January 30, 2020

2951 words 14 mins read

Paper Group ANR 321

Paper Group ANR 321

Learning to Flip Successive Cancellation Decoding of Polar Codes with LSTM Networks. Tactile Hallucinations on Artificial Skin Induced by Homeostasis in a Deep Boltzmann Machine. Something’s Brewing! Early Prediction of Controversy-causing Posts from Discussion Features. Convolutional Auto-encoding of Sentence Topics for Image Paragraph Generation. …

Learning to Flip Successive Cancellation Decoding of Polar Codes with LSTM Networks

Title Learning to Flip Successive Cancellation Decoding of Polar Codes with LSTM Networks
Authors Xianbin Wang, Huazi Zhang, Rong Li, Lingchen Huang, Shengchen Dai, Yourui Huangfu, Jun Wang
Abstract The key to successive cancellation (SC) flip decoding of polar codes is to accurately identify the first error bit. The optimal flipping strategy is considered difficult due to lack of an analytical solution. Alternatively, we propose a deep learning aided SC flip algorithm. Specifically, before each SC decoding attempt, a long short-term memory (LSTM) network is exploited to either (i) locate the first error bit, or (ii) undo a previous wrong' flip. In each SC attempt, the sequence of log likelihood ratios (LLRs) derived in the previous SC attempt is exploited to decide which action to take. Accordingly, a two-stage training method of the LSTM network is proposed, i.e., learn to locate first error bits in the first stage, and then to undo wrong’ flips in the second stage. Simulation results show that the proposed approach identifies error bits more accurately and achieves better performance than the state-of-the-art SC flip algorithms.
Tasks
Published 2019-02-22
URL http://arxiv.org/abs/1902.08394v2
PDF http://arxiv.org/pdf/1902.08394v2.pdf
PWC https://paperswithcode.com/paper/learning-to-flip-successive-cancellation
Repo
Framework

Tactile Hallucinations on Artificial Skin Induced by Homeostasis in a Deep Boltzmann Machine

Title Tactile Hallucinations on Artificial Skin Induced by Homeostasis in a Deep Boltzmann Machine
Authors Michael Deistler, Yagmur Yener, Florian Bergner, Pablo Lanillos, Gordon Cheng
Abstract Perceptual hallucinations are present in neurological and psychiatric disorders and amputees. While the hallucinations can be drug-induced, it has been described that they can even be provoked in healthy subjects. Understanding their manifestation could thus unveil how the brain processes sensory information and might evidence the generative nature of perception. In this work, we investigate the generation of tactile hallucinations on biologically inspired, artificial skin. To model tactile hallucinations, we apply homeostasis, a change in the excitability of neurons during sensory deprivation, in a Deep Boltzmann Machine (DBM). We find that homeostasis prompts hallucinations of previously learned patterns on the artificial skin in the absence of sensory input. Moreover, we show that homeostasis is capable of inducing the formation of meaningful latent representations in a DBM and that it significantly increases the quality of the reconstruction of these latent states. Through this, our work provides a possible explanation for the nature of tactile hallucinations and highlights homeostatic processes as a potential underlying mechanism.
Tasks
Published 2019-06-25
URL https://arxiv.org/abs/1906.10592v2
PDF https://arxiv.org/pdf/1906.10592v2.pdf
PWC https://paperswithcode.com/paper/tactile-hallucinations-on-artificial-skin
Repo
Framework

Something’s Brewing! Early Prediction of Controversy-causing Posts from Discussion Features

Title Something’s Brewing! Early Prediction of Controversy-causing Posts from Discussion Features
Authors Jack Hessel, Lillian Lee
Abstract Controversial posts are those that split the preferences of a community, receiving both significant positive and significant negative feedback. Our inclusion of the word “community” here is deliberate: what is controversial to some audiences may not be so to others. Using data from several different communities on reddit.com, we predict the ultimate controversiality of posts, leveraging features drawn from both the textual content and the tree structure of the early comments that initiate the discussion. We find that even when only a handful of comments are available, e.g., the first 5 comments made within 15 minutes of the original post, discussion features often add predictive capacity to strong content-and-rate only baselines. Additional experiments on domain transfer suggest that conversation-structure features often generalize to other communities better than conversation-content features do.
Tasks
Published 2019-04-15
URL http://arxiv.org/abs/1904.07372v1
PDF http://arxiv.org/pdf/1904.07372v1.pdf
PWC https://paperswithcode.com/paper/somethings-brewing-early-prediction-of
Repo
Framework

Convolutional Auto-encoding of Sentence Topics for Image Paragraph Generation

Title Convolutional Auto-encoding of Sentence Topics for Image Paragraph Generation
Authors Jing Wang, Yingwei Pan, Ting Yao, Jinhui Tang, Tao Mei
Abstract Image paragraph generation is the task of producing a coherent story (usually a paragraph) that describes the visual content of an image. The problem nevertheless is not trivial especially when there are multiple descriptive and diverse gists to be considered for paragraph generation, which often happens in real images. A valid question is how to encapsulate such gists/topics that are worthy of mention from an image, and then describe the image from one topic to another but holistically with a coherent structure. In this paper, we present a new design — Convolutional Auto-Encoding (CAE) that purely employs convolutional and deconvolutional auto-encoding framework for topic modeling on the region-level features of an image. Furthermore, we propose an architecture, namely CAE plus Long Short-Term Memory (dubbed as CAE-LSTM), that novelly integrates the learnt topics in support of paragraph generation. Technically, CAE-LSTM capitalizes on a two-level LSTM-based paragraph generation framework with attention mechanism. The paragraph-level LSTM captures the inter-sentence dependency in a paragraph, while sentence-level LSTM is to generate one sentence which is conditioned on each learnt topic. Extensive experiments are conducted on Stanford image paragraph dataset, and superior results are reported when comparing to state-of-the-art approaches. More remarkably, CAE-LSTM increases CIDEr performance from 20.93% to 25.15%.
Tasks
Published 2019-08-01
URL https://arxiv.org/abs/1908.00249v1
PDF https://arxiv.org/pdf/1908.00249v1.pdf
PWC https://paperswithcode.com/paper/convolutional-auto-encoding-of-sentence
Repo
Framework

Geometric Brain Surface Network For Brain Cortical Parcellation

Title Geometric Brain Surface Network For Brain Cortical Parcellation
Authors Wen Zhang, Yalin Wang
Abstract A large number of surface-based analyses on brain imaging data adopt some specific brain atlases to better assess structural and functional changes in one or more brain regions. In these analyses, it is necessary to obtain an anatomically correct surface parcellation scheme in an individual brain by referring to the given atlas. Traditional ways to accomplish this goal are through a designed surface-based registration or hand-crafted surface features, although both of them are time-consuming. A recent deep learning approach depends on a regular spherical parameterization of the mesh, which is computationally prohibitive in some cases and may also demand further post-processing to refine the network output. Therefore, an accurate and fully-automatic cortical surface parcellation scheme directly working on the original brain surfaces would be highly advantageous. In this study, we propose an end-to-end deep brain cortical parcellation network, called \textbf{DBPN}. Through intrinsic and extrinsic graph convolution kernels, DBPN dynamically deciphers neighborhood graph topology around each vertex and encodes the deciphered knowledge into node features. Eventually, a non-linear mapping between the node features and parcellation labels is constructed. Our model is a two-stage deep network which contains a coarse parcellation network with a U-shape structure and a refinement network to fine-tune the coarse results. We evaluate our model in a large public dataset and our work achieves superior performance than state-of-the-art baseline methods in both accuracy and efficiency
Tasks
Published 2019-09-13
URL https://arxiv.org/abs/1909.13834v1
PDF https://arxiv.org/pdf/1909.13834v1.pdf
PWC https://paperswithcode.com/paper/geometric-brain-surface-network-for-brain
Repo
Framework

Exact expressions for double descent and implicit regularization via surrogate random design

Title Exact expressions for double descent and implicit regularization via surrogate random design
Authors Michał Dereziński, Feynman Liang, Michael W. Mahoney
Abstract Double descent refers to the phase transition that is exhibited by the generalization error of unregularized learning models when varying the ratio between the number of parameters and the number of training samples. The recent success of highly over-parameterized machine learning models such as deep neural networks has motivated a theoretical analysis of the double descent phenomenon in classical models such as linear regression which can also generalize well in the over-parameterized regime. We build on recent advances in Randomized Numerical Linear Algebra (RandNLA) to provide the first exact non-asymptotic expressions for double descent of the minimum norm linear estimator. Our approach involves constructing what we call a surrogate random design to replace the standard i.i.d. design of the training sample. This surrogate design admits exact expressions for the mean squared error of the estimator while preserving the key properties of the standard design. We also establish an exact implicit regularization result for over-parameterized training samples. In particular, we show that, for the surrogate design, the implicit bias of the unregularized minimum norm estimator precisely corresponds to solving a ridge-regularized least squares problem on the population distribution.
Tasks
Published 2019-12-10
URL https://arxiv.org/abs/1912.04533v2
PDF https://arxiv.org/pdf/1912.04533v2.pdf
PWC https://paperswithcode.com/paper/exact-expressions-for-double-descent-and
Repo
Framework

GANs ‘N Lungs: improving pneumonia prediction

Title GANs ‘N Lungs: improving pneumonia prediction
Authors Tatiana Malygina, Elena Ericheva, Ivan Drokin
Abstract We propose a novel method to improve deep learning model performance on highly-imbalanced tasks. The proposed method is based on CycleGAN to achieve balanced dataset. We show that data augmentation with GAN helps to improve accuracy of pneumonia binary classification task even if the generative network was trained on the same training dataset.
Tasks Data Augmentation
Published 2019-08-01
URL https://arxiv.org/abs/1908.00433v1
PDF https://arxiv.org/pdf/1908.00433v1.pdf
PWC https://paperswithcode.com/paper/gans-n-lungs-improving-pneumonia-prediction
Repo
Framework

Fashion Outfit Complementary Item Retrieval

Title Fashion Outfit Complementary Item Retrieval
Authors Yen-Liang Lin, Son Tran, Larry S. Davis
Abstract Complementary fashion item recommendation is critical for fashion outfit completion. Existing methods mainly focus on outfit compatibility prediction but not in a retrieval setting. We propose a new framework for outfit complementary item retrieval. Specifically, a category-based subspace attention network is presented, which is a scalable approach for learning the subspace attentions. In addition, we introduce an outfit ranking loss that better models the item relationships of an entire outfit. We evaluate our method on the outfit compatibility, FITB and new retrieval tasks. Experimental results demonstrate that our approach outperforms state-of-the-art methods in both compatibility prediction and complementary item retrieval
Tasks
Published 2019-12-19
URL https://arxiv.org/abs/1912.08967v2
PDF https://arxiv.org/pdf/1912.08967v2.pdf
PWC https://paperswithcode.com/paper/fashion-outfit-complementary-item-retrieval
Repo
Framework

Early Diagnosis of Pneumonia with Deep Learning

Title Early Diagnosis of Pneumonia with Deep Learning
Authors Can Jozef Saul, Deniz Yagmur Urey, Can Doruk Taktakoglu
Abstract Pneumonia has been one of the fatal diseases and has the potential to result in severe consequences within a short period of time, due to the flow of fluid in lungs, which leads to drowning. If not acted upon by drugs at the right time, pneumonia may result in death of individuals. Therefore, the early diagnosis is a key factor along the progress of the disease. This paper focuses on the biological progress of pneumonia and its detection by x-ray imaging, overviews the studies conducted on enhancing the level of diagnosis, and presents the methodology and results of an automation of xray images based on various parameters in order to detect the disease at very early stages. In this study we propose our deep learning architecture for the classification task, which is trained with modified images, through multiple steps of preprocessing. Our classification method uses convolutional neural networks and residual network architecture for classifying the images. Our findings yield an accuracy of 78.73%, surpassing the previously top scoring accuracy of 76.8%.
Tasks
Published 2019-04-01
URL http://arxiv.org/abs/1904.00937v1
PDF http://arxiv.org/pdf/1904.00937v1.pdf
PWC https://paperswithcode.com/paper/early-diagnosis-of-pneumonia-with-deep
Repo
Framework

Computer Systems Have 99 Problems, Let’s Not Make Machine Learning Another One

Title Computer Systems Have 99 Problems, Let’s Not Make Machine Learning Another One
Authors David Mohaisen, Songqing Chen
Abstract Machine learning techniques are finding many applications in computer systems, including many tasks that require decision making: network optimization, quality of service assurance, and security. We believe machine learning systems are here to stay, and to materialize on their potential we advocate a fresh look at various key issues that need further attention, including security as a requirement and system complexity, and how machine learning systems affect them. We also discuss reproducibility as a key requirement for sustainable machine learning systems, and leads to pursuing it.
Tasks Decision Making
Published 2019-11-28
URL https://arxiv.org/abs/1911.12593v1
PDF https://arxiv.org/pdf/1911.12593v1.pdf
PWC https://paperswithcode.com/paper/computer-systems-have-99-problems-lets-not
Repo
Framework

$H_\infty$ Model-free Reinforcement Learning with Robust Stability Guarantee

Title $H_\infty$ Model-free Reinforcement Learning with Robust Stability Guarantee
Authors Minghao Han, Yuan Tian, Lixian Zhang, Jun Wang, Wei Pan
Abstract Reinforcement learning is showing great potentials in robotics applications, including autonomous driving, robot manipulation and locomotion. However, with complex uncertainties in the real-world environment, it is difficult to guarantee the successful generalization and sim-to-real transfer of learned policies theoretically. In this paper, we introduce and extend the idea of robust stability and $H_\infty$ control to design policies with both stability and robustness guarantee. Specifically, a sample-based approach for analyzing the Lyapunov stability and performance robustness of a learning-based control system is proposed. Based on the theoretical results, a maximum entropy algorithm is developed for searching Lyapunov function and designing a policy with provable robust stability guarantee. Without any specific domain knowledge, our method can find a policy that is robust to various uncertainties and generalizes well to different test environments. In our experiments, we show that our method achieves better robustness to both large impulsive disturbances and parametric variations in the environment than the state-of-art results in both robust and generic RL, as well as classic control. Anonymous code is available to reproduce the experimental results at https://github.com/RobustStabilityGuaranteeRL/RobustStabilityGuaranteeRL.
Tasks Autonomous Driving
Published 2019-11-07
URL https://arxiv.org/abs/1911.02875v2
PDF https://arxiv.org/pdf/1911.02875v2.pdf
PWC https://paperswithcode.com/paper/h_inf-model-free-reinforcement-learning-with
Repo
Framework

Similarities between policy gradient methods (PGM) in Reinforcement learning (RL) and supervised learning (SL)

Title Similarities between policy gradient methods (PGM) in Reinforcement learning (RL) and supervised learning (SL)
Authors Eric Benhamou
Abstract Reinforcement learning (RL) is about sequential decision making and is traditionally opposed to supervised learning (SL) and unsupervised learning (USL). In RL, given the current state, the agent makes a decision that may influence the next state as opposed to SL (and USL) where, the next state remains the same, regardless of the decisions taken, either in batch or online learning. Although this difference is fundamental between SL and RL, there are connections that have been overlooked. In particular, we prove in this paper that gradient policy method can be cast as a supervised learning problem where true label are replaced with discounted rewards. We provide a new proof of policy gradient methods (PGM) that emphasizes the tight link with the cross entropy and supervised learning. We provide a simple experiment where we interchange label and pseudo rewards. We conclude that other relationships with SL could be made if we modify the reward functions wisely.
Tasks Decision Making, Policy Gradient Methods
Published 2019-04-12
URL https://arxiv.org/abs/1904.06260v3
PDF https://arxiv.org/pdf/1904.06260v3.pdf
PWC https://paperswithcode.com/paper/similarities-between-policy-gradient-methods
Repo
Framework

Intrinsic dimension estimation for locally undersampled data

Title Intrinsic dimension estimation for locally undersampled data
Authors Vittorio Erba, Marco Gherardi, Pietro Rotondo
Abstract High-dimensional data are ubiquitous in contemporary science and finding methods to compress them is one of the primary goals of machine learning. Given a dataset lying in a high-dimensional space (in principle hundreds to several thousands of dimensions), it is often useful to project it onto a lower-dimensional manifold, without loss of information. Identifying the minimal dimension of such manifold is a challenging problem known in the literature as intrinsic dimension estimation (IDE). Traditionally, most IDE algorithms are either based on multiscale principal component analysis (PCA) or on the notion of correlation dimension (and more in general on k-nearest-neighbors distances). These methods are affected, in different ways, by a severe curse of dimensionality. In particular, none of the existing algorithms can provide accurate ID estimates in the extreme locally undersampled regime, i.e. in the limit where the number of samples in any local patch of the manifold is less than (or of the same order of) the ID of the dataset. Here we introduce a new ID estimator that leverages on simple properties of the tangent space of a manifold to overcome these shortcomings. The method is based on the full correlation integral, going beyond the limit of small radius used for the estimation of the correlation dimension. Our estimator alleviates the extreme undersampling problem, intractable with other methods. Based on this insight, we explore a multiscale generalization of the algorithm. We show that it is capable of (i) identifying multiple dimensionalities in a dataset, and (ii) providing accurate estimates of the ID of extremely curved manifolds. In particular, we test the method on manifolds generated from global transformations of high-contrast images, relevant for invariant object recognition and considered a challenge for state-of-the-art ID estimators.
Tasks Object Recognition
Published 2019-06-18
URL https://arxiv.org/abs/1906.07670v2
PDF https://arxiv.org/pdf/1906.07670v2.pdf
PWC https://paperswithcode.com/paper/intrinsic-dimension-estimation-for-locally
Repo
Framework

Commuting Conditional GANs for Robust Multi-Modal Fusion

Title Commuting Conditional GANs for Robust Multi-Modal Fusion
Authors Siddharth Roheda, Hamid Krim, Benjamin S. Riggan
Abstract This paper presents a data driven approach to multi-modal fusion, where optimal features for each sensor are selected from a common hidden space between the different modalities. The existence of such a hidden space is then used in order to detect damaged sensors and safeguard the performance of the system. Experimental results show that such an approach can make the system robust against noisy/damaged sensors, without requiring human intervention to inform the system about the damage.
Tasks
Published 2019-06-10
URL https://arxiv.org/abs/1906.04115v2
PDF https://arxiv.org/pdf/1906.04115v2.pdf
PWC https://paperswithcode.com/paper/commuting-conditional-gans-for-robust-multi
Repo
Framework

Stochastic Variational Inference via Upper Bound

Title Stochastic Variational Inference via Upper Bound
Authors Chunlin Ji, Haige Shen
Abstract Stochastic variational inference (SVI) plays a key role in Bayesian deep learning. Recently various divergences have been proposed to design the surrogate loss for variational inference. We present a simple upper bound of the evidence as the surrogate loss. This evidence upper bound (EUBO) equals to the log marginal likelihood plus the KL-divergence between the posterior and the proposal. We show that the proposed EUBO is tighter than previous upper bounds introduced by $\chi$-divergence or $\alpha$-divergence. To facilitate scalable inference, we present the numerical approximation of the gradient of the EUBO and apply the SGD algorithm to optimize the variational parameters iteratively. Simulation study with Bayesian logistic regression shows that the upper and lower bounds well sandwich the evidence and the proposed upper bound is favorably tight. For Bayesian neural network, the proposed EUBO-VI algorithm outperforms state-of-the-art results for various examples.
Tasks
Published 2019-12-02
URL https://arxiv.org/abs/1912.00650v1
PDF https://arxiv.org/pdf/1912.00650v1.pdf
PWC https://paperswithcode.com/paper/stochastic-variational-inference-via-upper
Repo
Framework
comments powered by Disqus