PyTorchKR Papers

최신 논문

총 8,922편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2202.09195

A Review on Methods and Applications in Multimodal Deep Learning

Deep Learning has implemented a wide range of applications and has become increasingly popular in recent years. The goal of multimodal deep learning (MMDL) is to create models that can process and link information using various modalities. Despite the extensive development made for unimodal learning, it still cannot cover all the aspects of human learning. Multimodal learning helps to understand and analyze better when various senses are engaged in the processing of information. This paper focuses on multiple types of modalities, i.e., image, video, text, audio, body gestures, facial expressions, and physiological signals. Detailed analysis of the baseline approaches and an in-depth study of recent advancements during the last five years (2017 to 2021) in multimodal deep learning applications has been provided. A fine-grained taxonomy of various multimodal deep learning methods is proposed, elaborating on different applications in more depth. Lastly, main issues are highlighted separately for each domain, along with their possible future research directions.

arXiv:2202.09046공식 코드 · Papers with Code

Microplankton life histories revealed by holographic microscopy and deep learning

The marine microbial food web plays a central role in the global carbon cycle. Our mechanistic understanding of the ocean, however, is biased towards its larger constituents, while rates and biomass fluxes in the microbial food web are mainly inferred from indirect measurements and ensemble averages. Yet, resolution at the level of the individual microplankton is required to advance our understanding of the oceanic food web. Here, we demonstrate that, by combining holographic microscopy with deep learning, we can follow microplanktons throughout their lifespan, continuously measuring their three dimensional position and dry mass. The deep learning algorithms circumvent the computationally intensive processing of holographic data and allow rapid measurements over extended time periods. This permits us to reliably estimate growth rates, both in terms of dry mass increase and cell divisions, as well as to measure trophic interactions between species such as predation events. The individual resolution provides information about selectivity, individual feeding rates and handling times for individual microplanktons. This method is particularly useful to explore the flux of carbon through micro-zooplankton, the most important and least known group of primary consumers in the global oceans. We exemplify this by detailed descriptions of micro-zooplankton feeding events, cell divisions, and long term monitoring of single cells from division to division.

arXiv:2202.08955

R2-D2: Repetitive Reprediction Deep Decipher for Semi-Supervised Deep Learning

Most recent semi-supervised deep learning (deep SSL) methods used a similar paradigm: use network predictions to update pseudo-labels and use pseudo-labels to update network parameters iteratively. However, they lack theoretical support and cannot explain why predictions are good candidates for pseudo-labels in the deep learning paradigm. In this paper, we propose a principled end-to-end framework named deep decipher (D2) for SSL. Within the D2 framework, we prove that pseudo-labels are related to network predictions by an exponential link function, which gives a theoretical support for using predictions as pseudo-labels. Furthermore, we demonstrate that updating pseudo-labels by network predictions will make them uncertain. To mitigate this problem, we propose a training strategy called repetitive reprediction (R2). Finally, the proposed R2-D2 method is tested on the large-scale ImageNet dataset and outperforms state-of-the-art methods by 5 percentage points.

arXiv:2202.08546

An overview of deep learning in medical imaging

Machine learning (ML) has seen enormous consideration during the most recent decade. This success started in 2012 when an ML model accomplished a remarkable triumph in the ImageNet Classification, the world's most famous competition for computer vision. This model was a kind of convolutional neural system (CNN) called deep learning (DL). Since then, researchers have started to participate efficiently in DL's fastest developing area of research. These days, DL systems are cutting-edge ML systems spanning a broad range of disciplines, from human language processing to video analysis, and commonly used in the scholarly world and enterprise sector. Recent advances can bring tremendous improvement to the medical field. Improved and innovative methods for data processing, image analysis and can significantly improve the diagnostic technologies and medicinal services gradually. A quick review of current developments with relevant problems in the field of DL used for medical imaging has been provided. The primary purposes of the review are four: (i) provide a brief prolog to DL by discussing different DL models, (ii) review of the DL usage for medical image analysis (classification, detection, segmentation, and registration), (iii) review seven main application fields of DL in medical imaging, (iv) give an initial stage to those keen on adding to the research area about DL in clinical imaging by providing links of some useful informative assets, such as freely available DL codes, public datasets Table 7, and medical imaging competition sources Table 8 and end our survey by outlining distinct continuous difficulties, lessons learned and future of DL in the field of medical science.

arXiv:2202.08679공식 코드 · Papers with Code

Where Is My Training Bottleneck? Hidden Trade-Offs in Deep Learning Preprocessing Pipelines

Preprocessing pipelines in deep learning aim to provide sufficient data throughput to keep the training processes busy. Maximizing resource utilization is becoming more challenging as the throughput of training processes increases with hardware innovations (e.g., faster GPUs, TPUs, and inter-connects) and advanced parallelization techniques that yield better scalability. At the same time, the amount of training data needed in order to train increasingly complex models is growing. As a consequence of this development, data preprocessing and provisioning are becoming a severe bottleneck in end-to-end deep learning pipelines. In this paper, we provide an in-depth analysis of data preprocessing pipelines from four different machine learning domains. We introduce a new perspective on efficiently preparing datasets for end-to-end deep learning pipelines and extract individual trade-offs to optimize throughput, preprocessing time, and storage consumption. Additionally, we provide an open-source profiling library that can automatically decide on a suitable preprocessing strategy to maximize throughput. By applying our generated insights to real-world use-cases, we obtain an increased throughput of 3x to 13x compared to an untuned system while keeping the pipeline functionally identical. These findings show the enormous potential of data pipeline tuning.

arXiv:2202.08436

PENCIL: Deep Learning with Noisy Labels

Deep learning has achieved excellent performance in various computer vision tasks, but requires a lot of training examples with clean labels. It is easy to collect a dataset with noisy labels, but such noise makes networks overfit seriously and accuracies drop dramatically. To address this problem, we propose an end-to-end framework called PENCIL, which can update both network parameters and label estimations as label distributions. PENCIL is independent of the backbone network structure and does not need an auxiliary clean dataset or prior information about noise, thus it is more general and robust than existing methods and is easy to apply. PENCIL can even be used repeatedly to obtain better performance. PENCIL outperforms previous state-of-the-art methods by large margins on both synthetic and real-world datasets with different noise types and noise rates. And PENCIL is also effective in multi-label classification tasks through adding a simple attention structure on backbone networks. Experiments show that PENCIL is robust on clean datasets, too.

arXiv:2202.08341공식 코드 · Papers with Code

Anomalib: A Deep Learning Library for Anomaly Detection

This paper introduces anomalib, a novel library for unsupervised anomaly detection and localization. With reproducibility and modularity in mind, this open-source library provides algorithms from the literature and a set of tools to design custom anomaly detection algorithms via a plug-and-play approach. Anomalib comprises state-of-the-art anomaly detection algorithms that achieve top performance on the benchmarks and that can be used off-the-shelf. In addition, the library provides components to design custom algorithms that could be tailored towards specific needs. Additional tools, including experiment trackers, visualizers, and hyper-parameter optimizers, make it simple to design and implement anomaly detection models. The library also supports OpenVINO model optimization and quantization for real-time deployment. Overall, anomalib is an extensive library for the design, implementation, and deployment of unsupervised anomaly detection models from data to the edge.

arXiv:2202.08384

Limitations of Neural Collapse for Understanding Generalization in Deep Learning

The recent work of Papyan, Han, & Donoho (2020) presented an intriguing "Neural Collapse" phenomenon, showing a structural property of interpolating classifiers in the late stage of training. This opened a rich area of exploration studying this phenomenon. Our motivation is to study the upper limits of this research program: How far will understanding Neural Collapse take us in understanding deep learning? First, we investigate its role in generalization. We refine the Neural Collapse conjecture into two separate conjectures: collapse on the train set (an optimization property) and collapse on the test distribution (a generalization property). We find that while Neural Collapse often occurs on the train set, it does not occur on the test set. We thus conclude that Neural Collapse is primarily an optimization phenomenon, with as-yet-unclear connections to generalization. Second, we investigate the role of Neural Collapse in feature learning. We show simple, realistic experiments where training longer leads to worse last-layer features, as measured by transfer-performance on a downstream task. This suggests that neural collapse is not always desirable for representation learning, as previously claimed. Finally, we give preliminary evidence of a "cascading collapse" phenomenon, wherein some form of Neural Collapse occurs not only for the last layer, but in earlier layers as well. We hope our work encourages the community to continue the rich line of Neural Collapse research, while also considering its inherent limitations.

arXiv:2202.08371

The Quarks of Attention

Attention plays a fundamental role in both natural and artificial intelligence systems. In deep learning, attention-based neural architectures, such as transformer architectures, are widely used to tackle problems in natural language processing and beyond. Here we investigate the fundamental building blocks of attention and their computational properties. Within the standard model of deep learning, we classify all possible fundamental building blocks of attention in terms of their source, target, and computational mechanism. We identify and study three most important mechanisms: additive activation attention, multiplicative output attention (output gating), and multiplicative synaptic attention (synaptic gating). The gating mechanisms correspond to multiplicative extensions of the standard model and are used across all current attention-based deep learning architectures. We study their functional properties and estimate the capacity of several attentional building blocks in the case of linear and polynomial threshold gates. Surprisingly, additive activation attention plays a central role in the proofs of the lower bounds. Attention mechanisms reduce the depth of certain basic circuits and leverage the power of quadratic activations without incurring their full cost.

arXiv:2011.05144

Image transmission through a flexible multimode fiber by deep learning

When multimode optical fibers are perturbed, the data that is transmitted through them is scrambled. This presents a major difficulty for many possible applications, such as multimode fiber-based telecommunication and endoscopy. To overcome this challenge, a deep learning approach that generalizes over mechanical perturbations is presented. Using this approach, successful reconstruction of the input images from intensity-only measurements of speckle patterns at the output of a 1.5 meter-long randomly perturbed multimode fiber is demonstrated. The model's success is explained by hidden correlations in the speckle of random fiber conformations.

arXiv:2202.07998

DeepTx: Deep Learning Beamforming with Channel Prediction

Machine learning algorithms have recently been considered for many tasks in the field of wireless communications. Previously, we have proposed the use of a deep fully convolutional neural network (CNN) for receiver processing and shown it to provide considerable performance gains. In this study, we focus on machine learning algorithms for the transmitter. In particular, we consider beamforming and propose a CNN which, for a given uplink channel estimate as input, outputs downlink channel information to be used for beamforming. The CNN is trained in a supervised manner considering both uplink and downlink transmissions with a loss function that is based on UE receiver performance. The main task of the neural network is to predict the channel evolution between uplink and downlink slots, but it can also learn to handle inefficiencies and errors in the whole chain, including the actual beamforming phase. The provided numerical experiments demonstrate the improved beamforming performance.

arXiv:2202.07896

Aryl: An Elastic Cluster Scheduler for Deep Learning

Companies build separate training and inference GPU clusters for deep learning, and use separate schedulers to manage them. This leads to problems for both training and inference: inference clusters have low GPU utilization when the traffic load is low; training jobs often experience long queueing time due to lack of resources. We introduce Aryl, a new cluster scheduler to address these problems. Aryl introduces capacity loaning to loan idle inference GPU servers for training jobs. It further exploits elastic scaling that scales a training job's GPU allocation to better utilize loaned resources. Capacity loaning and elastic scaling create new challenges to cluster management. When the loaned servers need to be returned, we need to minimize the number of job preemptions; when more GPUs become available, we need to allocate them to elastic jobs and minimize the job completion time (JCT). Aryl addresses these combinatorial problems using principled heuristics. It introduces the notion of server preemption cost which it greedily reduces during server reclaiming. It further relies on the JCT reduction value defined for each additional worker for an elastic job to solve the scheduling problem as a multiple-choice knapsack problem. Prototype implementation on a 64-GPU testbed and large-scale simulation with 15-day traces of over 50,000 production jobs show that Aryl brings 1.53x and 1.50x reductions in average queuing time and JCT, and improves cluster usage by up to 26.9% over the cluster scheduler without capacity loaning or elastic scaling.

arXiv:2202.07796

Low Latency Real-Time Seizure Detection Using Transfer Deep Learning

Scalp electroencephalogram (EEG) signals inherently have a low signal-to-noise ratio due to the way the signal is electrically transduced. Temporal and spatial information must be exploited to achieve accurate detection of seizure events. Most popular approaches to seizure detection using deep learning do not jointly model this information or require multiple passes over the signal, which makes the systems inherently non-causal. In this paper, we exploit both simultaneously by converting the multichannel signal to a grayscale image and using transfer learning to achieve high performance. The proposed system is trained end-to-end with only very simple pre- and postprocessing operations which are computationally lightweight and have low latency, making them conducive to clinical applications that require real-time processing. We have achieved a performance of 42.05% sensitivity with 5.78 false alarms per 24 hours on the development dataset of v1.5.2 of the Temple University Hospital Seizure Detection Corpus. On a single-core CPU operating at 1.7 GHz, the system runs faster than real-time (0.58 xRT), uses 16 Gbytes of memory, and has a latency of 300 msec.

arXiv:2202.07848

Singularity: Planet-Scale, Preemptive and Elastic Scheduling of AI Workloads

Lowering costs by driving high utilization across deep learning workloads is a crucial lever for cloud providers. We present Singularity, Microsoft's globally distributed scheduling service for highly-efficient and reliable execution of deep learning training and inference workloads. At the heart of Singularity is a novel, workload-aware scheduler that can transparently preempt and elastically scale deep learning workloads to drive high utilization without impacting their correctness or performance, across a global fleet of AI accelerators (e.g., GPUs, FPGAs). All jobs in Singularity are preemptable, migratable, and dynamically resizable (elastic) by default: a live job can be dynamically and transparently (a) preempted and migrated to a different set of nodes, cluster, data center or a region and resumed exactly from the point where the execution was preempted, and (b) resized (i.e., elastically scaled-up/down) on a varying set of accelerators of a given type. Our mechanisms are transparent in that they do not require the user to make any changes to their code or require using any custom libraries that may limit flexibility. Additionally, our approach significantly improves the reliability of deep learning workloads. We show that the resulting efficiency and reliability gains with Singularity are achieved with negligible impact on the steady-state performance. Finally, our design approach is agnostic of DNN architectures and handles a variety of parallelism strategies (e.g., data/pipeline/model parallelism).

arXiv:2202.07118공식 코드 · Papers with Code

Multi-task UNet: Jointly Boosting Saliency Prediction and Disease Classification on Chest X-ray Images

Human visual attention has recently shown its distinct capability in boosting machine learning models. However, studies that aim to facilitate medical tasks with human visual attention are still scarce. To support the use of visual attention, this paper describes a novel deep learning model for visual saliency prediction on chest X-ray (CXR) images. To cope with data deficiency, we exploit the multi-task learning method and tackles disease classification on CXR simultaneously. For a more robust training process, we propose a further optimized multi-task learning scheme to better handle model overfitting. Experiments show our proposed deep learning model with our new learning scheme can outperform existing methods dedicated either for saliency prediction or image classification. The code used in this paper is available at https://github.com/hz-zhu/MT-UNet.

arXiv:2202.07201

Holistic Adversarial Robustness of Deep Learning Models

Adversarial robustness studies the worst-case performance of a machine learning model to ensure safety and reliability. With the proliferation of deep-learning-based technology, the potential risks associated with model development and deployment can be amplified and become dreadful vulnerabilities. This paper provides a comprehensive overview of research topics and foundational principles of research methods for adversarial robustness of deep learning models, including attacks, defenses, verification, and novel applications.

arXiv:2202.07183

A Survey of Neural Trojan Attacks and Defenses in Deep Learning

Artificial Intelligence (AI) relies heavily on deep learning - a technology that is becoming increasingly popular in real-life applications of AI, even in the safety-critical and high-risk domains. However, it is recently discovered that deep learning can be manipulated by embedding Trojans inside it. Unfortunately, pragmatic solutions to circumvent the computational requirements of deep learning, e.g. outsourcing model training or data annotation to third parties, further add to model susceptibility to the Trojan attacks. Due to the key importance of the topic in deep learning, recent literature has seen many contributions in this direction. We conduct a comprehensive review of the techniques that devise Trojan attacks for deep learning and explore their defenses. Our informative survey systematically organizes the recent literature and discusses the key concepts of the methods while assuming minimal knowledge of the domain on the readers part. It provides a comprehensible gateway to the broader community to understand the recent developments in Neural Trojans.

arXiv:2202.06764

Low-latency Monaural Speech Enhancement with Deep Filter-bank Equalizer

It is highly desirable that speech enhancement algorithms can achieve good performance while keeping low latency for many applications, such as digital hearing aids, acoustically transparent hearing devices, and public address systems. To improve the performance of traditional low-latency speech enhancement algorithms, a deep filter-bank equalizer (FBE) framework was proposed, which integrated a deep learning-based subband noise reduction network with a deep learning-based shortened digital filter mapping network. In the first network, a deep learning model was trained with a controllable small frame shift to satisfy the low-latency demand, i.e., $\le$ 4 ms, so as to obtain (complex) subband gains, which could be regarded as an adaptive digital filter in each frame. In the second network, to reduce the latency, this adaptive digital filter was implicitly shortened by a deep learning-based framework, and was then applied to noisy speech to reconstruct the enhanced speech without the overlap-add method. Experimental results on the WSJ0-SI84 corpus indicated that the proposed deep FBE with only 4-ms latency achieved much better performance than traditional low-latency speech enhancement algorithms in terms of the indices such as PESQ, STOI, and the amount of noise reduction.

arXiv:2202.05972

Low-light Image Enhancement by Retinex Based Algorithm Unrolling and Adjustment

Motivated by their recent advances, deep learning techniques have been widely applied to low-light image enhancement (LIE) problem. Among which, Retinex theory based ones, mostly following a decomposition-adjustment pipeline, have taken an important place due to its physical interpretation and promising performance. However, current investigations on Retinex based deep learning are still not sufficient, ignoring many useful experiences from traditional methods. Besides, the adjustment step is either performed with simple image processing techniques, or by complicated networks, both of which are unsatisfactory in practice. To address these issues, we propose a new deep learning framework for the LIE problem. The proposed framework contains a decomposition network inspired by algorithm unrolling, and adjustment networks considering both global brightness and local brightness sensitivity. By virtue of algorithm unrolling, both implicit priors learned from data and explicit priors borrowed from traditional methods can be embedded in the network, facilitate to better decomposition. Meanwhile, the consideration of global and local brightness can guide designing simple yet effective network modules for adjustment. Besides, to avoid manually parameter tuning, we also propose a self-supervised fine-tuning strategy, which can always guarantee a promising performance. Experiments on a series of typical LIE datasets demonstrated the effectiveness of the proposed method, both quantitatively and visually, as compared with existing methods.

arXiv:2202.06431

AI can evolve without labels: self-evolving vision transformer for chest X-ray diagnosis through knowledge distillation

Although deep learning-based computer-aided diagnosis systems have recently achieved expert-level performance, developing a robust deep learning model requires large, high-quality data with manual annotation, which is expensive to obtain. This situation poses the problem that the chest x-rays collected annually in hospitals cannot be used due to the lack of manual labeling by experts, especially in deprived areas. To address this, here we present a novel deep learning framework that uses knowledge distillation through self-supervised learning and self-training, which shows that the performance of the original model trained with a small number of labels can be gradually improved with more unlabeled data. Experimental results show that the proposed framework maintains impressive robustness against a real-world environment and has general applicability to several diagnostic tasks such as tuberculosis, pneumothorax, and COVID-19. Notably, we demonstrated that our model performs even better than those trained with the same amount of labeled data. The proposed framework has a great potential for medical imaging, where plenty of data is accumulated every year, but ground truth annotations are expensive to obtain.