PyTorchKR Papers

최신 논문

총 8,922편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2008.11865

Traces of Class/Cross-Class Structure Pervade Deep Learning Spectra

Numerous researchers recently applied empirical spectral analysis to the study of modern deep learning classifiers. We identify and discuss an important formal class/cross-class structure and show how it lies at the origin of the many visually striking features observed in deepnet spectra, some of which were reported in recent articles, others are unveiled here for the first time. These include spectral outliers, "spikes", and small but distinct continuous distributions, "bumps", often seen beyond the edge of a "main bulk". The significance of the cross-class structure is illustrated in three ways: (i) we prove the ratio of outliers to bulk in the spectrum of the Fisher information matrix is predictive of misclassification, in the context of multinomial logistic regression; (ii) we demonstrate how, gradually with depth, a network is able to separate class-distinctive information from class variability, all while orthogonalizing the class-distinctive information; and (iii) we propose a correction to KFAC, a well-known second-order optimization algorithm for training deepnets.

arXiv:2008.11849

SparseRT: Accelerating Unstructured Sparsity on GPUs for Deep Learning Inference

In recent years, there has been a flurry of research in deep neural network pruning and compression. Early approaches prune weights individually. However, it is difficult to take advantage of the resulting unstructured sparsity patterns on modern hardware like GPUs. As a result, pruning strategies which impose sparsity structures in the weights have become more popular. However,these structured pruning approaches typically lead to higher losses in accuracy than unstructured pruning. In this paper, we present SparseRT, a code generator that leverage unstructured sparsity to accelerate sparse linear algebra operations in deep learning inference on GPUs. For 1x1 convolutions and fully connected layers, we demonstrate geometric mean of speedups of 3.4x over the equivalent dense computation at 90% sparsity and 5.4x at 95% sparsity when evaluated on hundreds of test cases in deep learning. For sparse 3x3 convolutions, we show speedups of over 5x on use cases in ResNet-50.

arXiv:2008.11757

Deep Learning for Constrained Utility Maximisation

This paper proposes two algorithms for solving stochastic control problems with deep learning, with a focus on the utility maximisation problem. The first algorithm solves Markovian problems via the Hamilton Jacobi Bellman (HJB) equation. We solve this highly nonlinear partial differential equation (PDE) with a second order backward stochastic differential equation (2BSDE) formulation. The convex structure of the problem allows us to describe a dual problem that can either verify the original primal approach or bypass some of the complexity. The second algorithm utilises the full power of the duality method to solve non-Markovian problems, which are often beyond the scope of stochastic control solvers in the existing literature. We solve an adjoint BSDE that satisfies the dual optimality conditions. We apply these algorithms to problems with power, log and non-HARA utilities in the Black-Scholes, the Heston stochastic volatility, and path dependent volatility models. Numerical experiments show highly accurate results with low computational cost, supporting our proposed algorithms.

arXiv:2008.12260공식 코드 · Papers with Code

Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning

Pollux improves scheduling performance in deep learning (DL) clusters by adaptively co-optimizing inter-dependent factors both at the per-job level and at the cluster-wide level. Most existing schedulers expect users to specify the number of resources for each job, often leading to inefficient resource use. Some recent schedulers choose job resources for users, but do so without awareness of how DL training can be re-optimized to better utilize the provided resources. Pollux simultaneously considers both aspects. By monitoring the status of each job during training, Pollux models how their goodput (a novel metric we introduce that combines system throughput with statistical efficiency) would change by adding or removing resources. Leveraging these information, Pollux dynamically (re-)assigns resources to improve cluster-wide goodput, while respecting fairness and continually optimizing each DL job to better utilize those resources. In experiments with real DL jobs and with trace-driven simulations, Pollux reduces average job completion times by 37-50% relative to state-of-the-art DL schedulers, even when they are provided with ideal resource and training configurations for every job. Pollux promotes fairness among DL jobs competing for resources based on a more meaningful measure of useful job progress, and reveals a new opportunity for reducing DL cost in cloud environments. Pollux is implemented and publicly available as part of an open-source project at https://github.com/petuum/adaptdl.

arXiv:2008.12080

Identifying and Tracking Solar Magnetic Flux Elements with Deep Learning

Deep learning has drawn a lot of interest in recent years due to its effectiveness in processing big and complex observational data gathered from diverse instruments. Here we propose a new deep learning method, called SolarUnet, to identify and track solar magnetic flux elements or features in observed vector magnetograms based on the Southwest Automatic Magnetic Identification Suite (SWAMIS). Our method consists of a data pre-processing component that prepares training data from the SWAMIS tool, a deep learning model implemented as a U-shaped convolutional neural network for fast and accurate image segmentation, and a post-processing component that prepares tracking results. SolarUnet is applied to data from the 1.6 meter Goode Solar Telescope at the Big Bear Solar Observatory. When compared to the widely used SWAMIS tool, SolarUnet is faster while agreeing mostly with SWAMIS on feature size and flux distributions, and complementing SWAMIS in tracking long-lifetime features. Thus, the proposed physics-guided deep learning-based tool can be considered as an alternative method for solar magnetic tracking.

arXiv:2008.12066

Minimal Adversarial Examples for Deep Learning on 3D Point Clouds

With recent developments of convolutional neural networks, deep learning for 3D point clouds has shown significant progress in various 3D scene understanding tasks, e.g., object recognition, semantic segmentation. In a safety-critical environment, it is however not well understood how such deep learning models are vulnerable to adversarial examples. In this work, we explore adversarial attacks for point cloud-based neural networks. We propose a unified formulation for adversarial point cloud generation that can generalise two different attack strategies. Our method generates adversarial examples by attacking the classification ability of point cloud-based networks while considering the perceptibility of the examples and ensuring the minimal level of point manipulations. Experimental results show that our method achieves the state-of-the-art performance with higher than 89% and 90% of attack success rate on synthetic and real-world data respectively, while manipulating only about 4% of the total points.

arXiv:2008.11546

Towards Structured Prediction in Bioinformatics with Deep Learning

Using machine learning, especially deep learning, to facilitate biological research is a fascinating research direction. However, in addition to the standard classification or regression problems, in bioinformatics, we often need to predict more complex structured targets, such as 2D images and 3D molecular structures. The above complex prediction tasks are referred to as structured prediction. Structured prediction is more complicated than the traditional classification but has much broader applications, considering that most of the original bioinformatics problems have complex output objects. Due to the properties of those structured prediction problems, such as having problem-specific constraints and dependency within the labeling space, the straightforward application of existing deep learning models can lead to unsatisfactory results. Here, we argue that the following ideas can help resolve structured prediction problems in bioinformatics. Firstly, we can combine deep learning with other classic algorithms, such as probabilistic graphical models, which model the problem structure explicitly. Secondly, we can design the problem-specific deep learning architectures or methods by considering the structured labeling space and problem constraints, either explicitly or implicitly. We demonstrate our ideas with six projects from four bioinformatics subfields, including sequencing analysis, structure prediction, function annotation, and network analysis. The structured outputs cover 1D signals, 2D images, 3D structures, hierarchical labeling, and heterogeneous networks. With the help of the above ideas, all of our methods can achieve SOTA performance on the corresponding problems. The success of these projects motivates us to extend our work towards other more challenging but important problems, such as health-care problems, which can directly benefit people's health and wellness.

arXiv:2008.11632

GuardNN: Secure Accelerator Architecture for Privacy-Preserving Deep Learning

This paper proposes GuardNN, a secure DNN accelerator that provides hardware-based protection for user data and model parameters even in an untrusted environment. GuardNN shows that the architecture and protection can be customized for a specific application to provide strong confidentiality and integrity guarantees with negligible overhead. The design of the GuardNN instruction set reduces the TCB to just the accelerator and allows confidentiality protection even when the instructions from a host cannot be trusted. GuardNN minimizes the overhead of memory encryption and integrity verification by customizing the off-chip memory protection for the known memory access patterns of a DNN accelerator. GuardNN is prototyped on an FPGA, demonstrating effective confidentiality protection with ~3% performance overhead for inference.

arXiv:2008.09768공식 코드 · Papers with Code

Hierarchical Deep Learning of Multiscale Differential Equation Time-Steppers

Nonlinear differential equations rarely admit closed-form solutions, thus requiring numerical time-stepping algorithms to approximate solutions. Further, many systems characterized by multiscale physics exhibit dynamics over a vast range of timescales, making numerical integration computationally expensive due to numerical stiffness. In this work, we develop a hierarchy of deep neural network time-steppers to approximate the flow map of the dynamical system over a disparate range of time-scales. The resulting model is purely data-driven and leverages features of the multiscale dynamics, enabling numerical integration and forecasting that is both accurate and highly efficient. Moreover, similar ideas can be used to couple neural network-based models with classical numerical time-steppers. Our multiscale hierarchical time-stepping scheme provides important advantages over current time-stepping algorithms, including (i) circumventing numerical stiffness due to disparate time-scales, (ii) improved accuracy in comparison with leading neural-network architectures, (iii) efficiency in long-time simulation/forecasting due to explicit training of slow time-scale dynamics, and (iv) a flexible framework that is parallelizable and may be integrated with standard numerical time-stepping algorithms. The method is demonstrated on a wide range of nonlinear dynamical systems, including the Van der Pol oscillator, the Lorenz system, the Kuramoto-Sivashinsky equation, and fluid flow pass a cylinder; audio and video signals are also explored. On the sequence generation examples, we benchmark our algorithm against state-of-the-art methods, such as LSTM, reservoir computing, and clockwork RNN. Despite the structural simplicity of our method, it outperforms competing methods on numerical integration.

arXiv:2008.09697

Perceptual underwater image enhancement with deep learning and physical priors

Underwater image enhancement, as a pre-processing step to improve the accuracy of the following object detection task, has drawn considerable attention in the field of underwater navigation and ocean exploration. However, most of the existing underwater image enhancement strategies tend to consider enhancement and detection as two independent modules with no interaction, and the practice of separate optimization does not always help the underwater object detection task. In this paper, we propose two perceptual enhancement models, each of which uses a deep enhancement model with a detection perceptor. The detection perceptor provides coherent information in the form of gradients to the enhancement model, guiding the enhancement model to generate patch level visually pleasing images or detection favourable images. In addition, due to the lack of training data, a hybrid underwater image synthesis model, which fuses physical priors and data-driven cues, is proposed to synthesize training data and generalise our enhancement model for real-world underwater images. Experimental results show the superiority of our proposed method over several state-of-the-art methods on both real-world and synthetic underwater datasets.

arXiv:2008.09647

Generating synthetic photogrammetric data for training deep learning based 3D point cloud segmentation models

At I/ITSEC 2019, the authors presented a fully-automated workflow to segment 3D photogrammetric point-clouds/meshes and extract object information, including individual tree locations and ground materials (Chen et al., 2019). The ultimate goal is to create realistic virtual environments and provide the necessary information for simulation. We tested the generalizability of the previously proposed framework using a database created under the U.S. Army's One World Terrain (OWT) project with a variety of landscapes (i.e., various buildings styles, types of vegetation, and urban density) and different data qualities (i.e., flight altitudes and overlap between images). Although the database is considerably larger than existing databases, it remains unknown whether deep-learning algorithms have truly achieved their full potential in terms of accuracy, as sizable data sets for training and validation are currently lacking. Obtaining large annotated 3D point-cloud databases is time-consuming and labor-intensive, not only from a data annotation perspective in which the data must be manually labeled by well-trained personnel, but also from a raw data collection and processing perspective. Furthermore, it is generally difficult for segmentation models to differentiate objects, such as buildings and tree masses, and these types of scenarios do not always exist in the collected data set. Thus, the objective of this study is to investigate using synthetic photogrammetric data to substitute real-world data in training deep-learning algorithms. We have investigated methods for generating synthetic UAV-based photogrammetric data to provide a sufficiently sized database for training a deep-learning algorithm with the ability to enlarge the data size for scenarios in which deep-learning models have difficulties.

arXiv:2008.10293

Bosch Deep Learning Hardware Benchmark

The widespread use of Deep Learning (DL) applications in science and industry has created a large demand for efficient inference systems. This has resulted in a rapid increase of available Hardware Accelerators (HWAs) making comparison challenging and laborious. To address this, several DL hardware benchmarks have been proposed aiming at a comprehensive comparison for many models, tasks, and hardware platforms. Here, we present our DL hardware benchmark which has been specifically developed for inference on embedded HWAs and tasks required for autonomous driving. In addition to previous benchmarks, we propose a new granularity level to evaluate common submodules of DL models, a twofold benchmark procedure that accounts for hardware and model optimizations done by HWA manufacturers, and an extended set of performance indicators that can help to identify a mismatch between a HWA and the DL models used in our benchmark.

arXiv:2008.10805

New Directions in Distributed Deep Learning: Bringing the Network at Forefront of IoT Design

In this paper, we first highlight three major challenges to large-scale adoption of deep learning at the edge: (i) Hardware-constrained IoT devices, (ii) Data security and privacy in the IoT era, and (iii) Lack of network-aware deep learning algorithms for distributed inference across multiple IoT devices. We then provide a unified view targeting three research directions that naturally emerge from the above challenges: (1) Federated learning for training deep networks, (2) Data-independent deployment of learning algorithms, and (3) Communication-aware distributed inference. We believe that the above research directions need a network-centric approach to enable the edge intelligence and, therefore, fully exploit the true potential of IoT.

arXiv:2008.10678

Probabilistic Deep Learning for Instance Segmentation

Probabilistic convolutional neural networks, which predict distributions of predictions instead of point estimates, led to recent advances in many areas of computer vision, from image reconstruction to semantic segmentation. Besides state of the art benchmark results, these networks made it possible to quantify local uncertainties in the predictions. These were used in active learning frameworks to target the labeling efforts of specialist annotators or to assess the quality of a prediction in a safety-critical environment. However, for instance segmentation problems these methods are not frequently used so far. We seek to close this gap by proposing a generic method to obtain model-inherent uncertainty estimates within proposal-free instance segmentation models. Furthermore, we analyze the quality of the uncertainty estimates with a metric adapted from semantic segmentation. We evaluate our method on the BBBC010 C.\ elegans dataset, where it yields competitive performance while also predicting uncertainty estimates that carry information about object-level inaccuracies like false splits and false merges. We perform a simulation to show the potential use of such uncertainty estimates in guided proofreading.

arXiv:2008.10351공식 코드 · Papers with Code

Model Generalization in Deep Learning Applications for Land Cover Mapping

Recent work has shown that deep learning models can be used to classify land-use data from geospatial satellite imagery. We show that when these deep learning models are trained on data from specific continents/seasons, there is a high degree of variability in model performance on out-of-sample continents/seasons. This suggests that just because a model accurately predicts land-use classes in one continent or season does not mean that the model will accurately predict land-use classes in a different continent or season. We then use clustering techniques on satellite imagery from different continents to visualize the differences in landscapes that make geospatial generalization particularly difficult, and summarize our takeaways for future satellite imagery-related applications.

arXiv:2008.09874

Multiple Classification with Split Learning

Privacy issues were raised in the process of training deep learning in medical, mobility, and other fields. To solve this problem, we present privacy-preserving distributed deep learning method that allow clients to learn a variety of data without direct exposure. We divided a single deep learning architecture into a common extractor, a cloud model and a local classifier for the distributed learning. First, the common extractor, which is used by local clients, extracts secure features from the input data. The secure features also take the role that the cloud model can employ various task and diverse types of data. The feature contain the most important information that helps to proceed various task. Second, the cloud model including most parts of the whole training model gets the embedded features from the massive local clients, and performs most of deep learning operations which takes severe computing cost. After the operations in cloud model finished, outputs of the cloud model send back to local clients. Finally, the local classifier determined classification results and delivers the results to local clients. When clients train models, our model does not directly expose sensitive information to exterior network. During the test, the average performance improvement was 2.63% over the existing local training model. However, in a distributed environment, there is a possibility of inversion attack due to exposed features. For this reason, we experimented with the common extractor to prevent data restoration. The quality of restoration of the original image was tested by adjusting the depth of the common extractor. As a result, we found that the deeper the common extractor, the restoration score decreased to 89.74.

arXiv:2008.09168공식 코드 · Papers with Code

A Systematic Assessment of Deep Learning Models for Molecule Generation

In recent years the scientific community has devoted much effort in the development of deep learning models for the generation of new molecules with desirable properties (i.e. drugs). This has produced many proposals in literature. However, a systematic comparison among the different VAE methods is still missing. For this reason, we propose an extensive testbed for the evaluation of generative models for drug discovery, and we present the results obtained by many of the models proposed in literature.

arXiv:2008.09586공식 코드 · Papers with Code

An Overview of Deep-Learning-Based Audio-Visual Speech Enhancement and Separation

Speech enhancement and speech separation are two related tasks, whose purpose is to extract either one or more target speech signals, respectively, from a mixture of sounds generated by several sources. Traditionally, these tasks have been tackled using signal processing and machine learning techniques applied to the available acoustic signals. Since the visual aspect of speech is essentially unaffected by the acoustic environment, visual information from the target speakers, such as lip movements and facial expressions, has also been used for speech enhancement and speech separation systems. In order to efficiently fuse acoustic and visual information, researchers have exploited the flexibility of data-driven approaches, specifically deep learning, achieving strong performance. The ceaseless proposal of a large number of techniques to extract features and fuse multimodal information has highlighted the need for an overview that comprehensively describes and discusses audio-visual speech enhancement and separation based on deep learning. In this paper, we provide a systematic survey of this research topic, focusing on the main elements that characterise the systems in the literature: acoustic features; visual features; deep learning methods; fusion techniques; training targets and objective functions. In addition, we review deep-learning-based methods for speech reconstruction from silent videos and audio-visual sound source separation for non-speech signals, since these methods can be more or less directly applied to audio-visual speech enhancement and separation. Finally, we survey commonly employed audio-visual speech datasets, given their central role in the development of data-driven approaches, and evaluation methods, because they are generally used to compare different systems and determine their performance.

arXiv:2008.09264공식 코드 · Papers with Code

CITISEN: A Deep Learning-Based Speech Signal-Processing Mobile Application

This study presents a deep learning-based speech signal-processing mobile application known as CITISEN. The CITISEN provides three functions: speech enhancement (SE), model adaptation (MA), and background noise conversion (BNC), allowing CITISEN to be used as a platform for utilizing and evaluating SE models and flexibly extend the models to address various noise environments and users. For SE, a pretrained SE model downloaded from the cloud server is used to effectively reduce noise components from instant or saved recordings provided by users. For encountering unseen noise or speaker environments, the MA function is applied to promote CITISEN. A few audio samples recording on a noisy environment are uploaded and used to adapt the pretrained SE model on the server. Finally, for BNC, CITISEN first removes the background noises through an SE model and then mixes the processed speech with new background noise. The novel BNC function can evaluate SE performance under specific conditions, cover people's tracks, and provide entertainment. The experimental results confirmed the effectiveness of SE, MA, and BNC functions. Compared with the noisy speech signals, the enhanced speech signals achieved about 6\% and 33\% of improvements, respectively, in terms of short-time objective intelligibility (STOI) and perceptual evaluation of speech quality (PESQ). With MA, the STOI and PESQ could be further improved by approximately 6\% and 11\%, respectively. Finally, the BNC experiment results indicated that the speech signals converted from noisy and silent backgrounds have a close scene identification accuracy and similar embeddings in an acoustic scene classification model. Therefore, the proposed BNC can effectively convert the background noise of a speech signal and be a data augmentation method when clean speech signals are unavailable.

arXiv:2008.09104

A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises

Since its renaissance, deep learning has been widely used in various medical imaging tasks and has achieved remarkable success in many medical imaging applications, thereby propelling us into the so-called artificial intelligence (AI) era. It is known that the success of AI is mostly attributed to the availability of big data with annotations for a single task and the advances in high performance computing. However, medical imaging presents unique challenges that confront deep learning approaches. In this survey paper, we first present traits of medical imaging, highlight both clinical needs and technical challenges in medical imaging, and describe how emerging trends in deep learning are addressing these issues. We cover the topics of network architecture, sparse and noisy labels, federating learning, interpretability, uncertainty quantification, etc. Then, we present several case studies that are commonly found in clinical practice, including digital pathology and chest, brain, cardiovascular, and abdominal imaging. Rather than presenting an exhaustive literature survey, we instead describe some prominent research highlights related to these case study applications. We conclude with a discussion and presentation of promising future directions.