PyTorchKR Papers

최신 논문

총 8,922편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2001.05070

Understanding Generalization in Deep Learning via Tensor Methods

Deep neural networks generalize well on unseen data though the number of parameters often far exceeds the number of training examples. Recently proposed complexity measures have provided insights to understanding the generalizability in neural networks from perspectives of PAC-Bayes, robustness, overparametrization, compression and so on. In this work, we advance the understanding of the relations between the network's architecture and its generalizability from the compression perspective. Using tensor analysis, we propose a series of intuitive, data-dependent and easily-measurable properties that tightly characterize the compressibility and generalizability of neural networks; thus, in practice, our generalization bound outperforms the previous compression-based ones, especially for neural networks using tensors as their weight kernels (e.g. CNNs). Moreover, these intuitive measurements provide further insights into designing neural network architectures with properties favorable for better/guaranteed generalizability. Our experimental results demonstrate that through the proposed measurable properties, our generalization error bound matches the trend of the test error well. Our theoretical analysis further provides justifications for the empirical success and limitations of some widely-used tensor-based compression approaches. We also discover the improvements to the compressibility and robustness of current neural networks when incorporating tensor operations via our proposed layer-wise structure.

arXiv:2001.05058공식 코드 · Papers with Code

Hippocampus Segmentation on Epilepsy and Alzheimer's Disease Studies with Multiple Convolutional Neural Networks

Hippocampus segmentation on magnetic resonance imaging is of key importance for the diagnosis, treatment decision and investigation of neuropsychiatric disorders. Automatic segmentation is an active research field, with many recent models using deep learning. Most current state-of-the art hippocampus segmentation methods train their methods on healthy or Alzheimer's disease patients from public datasets. This raises the question whether these methods are capable of recognizing the hippocampus on a different domain, that of epilepsy patients with hippocampus resection. In this paper we present a state-of-the-art, open source, ready-to-use, deep learning based hippocampus segmentation method. It uses an extended 2D multi-orientation approach, with automatic pre-processing and orientation alignment. The methodology was developed and validated using HarP, a public Alzheimer's disease hippocampus segmentation dataset. We test this methodology alongside other recent deep learning methods, in two domains: The HarP test set and an in-house epilepsy dataset, containing hippocampus resections, named HCUnicamp. We show that our method, while trained only in HarP, surpasses others from the literature in both the HarP test set and HCUnicamp in Dice. Additionally, Results from training and testing in HCUnicamp volumes are also reported separately, alongside comparisons between training and testing in epilepsy and Alzheimer's data and vice versa. Although current state-of-the-art methods, including our own, achieve upwards of 0.9 Dice in HarP, all tested methods, including our own, produced false positives in HCUnicamp resection regions, showing that there is still room for improvement for hippocampus segmentation methods when resection is involved.

arXiv:2001.05343

Causal Discovery from Incomplete Data: A Deep Learning Approach

As systems are getting more autonomous with the development of artificial intelligence, it is important to discover the causal knowledge from observational sensory inputs. By encoding a series of cause-effect relations between events, causal networks can facilitate the prediction of effects from a given action and analyze their underlying data generation mechanism. However, missing data are ubiquitous in practical scenarios. Directly performing existing casual discovery algorithms on partially observed data may lead to the incorrect inference. To alleviate this issue, we proposed a deep learning framework, dubbed Imputated Causal Learning (ICL), to perform iterative missing data imputation and causal structure discovery. Through extensive simulations on both synthetic and real data, we show that ICL can outperform state-of-the-art methods under different missing data mechanisms.

arXiv:2001.05266

Deep Learning for MIR Tutorial

Deep Learning has become state of the art in visual computing and continuously emerges into the Music Information Retrieval (MIR) and audio retrieval domain. In order to bring attention to this topic we propose an introductory tutorial on deep learning for MIR. Besides a general introduction to neural networks, the proposed tutorial covers a wide range of MIR relevant deep learning approaches. \textbf{Convolutional Neural Networks} are currently a de-facto standard for deep learning based audio retrieval. \textbf{Recurrent Neural Networks} have proven to be effective in onset detection tasks such as beat or audio-event detection. \textbf{Siamese Networks} have been shown effective in learning audio representations and distance functions specific for music similarity retrieval. We will incorporate both academic and industrial points of view into the tutorial. Accompanying the tutorial, we will create a Github repository for the content presented at the tutorial as well as references to state of the art work and literature for further reading. This repository will remain public after the conference.

arXiv:2001.04568

180-degree Outpainting from a Single Image

Presenting context images to a viewer's peripheral vision is one of the most effective techniques to enhance immersive visual experiences. However, most images only present a narrow view, since the field-of-view (FoV) of standard cameras is small. To overcome this limitation, we propose a deep learning approach that learns to predict a 180{\deg} panoramic image from a narrow-view image. Specifically, we design a foveated framework that applies different strategies on near-periphery and mid-periphery regions. Two networks are trained separately, and then are employed jointly to sequentially perform narrow-to-90{\deg} generation and 90{\deg}-to-180{\deg} generation. The generated outputs are then fused with their aligned inputs to produce expanded equirectangular images for viewing. Our experimental results show that single-view-to-panoramic image generation using deep learning is both feasible and promising.

arXiv:2001.04813

Review and Prospect: Deep Learning in Nuclear Magnetic Resonance Spectroscopy

Since the concept of Deep Learning (DL) was formally proposed in 2006, it had a major impact on academic research and industry. Nowadays, DL provides an unprecedented way to analyze and process data with demonstrated great results in computer vision, medical imaging, natural language processing, etc. In this Minireview, we summarize applications of DL in Nuclear Magnetic Resonance (NMR) spectroscopy and outline a perspective for DL as entirely new approaches that are likely to transform NMR spectroscopy into a much more efficient and powerful technique in chemistry and life science.

arXiv:2001.04689

Deep Learning for ECG Segmentation

We propose an algorithm for electrocardiogram (ECG) segmentation using a UNet-like full-convolutional neural network. The algorithm receives an arbitrary sampling rate ECG signal as an input, and gives a list of onsets and offsets of P and T waves and QRS complexes as output. Our method of segmentation differs from others in speed, a small number of parameters and a good generalization: it is adaptive to different sampling rates and it is generalized to various types of ECG monitors. The proposed approach is superior to other state-of-the-art segmentation methods in terms of quality. In particular, F1-measures for detection of onsets and offsets of P and T waves and for QRS-complexes are at least 97.8%, 99.5%, and 99.9%, respectively.

arXiv:2001.04663

Effects of annotation granularity in deep learning models for histopathological images

Pathological is crucial to cancer diagnosis. Usually, Pathologists draw their conclusion based on observed cell and tissue structure on histology slides. Rapid development in machine learning, especially deep learning have established robust and accurate classifiers. They are being used to analyze histopathological slides and assist pathologists in diagnosis. Most machine learning systems rely heavily on annotated data sets to gain experiences and knowledge to correctly and accurately perform various tasks such as classification and segmentation. This work investigates different granularity of annotations in histopathological data set including image-wise, bounding box, ellipse-wise, and pixel-wise to verify the influence of annotation in pathological slide on deep learning models. We design corresponding experiments to test classification and segmentation performance of deep learning models based on annotations with different annotation granularity. In classification, state-of-the-art deep learning-based classifiers perform better when trained by pixel-wise annotation dataset. On average, precision, recall and F1-score improves by 7.87%, 8.83% and 7.85% respectively. Thus, it is suggested that finer granularity annotations are better utilized by deep learning algorithms in classification tasks. Similarly, semantic segmentation algorithms can achieve 8.33% better segmentation accuracy when trained by pixel-wise annotations. Our study shows not only that finer-grained annotation can improve the performance of deep learning models, but also help extracts more accurate phenotypic information from histopathological slides. Intelligence systems trained on granular annotations may help pathologists inspecting certain regions for better diagnosis. The compartmentalized prediction approach similar to this work may contribute to phenotype and genotype association studies.

arXiv:2001.04552

Deep Learning Stereo Vision at the edge

We present an overview of the methodology used to build a new stereo vision solution that is suitable for System on Chip. This new solution was developed to bring computer vision capability to embedded devices that live in a power constrained environment. The solution is constructured as a hybrid between classical Stereo Vision techniques and deep learning approaches. The stereoscopic module is composed of two separate modules: one that accelerates the neural network we trained and one that accelerates the front-end part. The system is completely passive and does not require any structured light to obtain very compelling accuracy. With respect to the previous Stereo Vision solutions offered by the industries we offer a major improvement is robustness to noise. This is mainly possible due to the deep learning part of the chosen architecture. We submitted our result to Middlebury dataset challenge. It currently ranks as the best System on Chip solution. The system has been developed for low latency applications which require better than real time performance on high definition videos.

arXiv:2001.04325

A Novel Inspection System For Variable Data Printing Using Deep Learning

We present a novel approach for inspecting variable data prints (VDP) with an ultra-low false alarm rate (0.005%) and potential applicability to other real-world problems. The system is based on a comparison between two images: a reference image and an image captured by low-cost scanners. The comparison task is challenging as low-cost imaging systems create artifacts that may erroneously be classified as true (genuine) defects. To address this challenge we introduce two new fusion methods, for change detection applications, which are both fast and efficient. The first is an early fusion method that combines the two input images into a single pseudo-color image. The second, called Change-Detection Single Shot Detector (CD-SSD) leverages the SSD by fusing features in the middle of the network. We demonstrate the effectiveness of the proposed deep learning-based approach with a large dataset from real-world printing scenarios. Finally, we evaluate our models on a different domain of aerial imagery change detection (AICD). Our best method clearly outperforms the state-of-the-art baseline on this dataset.

arXiv:2001.04193공식 코드 · Papers with Code

Deep Learning for Person Re-identification: A Survey and Outlook

Person re-identification (Re-ID) aims at retrieving a person of interest across multiple non-overlapping cameras. With the advancement of deep neural networks and increasing demand of intelligent video surveillance, it has gained significantly increased interest in the computer vision community. By dissecting the involved components in developing a person Re-ID system, we categorize it into the closed-world and open-world settings. The widely studied closed-world setting is usually applied under various research-oriented assumptions, and has achieved inspiring success using deep learning techniques on a number of datasets. We first conduct a comprehensive overview with in-depth analysis for closed-world person Re-ID from three different perspectives, including deep feature representation learning, deep metric learning and ranking optimization. With the performance saturation under closed-world setting, the research focus for person Re-ID has recently shifted to the open-world setting, facing more challenging issues. This setting is closer to practical applications under specific scenarios. We summarize the open-world Re-ID in terms of five different aspects. By analyzing the advantages of existing methods, we design a powerful AGW baseline, achieving state-of-the-art or at least comparable performance on twelve datasets for FOUR different Re-ID tasks. Meanwhile, we introduce a new evaluation metric (mINP) for person Re-ID, indicating the cost for finding all the correct matches, which provides an additional criteria to evaluate the Re-ID system for real applications. Finally, some important yet under-investigated open issues are discussed.

arXiv:2001.04344

An Explainable Autoencoder For Collaborative Filtering Recommendation

Autoencoders are a common building block of Deep Learning architectures, where they are mainly used for representation learning. They have also been successfully used in Collaborative Filtering (CF) recommender systems to predict missing ratings. Unfortunately, like all black box machine learning models, they are unable to explain their outputs. Hence, while predictions from an Autoencoder-based recommender system might be accurate, it might not be clear to the user why a recommendation was generated. In this work, we design an explainable recommendation system using an Autoencoder model whose predictions can be explained using the neighborhood based explanation style. Our preliminary work can be considered to be the first step towards an explainable deep learning architecture based on Autoencoders.

arXiv:2001.04292

Geometric deep learning for computational mechanics Part I: Anisotropic Hyperelasticity

This paper is the first attempt to use geometric deep learning and Sobolev training to incorporate non-Euclidean microstructural data such that anisotropic hyperelastic material machine learning models can be trained in the finite deformation range. While traditional hyperelasticity models often incorporate homogenized measures of microstructural attributes, such as porosity averaged orientation of constitutes, these measures cannot reflect the topological structures of the attributes. We fill this knowledge gap by introducing the concept of weighted graph as a new mean to store topological information, such as the connectivity of anisotropic grains in assembles. Then, by leveraging a graph convolutional deep neural network architecture in the spectral domain, we introduce a mechanism to incorporate these non-Euclidean weighted graph data directly as input for training and for predicting the elastic responses of materials with complex microstructures. To ensure smoothness and prevent non-convexity of the trained stored energy functional, we introduce a Sobolev training technique for neural networks such that stress measure is obtained implicitly from taking directional derivatives of the trained energy functional. By optimizing the neural network to approximate both the energy functional output and the stress measure, we introduce a training procedure the improves efficiency and generalize the learned energy functional for different microstructures. The trained hybrid neural network model is then used to generate new stored energy functional for unseen microstructures in a parametric study to predict the influence of elastic anisotropy on the nucleation and propagation of fracture in the brittle regime.

arXiv:2001.04206

Towards High Performance Java-based Deep Learning Frameworks

The advent of modern cloud services along with the huge volume of data produced on a daily basis, have set the demand for fast and efficient data processing. This demand is common among numerous application domains, such as deep learning, data mining, and computer vision. Prior research has focused on employing hardware accelerators as a means to overcome this inefficiency. This trend has driven software development to target heterogeneous execution, and several modern computing systems have incorporated a mixture of diverse computing components, including GPUs and FPGAs. However, the specialization of the applications' code for heterogeneous execution is not a trivial task, as it requires developers to have hardware expertise in order to obtain high performance. The vast majority of the existing deep learning frameworks that support heterogeneous acceleration, rely on the implementation of wrapper calls from a high-level programming language to a low-level accelerator backend, such as OpenCL, CUDA or HLS. In this paper we have employed TornadoVM, a state-of-the-art heterogeneous programming framework to transparently accelerate Deep Netts; a Java-based deep learning framework. Our initial results demonstrate up to 8x performance speedup when executing the back propagation process of the network's training on AMD GPUs against the sequential execution of the original Deep Netts framework.

arXiv:2001.04413

Backward Feature Correction: How Deep Learning Performs Deep (Hierarchical) Learning

Deep learning is also known as hierarchical learning, where the learner _learns_ to represent a complicated target function by decomposing it into a sequence of simpler functions to reduce sample and time complexity. This paper formally analyzes how multi-layer neural networks can perform such hierarchical learning _efficiently_ and _automatically_ by SGD on the training objective. On the conceptual side, we present a theoretical characterizations of how certain types of deep (i.e. super-constant layer) neural networks can still be sample and time efficiently trained on some hierarchical tasks, when no existing algorithm (including layerwise training, kernel method, etc) is known to be efficient. We establish a new principle called "backward feature correction", where the errors in the lower-level features can be automatically corrected when training together with the higher-level layers. We believe this is a key behind how deep learning is performing deep (hierarchical) learning, as opposed to layerwise learning or simulating some non-hierarchical method. On the technical side, we show for every input dimension $d > 0$, there is a concept class of degree $\omega(1)$ multi-variate polynomials so that, using $\omega(1)$-layer neural networks as learners, SGD can learn any function from this class in $\mathsf{poly}(d)$ time to any $\frac{1}{\mathsf{poly}(d)}$ error, through learning to represent it as a composition of $\omega(1)$ layers of quadratic functions using "backward feature correction." In contrast, we do not know any other simpler algorithm (including layerwise training, applying kernel method sequentially, training a two-layer network, etc) that can learn this concept class in $\mathsf{poly}(d)$ time even to any $d^{-0.01}$ error. As a side result, we prove $d^{\omega(1)}$ lower bounds for several non-hierarchical learners, including any kernel methods.

arXiv:2001.03560

Biophysical models of cis-regulation as interpretable neural networks

The adoption of deep learning techniques in genomics has been hindered by the difficulty of mechanistically interpreting the models that these techniques produce. In recent years, a variety of post-hoc attribution methods have been proposed for addressing this neural network interpretability problem in the context of gene regulation. Here we describe a complementary way of approaching this problem. Our strategy is based on the observation that two large classes of biophysical models of cis-regulatory mechanisms can be expressed as deep neural networks in which nodes and weights have explicit physiochemical interpretations. We also demonstrate how such biophysical networks can be rapidly inferred, using modern deep learning frameworks, from the data produced by certain types of massively parallel reporter assays (MPRAs). These results suggest a scalable strategy for using MPRAs to systematically characterize the biophysical basis of gene regulation in a wide range of biological contexts. They also highlight gene regulation as a promising venue for the development of scientifically interpretable approaches to deep learning.

arXiv:2001.03350

Two Applications of Deep Learning in the Physical Layer of Communication Systems

Deep learning has proved itself to be a powerful tool to develop data-driven signal processing algorithms for challenging engineering problems. By learning the key features and characteristics of the input signals, instead of requiring a human to first identify and model them, learned algorithms can beat many man-made algorithms. In particular, deep neural networks are capable of learning the complicated features in nature-made signals, such as photos and audio recordings, and use them for classification and decision making. The situation is rather different in communication systems, where the information signals are man-made, the propagation channels are relatively easy to model, and we know how to operate close to the Shannon capacity limits. Does this mean that there is no role for deep learning in the development of future communication systems?

arXiv:2001.03274

Backdoor Attacks against Transfer Learning with Pre-trained Deep Learning Models

Transfer learning provides an effective solution for feasibly and fast customize accurate \textit{Student} models, by transferring the learned knowledge of pre-trained \textit{Teacher} models over large datasets via fine-tuning. Many pre-trained Teacher models used in transfer learning are publicly available and maintained by public platforms, increasing their vulnerability to backdoor attacks. In this paper, we demonstrate a backdoor threat to transfer learning tasks on both image and time-series data leveraging the knowledge of publicly accessible Teacher models, aimed at defeating three commonly-adopted defenses: \textit{pruning-based}, \textit{retraining-based} and \textit{input pre-processing-based defenses}. Specifically, (A) ranking-based selection mechanism to speed up the backdoor trigger generation and perturbation process while defeating \textit{pruning-based} and/or \textit{retraining-based defenses}. (B) autoencoder-powered trigger generation is proposed to produce a robust trigger that can defeat the \textit{input pre-processing-based defense}, while guaranteeing that selected neuron(s) can be significantly activated. (C) defense-aware retraining to generate the manipulated model using reverse-engineered model inputs. We launch effective misclassification attacks on Student models over real-world images, brain Magnetic Resonance Imaging (MRI) data and Electrocardiography (ECG) learning systems. The experiments reveal that our enhanced attack can maintain the $98.4\%$ and $97.2\%$ classification accuracy as the genuine model on clean image and time series inputs respectively while improving $27.9\%-100\%$ and $27.1\%-56.1\%$ attack success rate on trojaned image and time series inputs respectively in the presence of pruning-based and/or retraining-based defenses.

arXiv:2001.03354

Learning credit assignment

Deep learning has achieved impressive prediction accuracies in a variety of scientific and industrial domains. However, the nested non-linear feature of deep learning makes the learning highly non-transparent, i.e., it is still unknown how the learning coordinates a huge number of parameters to achieve a decision making. To explain this hierarchical credit assignment, we propose a mean-field learning model by assuming that an ensemble of sub-networks, rather than a single network, are trained for a classification task. Surprisingly, our model reveals that apart from some deterministic synaptic weights connecting two neurons at neighboring layers, there exist a large number of connections that can be absent, and other connections can allow for a broad distribution of their weight values. Therefore, synaptic connections can be classified into three categories: very important ones, unimportant ones, and those of variability that may partially encode nuisance factors. Therefore, our model learns the credit assignment leading to the decision, and predicts an ensemble of sub-networks that can accomplish the same task, thereby providing insights toward understanding the macroscopic behavior of deep learning through the lens of distinct roles of synaptic weights.

arXiv:2001.02992

Poly-time universality and limitations of deep learning

The goal of this paper is to characterize function distributions that deep learning can or cannot learn in poly-time. A universality result is proved for SGD-based deep learning and a non-universality result is proved for GD-based deep learning; this also gives a separation between SGD-based deep learning and statistical query algorithms: (1) {\it Deep learning with SGD is efficiently universal.} Any function distribution that can be learned from samples in poly-time can also be learned by a poly-size neural net trained with SGD on a poly-time initialization with poly-steps, poly-rate and possibly poly-noise. Therefore deep learning provides a universal learning paradigm: it was known that the approximation and estimation errors could be controlled with poly-size neural nets, using ERM that is NP-hard; this new result shows that the optimization error can also be controlled with SGD in poly-time. The picture changes for GD with large enough batches: (2) {\it Result (1) does not hold for GD:} Neural nets of poly-size trained with GD (full gradients or large enough batches) on any initialization with poly-steps, poly-range and at least poly-noise cannot learn any function distribution that has super-polynomial {\it cross-predictability,} where the cross-predictability gives a measure of ``average'' function correlation -- relations and distinctions to the statistical dimension are discussed. In particular, GD with these constraints can learn efficiently monomials of degree $k$ if and only if $k$ is constant. Thus (1) and (2) point to an interesting contrast: SGD is universal even with some poly-noise while full GD or SQ algorithms are not (e.g., parities).