PyTorchKR Papers

최신 논문

총 8,922편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1907.13418

Uncertainty Quantification in Deep Learning for Safer Neuroimage Enhancement

Deep learning (DL) has shown great potential in medical image enhancement problems, such as super-resolution or image synthesis. However, to date, little consideration has been given to uncertainty quantification over the output image. Here we introduce methods to characterise different components of uncertainty in such problems and demonstrate the ideas using diffusion MRI super-resolution. Specifically, we propose to account for $intrinsic$ uncertainty through a heteroscedastic noise model and for $parameter$ uncertainty through approximate Bayesian inference, and integrate the two to quantify $predictive$ uncertainty over the output image. Moreover, we introduce a method to propagate the predictive uncertainty on a multi-channelled image to derived scalar parameters, and separately quantify the effects of intrinsic and parameter uncertainty therein. The methods are evaluated for super-resolution of two different signal representations of diffusion MR images---DTIs and Mean Apparent Propagator MRI---and their derived quantities such as MD and FA, on multiple datasets of both healthy and pathological human brains. Results highlight three key benefits of uncertainty modelling for improving the safety of DL-based image enhancement systems. Firstly, incorporating uncertainty improves the predictive performance even when test data departs from training data. Secondly, the predictive uncertainty highly correlates with errors, and is therefore capable of detecting predictive "failures". Results demonstrate that such an uncertainty measure enables subject-specific and voxel-wise risk assessment of the output images. Thirdly, we show that the method for decomposing predictive uncertainty into its independent sources provides high-level "explanations" for the performance by quantifying how much uncertainty arises from the inherent difficulty of the task or the limited training examples.

arXiv:1907.13368공식 코드 · Papers with Code

Towards Digital Retina in Smart Cities: A Model Generation, Utilization and Communication Paradigm

The digital retina in smart cities is to select what the City Eye tells the City Brain, and convert the acquired visual data from front-end visual sensors to features in an intelligent sensing manner. By deploying deep learning and/or handcrafted models in front-end devices, the compact features can be extracted and subsequently delivered to back-end cloud for search and advanced analytics. In this context, we propose a model generation, utilization, and communication paradigm, aiming to address a set of unique challenges for better artificial intelligence services in smart cities. In particular, we present an integrated multiple deep learning models reuse and prediction strategy, which greatly increases the feasibility of the digital retina in processing and analyzing the large-scale visual data in smart cities. The promise of the proposed paradigm is demonstrated through a set of experiments.

arXiv:1907.13359공식 코드 · Papers with Code

Deep Neural Network Hyperparameter Optimization with Orthogonal Array Tuning

Deep learning algorithms have achieved excellent performance lately in a wide range of fields (e.g., computer version). However, a severe challenge faced by deep learning is the high dependency on hyper-parameters. The algorithm results may fluctuate dramatically under the different configuration of hyper-parameters. Addressing the above issue, this paper presents an efficient Orthogonal Array Tuning Method (OATM) for deep learning hyper-parameter tuning. We describe the OATM approach in five detailed steps and elaborate on it using two widely used deep neural network structures (Recurrent Neural Networks and Convolutional Neural Networks). The proposed method is compared to the state-of-the-art hyper-parameter tuning methods including manually (e.g., grid search and random search) and automatically (e.g., Bayesian Optimization) ones. The experiment results state that OATM can significantly save the tuning time compared to the state-of-the-art methods while preserving the satisfying performance. The codes are open in GitHub (https://github.com/xiangzhang1015/OATM)

arXiv:1907.13257

Optimizing Multi-GPU Parallelization Strategies for Deep Learning Training

Deploying deep learning (DL) models across multiple compute devices to train large and complex models continues to grow in importance because of the demand for faster and more frequent training. Data parallelism (DP) is the most widely used parallelization strategy, but as the number of devices in data parallel training grows, so does the communication overhead between devices. Additionally, a larger aggregate batch size per step leads to statistical efficiency loss, i.e., a larger number of epochs are required to converge to a desired accuracy. These factors affect overall training time and beyond a certain number of devices, the speedup from leveraging DP begins to scale poorly. In addition to DP, each training step can be accelerated by exploiting model parallelism (MP). This work explores hybrid parallelization, where each data parallel worker is comprised of more than one device, across which the model dataflow graph (DFG) is split using MP. We show that at scale, hybrid training will be more effective at minimizing end-to-end training time than exploiting DP alone. We project that for Inception-V3, GNMT, and BigLSTM, the hybrid strategy provides an end-to-end training speedup of at least 26.5%, 8%, and 22% respectively compared to what DP alone can achieve at scale.

arXiv:1907.13216

Deep Learning Training on the Edge with Low-Precision Posits

Recently, the posit numerical format has shown promise for DNN data representation and compute with ultra-low precision ([5..8]-bit). However, majority of studies focus only on DNN inference. In this work, we propose DNN training using posits and compare with the floating point training. We evaluate on both MNIST and Fashion MNIST corpuses, where 16-bit posits outperform 16-bit floating point for end-to-end DNN training.

arXiv:1907.12900

Slot Based Image Augmentation System for Object Detection

Object Detection has been a significant topic in computer vision. As the continuous development of Deep Learning, many advanced academic and industrial outcomes are established on localising and classifying the target objects, such as instance segmentation, video tracking and robotic vision. As the core concept of Deep Learning, Deep Neural Networks (DNNs) and associated training are highly integrated with task-driven modelling, having great effects on accurate detection. The main focus of improving detection performance is proposing DNNs with extra layers and novel topological connections to extract the desired features from input data. However, training these models can be computationally expensive and laborious progress as the complicated model architecture and enormous parameters. Besides, the dataset is another reason causing this issue and low detection accuracy, because of insufficient data samples or difficult instances. To address these training difficulties, this thesis presents two different approaches to improve the detection performance in the relatively light-weight way. As the intrinsic feature of data-driven in deep learning, the first approach is "slot-based image augmentation" to enrich the dataset with extra foreground and background combinations. Instead of the commonly used image flipping method, the proposed system achieved similar mAP improvement with less extra images which decrease training time. This proposed augmentation system has extra flexibility adapting to various scenarios and the performance-driven analysis provides an alternative aspect of conducting image augmentation

arXiv:1907.12740

Deep Learning in Video Multi-Object Tracking: A Survey

The problem of Multiple Object Tracking (MOT) consists in following the trajectory of different objects in a sequence, usually a video. In recent years, with the rise of Deep Learning, the algorithms that provide a solution to this problem have benefited from the representational power of deep models. This paper provides a comprehensive survey on works that employ Deep Learning models to solve the task of MOT on single-camera videos. Four main steps in MOT algorithms are identified, and an in-depth review of how Deep Learning was employed in each one of these stages is presented. A complete experimental comparison of the presented works on the three MOTChallenge datasets is also provided, identifying a number of similarities among the top-performing methods and presenting some possible future research directions.

arXiv:1907.12739

Deep Learning For Face Recognition: A Critical Analysis

Face recognition is a rapidly developing and widely applied aspect of biometric technologies. Its applications are broad, ranging from law enforcement to consumer applications, and industry efficiency and monitoring solutions. The recent advent of affordable, powerful GPUs and the creation of huge face databases has drawn research focus primarily on the development of increasingly deep neural networks designed for all aspects of face recognition tasks, ranging from detection and preprocessing to feature representation and classification in verification and identification solutions. However, despite these improvements, real-time, accurate face recognition is still a challenge, primarily due to the high computational cost associated with the use of Deep Convolutions Neural Networks (DCNN), and the need to balance accuracy requirements with time and resource constraints. Other significant issues affecting face recognition relate to occlusion, illumination and pose invariance, which causes a notable decline in accuracy in both traditional handcrafted solutions and deep neural networks. This survey will provide a critical analysis and comparison of modern state of the art methodologies, their benefits, and their limitations. It provides a comprehensive coverage of both deep and shallow solutions, as they stand today, and highlight areas requiring future development and improvement. This review is aimed at facilitating research into novel approaches, and further development of current methodologies by scientists and engineers, whilst imparting an informative and analytical perspective on currently available solutions to end users in industry, government and consumer contexts.

arXiv:1907.12647

Mapping road safety features from streetview imagery: A deep learning approach

Each year, around 6 million car accidents occur in the U.S. on average. Road safety features (e.g., concrete barriers, metal crash barriers, rumble strips) play an important role in preventing or mitigating vehicle crashes. Accurate maps of road safety features is an important component of safety management systems for federal or state transportation agencies, helping traffic engineers identify locations to invest on safety infrastructure. In current practice, mapping road safety features is largely done manually (e.g., observations on the road or visual interpretation of streetview imagery), which is both expensive and time consuming. In this paper, we propose a deep learning approach to automatically map road safety features from streetview imagery. Unlike existing Convolutional Neural Networks (CNNs) that classify each image individually, we propose to further add Recurrent Neural Network (Long Short Term Memory) to capture geographic context of images (spatial autocorrelation effect along linear road network paths). Evaluations on real world streetview imagery show that our proposed model outperforms several baseline methods.

arXiv:1907.12975

Deep Learning architectures for generalized immunofluorescence based nuclear image segmentation

Separating and labeling each instance of a nucleus (instance-aware segmentation) is the key challenge in segmenting single cell nuclei on fluorescence microscopy images. Deep Neural Networks can learn the implicit transformation of a nuclear image into a probability map indicating the class membership of each pixel (nucleus or background), but the use of post-processing steps to turn the probability map into a labeled object mask is error-prone. This especially accounts for nuclear images of tissue sections and nuclear images across varying tissue preparations. In this work, we aim to evaluate the performance of state-of-the-art deep learning architectures to segment nuclei in fluorescence images of various tissue origins and sample preparation types without post-processing. We compare architectures that operate on pixel to pixel translation and an architecture that operates on object detection and subsequent locally applied segmentation. In addition, we propose a novel strategy to create artificial images to extend the training set. We evaluate the influence of ground truth annotation quality, image scale and segmentation complexity on segmentation performance. Results show that three out of four deep learning architectures (U-Net, U-Net with ResNet34 backbone, Mask R-CNN) can segment fluorescent nuclear images on most of the sample preparation types and tissue origins with satisfactory segmentation performance. Mask R-CNN, an architecture designed to address instance aware segmentation tasks, outperforms other architectures. Equal nuclear mean size, consistent nuclear annotations and the use of artificially generated images result in overall acceptable precision and recall across different tissues and sample preparation types.

arXiv:1907.12368

Detecting Radical Text over Online Media using Deep Learning

Social Media has influenced the way people socially connect, interact and opinionize. The growth in technology has enhanced communication and dissemination of information. Unfortunately,many terror groups like jihadist communities have started consolidating a virtual community online for various purposes such as recruitment, online donations, targeting youth online and spread of extremist ideologies. Everyday a large number of articles, tweets, posts, posters, blogs, comments, views and news are posted online without a check which in turn imposes a threat to the security of any nation. However, different agencies are working on getting down this radical content from various online social media platforms. The aim of our paper is to utilise deep learning algorithm in detection of radicalization contrary to the existing works based on machine learning algorithms. An LSTM based feed forward neural network is employed to detect radical content. We collected total 61601 records from various online sources constituting news, articles and blogs. These records are annotated by domain experts into three categories: Radical(R), Non-Radical (NR) and Irrelevant(I) which are further applied to LSTM based network to classify radical content. A precision of 85.9% has been achieved with the proposed approach

arXiv:1907.12363

A comparison of Deep Learning performances with other machine learning algorithms on credit scoring unbalanced data

Training models on highly unbalanced data is admitted to be a challenging task for machine learning algorithms. Current studies on deep learning mainly focus on data sets with balanced class labels or unbalanced data, but with massive amount of samples available, like in speech recognition. However, the capacities of deep learning on imbalanced data with little samples is not deeply investigated in literature, while it is a very common application context in numerous industries. To contribute to fill this gap, this paper compares the performances of several popular machine learning algorithms previously applied with success to unbalanced data set with deep learning algorithms. We conduct those experiments on a highly unbalanced data set, used for credit scoring. We evaluate various configuration including neural network optimization techniques and try to determine their capacities when they operate with imbalanced corpora.

arXiv:1907.11849

Genetic Deep Learning for Lung Cancer Screening

Convolutional neural networks (CNNs) have shown great promise in improving computer aided detection (CADe). From classifying tumors found via mammography as benign or malignant to automated detection of colorectal polyps in CT colonography, these advances have helped reduce the need for further evaluation with invasive testing and prevent errors from missed diagnoses by acting as a second observer in today's fast paced and high volume clinical environment. CADe methods have become faster and more precise thanks to innovations in deep learning over the past several years. With advancements such as the inception module and utilization of residual connections, the approach to designing CNN architectures has become an art. It is customary to use proven models and fine tune them for particular tasks given a dataset, often requiring tedious work. We investigated using a genetic algorithm (GA) to conduct a neural architectural search (NAS) to generate a novel CNN architecture to find early stage lung cancer in chest x-rays (CXR). Using a dataset of over twelve thousand biopsy proven cases of lung cancer, the trained classification model achieved an accuracy of 97.15% with a PPV of 99.88% and a NPV of 94.81%, beating models such as Inception-V3 and ResNet-152 while simultaneously reducing the number of parameters a factor of 4 and 14, respectively.

arXiv:1907.11836

Deep Learning for CSI Feedback Based on Superimposed Coding

Massive multiple-input multiple-output (MIMO) with frequency division duplex (FDD) mode is a promising approach to increasing system capacity and link robustness for the fifth generation (5G) wireless cellular systems. The premise of these advantages is the accurate downlink channel state information (CSI) fed back from user equipment. However, conventional feedback methods have difficulties in reducing feedback overhead due to significant amount of base station (BS) antennas in massive MIMO systems. Recently, deep learning (DL)-based CSI feedback conquers many difficulties, yet still shows insufficiency to decrease the occupation of uplink bandwidth resources. In this paper, to solve this issue, we combine DL and superimposed coding (SC) for CSI feedback, in which the downlink CSI is spread and then superimposed on uplink user data sequences (UL-US) toward the BS. Then, a multi-task neural network (NN) architecture is proposed at BS to recover the downlink CSI and UL-US by unfolding two iterations of the minimum mean-squared error (MMSE) criterion-based interference reduction. In addition, for a network training, a subnet-by-subnet approach is exploited to facilitate the parameter tuning and expedite the convergence rate. Compared with standalone SC-based CSI scheme, our multi-task NN, trained in a specific signal-to-noise ratio (SNR) and power proportional coefficient (PPC), consistently improves the estimation of downlink CSI with similar or better UL-US detection under SNR and PPC varying.

arXiv:1907.11813

Learning to design from humans: Imitating human designers through deep learning

Humans as designers have quite versatile problem-solving strategies. Computer agents on the other hand can access large scale computational resources to solve certain design problems. Hence, if agents can learn from human behavior, a synergetic human-agent problem solving team can be created. This paper presents an approach to extract human design strategies and implicit rules, purely from historical human data, and use that for design generation. A two-step framework that learns to imitate human design strategies from observation is proposed and implemented. This framework makes use of deep learning constructs to learn to generate designs without any explicit information about objective and performance metrics. The framework is designed to interact with the problem through a visual interface as humans did when solving the problem. It is trained to imitate a set of human designers by observing their design state sequences without inducing problem-specific modelling bias or extra information about the problem. Furthermore, an end-to-end agent is developed that uses this deep learning framework as its core in conjunction with image processing to map pixel-to-design moves as a mechanism to generate designs. Finally, the designs generated by a computational team of these agents are then compared to actual human data for teams solving a truss design problem. Results demonstrates that these agents are able to create feasible and efficient truss designs without guidance, showing that this methodology allows agents to learn effective design strategies.

arXiv:1907.11711

Deep MRI Reconstruction: Unrolled Optimization Algorithms Meet Neural Networks

Image reconstruction from undersampled k-space data has been playing an important role for fast MRI. Recently, deep learning has demonstrated tremendous success in various fields and also shown potential to significantly speed up MR reconstruction with reduced measurements. This article gives an overview of deep learning-based image reconstruction methods for MRI. Three types of deep learning-based approaches are reviewed, the data-driven, model-driven and integrated approaches. The main structure of each network in three approaches is explained and the analysis of common parts of reviewed networks and differences in-between are highlighted. Based on the review, a number of signal processing issues are discussed for maximizing the potential of deep reconstruction for fast MRI. the discussion may facilitate further development of "optimal" network and performance analysis from a theoretical point of view.

arXiv:1907.11377공식 코드 · Papers with Code

Deep Learning Detection of Inaccurate Smart Electricity Meters: A Case Study

Detecting inaccurate smart meters and targeting them for replacement can save significant resources. For this purpose, a novel deep-learning method was developed based on long short-term memory (LSTM) and a modified convolutional neural network (CNN) to predict electricity usage trajectories based on historical data. From the significant difference between the predicted trajectory and the observed one, the meters that cannot measure electricity accurately are located. In a case study, a proof of principle was demonstrated in detecting inaccurate meters with high accuracy for practical usage to prevent unnecessary replacement and increase the service life span of smart meters.

arXiv:1907.11617공식 코드 · Papers with Code

Analog forecasting of extreme-causing weather patterns using deep learning

Numerical weather prediction (NWP) models require ever-growing computing time/resources, but still, have difficulties with predicting weather extremes. Here we introduce a data-driven framework that is based on analog forecasting (prediction using past similar patterns) and employs a novel deep learning pattern-recognition technique (capsule neural networks, CapsNets) and impact-based auto-labeling strategy. CapsNets are trained on mid-tropospheric large-scale circulation patterns (Z500) labeled $0-4$ depending on the existence and geographical region of surface temperature extremes over North America several days ahead. The trained networks predict the occurrence/region of cold or heat waves, only using Z500, with accuracies (recalls) of $69\%-45\%$ $(77\%-48\%)$ or $62\%-41\%$ $(73\%-47\%)$ $1-5$ days ahead. CapsNets outperform simpler techniques such as convolutional neural networks and logistic regression. Using both temperature and Z500, accuracies (recalls) with CapsNets increase to $\sim 80\%$ $(88\%)$, showing the promises of multi-modal data-driven frameworks for accurate/fast extreme weather predictions, which can augment NWP efforts in providing early warnings.

arXiv:1907.11040

DeepDrawing: A Deep Learning Approach to Graph Drawing

Node-link diagrams are widely used to facilitate network explorations. However, when using a graph drawing technique to visualize networks, users often need to tune different algorithm-specific parameters iteratively by comparing the corresponding drawing results in order to achieve a desired visual effect. This trial and error process is often tedious and time-consuming, especially for non-expert users. Inspired by the powerful data modelling and prediction capabilities of deep learning techniques, we explore the possibility of applying deep learning techniques to graph drawing. Specifically, we propose using a graph-LSTM-based approach to directly map network structures to graph drawings. Given a set of layout examples as the training dataset, we train the proposed graph-LSTM-based model to capture their layout characteristics. Then, the trained model is used to generate graph drawings in a similar style for new networks. We evaluated the proposed approach on two special types of layouts (i.e., grid layouts and star layouts) and two general types of layouts (i.e., ForceAtlas2 and PivotMDS) in both qualitative and quantitative ways. The results provide support for the effectiveness of our approach. We also conducted a time cost assessment on the drawings of small graphs with 20 to 50 nodes. We further report the lessons we learned and discuss the limitations and future work.

arXiv:1907.11210

HUGE2: a Highly Untangled Generative-model Engine for Edge-computing

As a type of prominent studies in deep learning, generative models have been widely investigated in research recently. Two research branches of the deep learning models, the Generative Networks (GANs, VAE) and the Semantic Segmentation, rely highly on the upsampling operations, especially the transposed convolution and the dilated convolution. However, these two types of convolutions are intrinsically different from standard convolution regarding the insertion of zeros in input feature maps or in kernels respectively. This distinct nature severely degrades the performance of the existing deep learning engine or frameworks, such as Darknet, Tensorflow, and PyTorch, which are mainly developed for the standard convolution. Another trend in deep learning realm is to deploy the model onto edge/ embedded devices, in which the memory resource is scarce. In this work, we propose a Highly Untangled Generative-model Engine for Edge-computing or HUGE2 for accelerating these two special convolutions on the edge-computing platform by decomposing the kernels and untangling these smaller convolutions by performing basic matrix multiplications. The methods we propose use much smaller memory footprint, hence much fewer memory accesses, and the data access patterns also dramatically increase the reusability of the data already fetched in caches, hence increasing the localities of caches. Our engine achieves a speedup of nearly 5x on embedded CPUs, and around 10x on embedded GPUs, and more than 50% reduction of memory access.