediction
Authors Merten Stender, Merten Tiedemann, David Spieler, Daniel Schoepflin, Norbert Hofffmann, Sebastian Oberst
Abstract Despite significant advances in numerical modeling of brake squeal, the majority of industrial research and design is still conducted experimentally. In this work we report on novel strategies for handling data-intensive vibration testings and gaining better insights into brake system vibrations. To this end, we propose machine learning-based methods to detect and characterize vibrations, understand sensitivities and predict brake squeal. Our aim is to illustrate how interdisciplinary approaches can leverage the potential of data science techniques for classical mechanical engineering challenges. In the first part, a deep learning brake squeal detector is developed to identify several classes of typical sounds in vibration recordings. The detection method is rooted in recent computer vision techniques for object detection. It allows to overcome limitations of classical approaches that rely on spectral properties of the recorded vibrations. Results indicate superior detection and characterization quality when compared to state-of-the-art brake squeal detectors. In the second part, deep recurrent neural networks are employed to learn the parametric patterns that determine the dynamic stability of the brake system during operation. Given a set of multivariate loading conditions, the models learn to predict the vibrational behavior of the structure. The validated models represent virtual twins for the squeal behavior of a specific brake system. It is found that those models can predict the occurrence and onset of brake squeal with high accuracy. Hence, the deep learning models can identify the complicated patterns and temporal dependencies in the loading conditions that drive the dynamical structure into regimes of instability. Large data sets from commercial brake system testing are used to train and validate the deep learning models.
Tasks Object Detection
Published 2020-01-02
URL https://arxiv.org/abs/2001.01596v1
PDF https://arxiv.org/pdf/2001.01596v1.pdf
PWC https://paperswithcode.com/paper/deep-learning-for-brake-squeal-vibration

Indirect Adversarial Attacks via Poisoning Neighbors for Graph Convolutional Networks

Title Indirect Adversarial Attacks via Poisoning Neighbors for Graph Convolutional Networks
Authors Tsubasa Takahashi
Abstract Graph convolutional neural networks, which learn aggregations over neighbor nodes, have achieved great performance in node classification tasks. However, recent studies reported that such graph convolutional node classifier can be deceived by adversarial perturbations on graphs. Abusing graph convolutions, a node’s classification result can be influenced by poisoning its neighbors. Given an attributed graph and a node classifier, how can we evaluate robustness against such indirect adversarial attacks? Can we generate strong adversarial perturbations which are effective on not only one-hop neighbors, but more far from the target? In this paper, we demonstrate that the node classifier can be deceived with high-confidence by poisoning just a single node even two-hops or more far from the target. Towards achieving the attack, we propose a new approach which searches smaller perturbations on just a single node far from the target. In our experiments, our proposed method shows 99% attack success rate within two-hops from the target in two datasets. We also demonstrate that m-layer graph convolutional neural networks have chance to be deceived by our indirect attack within m-hop neighbors. The proposed attack can be used as a benchmark in future defense attempts to develop graph convolutional neural networks with having adversary robustness.
Tasks Node Classification
Published 2020-02-19
URL https://arxiv.org/abs/2002.08012v1
PDF https://arxiv.org/pdf/2002.08012v1.pdf
PWC https://paperswithcode.com/paper/indirect-adversarial-attacks-via-poisoning

Line Hypergraph Convolution Network: Applying Graph Convolution for Hypergraphs

Title Line Hypergraph Convolution Network: Applying Graph Convolution for Hypergraphs
Authors Sambaran Bandyopadhyay, Kishalay Das, M. Narasimha Murty
Abstract Network representation learning and node classification in graphs got significant attention due to the invent of different types graph neural networks. Graph convolution network (GCN) is a popular semi-supervised technique which aggregates attributes within the neighborhood of each node. Conventional GCNs can be applied to simple graphs where each edge connects only two nodes. But many modern days applications need to model high order relationships in a graph. Hypergraphs are effective data types to handle such complex relationships. In this paper, we propose a novel technique to apply graph convolution on hypergraphs with variable hyperedge sizes. We use the classical concept of line graph of a hypergraph for the first time in the hypergraph learning literature. Then we propose to use graph convolution on the line graph of a hypergraph. Experimental analysis on multiple real world network datasets shows the merit of our approach compared to state-of-the-arts.
Tasks Node Classification, Representation Learning
Published 2020-02-09
URL https://arxiv.org/abs/2002.03392v1
PDF https://arxiv.org/pdf/2002.03392v1.pdf
PWC https://paperswithcode.com/paper/line-hypergraph-convolution-network-applying

Read Beyond the Lines: Understanding the Implied Textual Meaning via a Skim and Intensive Reading Model

Title Read Beyond the Lines: Understanding the Implied Textual Meaning via a Skim and Intensive Reading Model
Authors Guoxiu He, Zhe Gao, Zhuoren Jiang, Yangyang Kang, Changlong Sun, Xiaozhong Liu, Wei Lu
Abstract The nonliteral interpretation of a text is hard to be understood by machine models due to its high context-sensitivity and heavy usage of figurative language. In this study, inspired by human reading comprehension, we propose a novel, simple, and effective deep neural framework, called Skim and Intensive Reading Model (SIRM), for figuring out implied textual meaning. The proposed SIRM consists of two main components, namely the skim reading component and intensive reading component. N-gram features are quickly extracted from the skim reading component, which is a combination of several convolutional neural networks, as skim (entire) information. An intensive reading component enables a hierarchical investigation for both local (sentence) and global (paragraph) representation, which encapsulates the current embedding and the contextual information with a dense connection. More specifically, the contextual information includes the near-neighbor information and the skim information mentioned above. Finally, besides the normal training loss function, we employ an adversarial loss function as a penalty over the skim reading component to eliminate noisy information arisen from special figurative words in the training data. To verify the effectiveness, robustness, and efficiency of the proposed architecture, we conduct extensive comparative experiments on several sarcasm benchmarks and an industrial spam dataset with metaphors. Experimental results indicate that (1) the proposed model, which benefits from context modeling and consideration of figurative language, outperforms existing state-of-the-art solutions, with comparable parameter scale and training speed; (2) the SIRM yields superior robustness in terms of parameter size sensitivity; (3) compared with ablation and addition variants of the SIRM, the final framework is efficient enough.
Tasks Reading Comprehension
Published 2020-01-03
URL https://arxiv.org/abs/2001.00572v2
PDF https://arxiv.org/pdf/2001.00572v2.pdf
PWC https://paperswithcode.com/paper/read-beyond-the-lines-understanding-the

LogicGAN: Logic-guided Generative Adversarial Networks

Title LogicGAN: Logic-guided Generative Adversarial Networks
Authors Laura Graves, Vineel Nagisetty, Joseph Scott, Vijay Ganesh
Abstract Generative Adversarial Networks (GANs) are a revolutionary class of Deep Neural Networks (DNNs) that have been successfully used to generate realistic images, music, text, and other data. However, it is well known that GAN training can be notoriously resource-intensive and presents many challenges. Further, a potential weakness in GANs is that discriminator DNNs typically provide only one value (loss) of corrective feedback to generator DNNs (namely, the discriminator’s assessment of the generated example). By contrast, we propose a new class of GAN we refer to as LogicGAN, that leverages recent advances in (logic-based) explainable AI (xAI) systems to provide a “richer” form of corrective feedback from discriminators to generators. Specifically, we modify the gradient descent process using xAI systems that specify the reason as to why the discriminator made the classification it did, thus providing the richer corrective feedback that helps the generator to better fool the discriminator. Using our approach, we show that LogicGANs learn much faster on MNIST data, achieving an improvement in data efficiency of 45% in single and 12.73% in multi-class setting over standard GANs while maintaining the same quality as measured by Fr'echet Inception Distance. Further, we argue that LogicGAN enables users greater control over how models learn than standard GAN systems.
Published 2020-02-24
URL https://arxiv.org/abs/2002.10438v1
PDF https://arxiv.org/pdf/2002.10438v1.pdf
PWC https://paperswithcode.com/paper/logicgan-logic-guided-generative-adversarial
comments powered by Disqus