PyTorchKR Papers

최신 논문

총 10,163편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2007.00151공식 코드 · Papers with Code

Early-Learning Regularization Prevents Memorization of Noisy Labels

We propose a novel framework to perform classification via deep learning in the presence of noisy annotations. When trained on noisy labels, deep neural networks have been observed to first fit the training data with clean labels during an "early learning" phase, before eventually memorizing the examples with false labels. We prove that early learning and memorization are fundamental phenomena in high-dimensional classification tasks, even in simple linear models, and give a theoretical explanation in this setting. Motivated by these findings, we develop a new technique for noisy classification tasks, which exploits the progress of the early learning phase. In contrast with existing approaches, which use the model output during early learning to detect the examples with clean labels, and either ignore or attempt to correct the false labels, we take a different route and instead capitalize on early learning via regularization. There are two key elements to our approach. First, we leverage semi-supervised learning techniques to produce target probabilities based on the model outputs. Second, we design a regularization term that steers the model towards these targets, implicitly preventing memorization of the false labels. The resulting framework is shown to provide robustness to noisy annotations on several standard benchmarks and real-world datasets, where it achieves results comparable to the state of the art.

arXiv:2007.00051

Extracurricular Learning: Knowledge Transfer Beyond Empirical Distribution

Knowledge distillation has been used to transfer knowledge learned by a sophisticated model (teacher) to a simpler model (student). This technique is widely used to compress model complexity. However, in most applications the compressed student model suffers from an accuracy gap with its teacher. We propose extracurricular learning, a novel knowledge distillation method, that bridges this gap by (1) modeling student and teacher output distributions; (2) sampling examples from an approximation to the underlying data distribution; and (3) matching student and teacher output distributions over this extended set including uncertain samples. We conduct rigorous evaluations on regression and classification tasks and show that compared to the standard knowledge distillation, extracurricular learning reduces the gap by 46% to 68%. This leads to major accuracy improvements compared to the empirical risk minimization-based training for various recent neural network architectures: 16% regression error reduction on the MPIIGaze dataset, +3.4% to +9.1% improvement in top-1 classification accuracy on the CIFAR100 dataset, and +2.9% top-1 improvement on the ImageNet dataset.

arXiv:2007.00644공식 코드 · Papers with Code

Measuring Robustness to Natural Distribution Shifts in Image Classification

We study how robust current ImageNet models are to distribution shifts arising from natural variations in datasets. Most research on robustness focuses on synthetic image perturbations (noise, simulated weather artifacts, adversarial examples, etc.), which leaves open how robustness on synthetic distribution shift relates to distribution shift arising in real data. Informed by an evaluation of 204 ImageNet models in 213 different test conditions, we find that there is often little to no transfer of robustness from current synthetic to natural distribution shift. Moreover, most current techniques provide no robustness to the natural distribution shifts in our testbed. The main exception is training on larger and more diverse datasets, which in multiple cases increases robustness, but is still far from closing the performance gaps. Our results indicate that distribution shifts arising in real data are currently an open research problem. We provide our testbed and data as a resource for future work at https://modestyachts.github.io/imagenet-testbed/ .

arXiv:2007.00584

HACT-Net: A Hierarchical Cell-to-Tissue Graph Neural Network for Histopathological Image Classification

Cancer diagnosis, prognosis, and therapeutic response prediction are heavily influenced by the relationship between the histopathological structures and the function of the tissue. Recent approaches acknowledging the structure-function relationship, have linked the structural and spatial patterns of cell organization in tissue via cell-graphs to tumor grades. Though cell organization is imperative, it is insufficient to entirely represent the histopathological structure. We propose a novel hierarchical cell-to-tissue-graph (HACT) representation to improve the structural depiction of the tissue. It consists of a low-level cell-graph, capturing cell morphology and interactions, a high-level tissue-graph, capturing morphology and spatial distribution of tissue parts, and cells-to-tissue hierarchies, encoding the relative spatial distribution of the cells with respect to the tissue distribution. Further, a hierarchical graph neural network (HACT-Net) is proposed to efficiently map the HACT representations to histopathological breast cancer subtypes. We assess the methodology on a large set of annotated tissue regions of interest from H\&E stained breast carcinoma whole-slides. Upon evaluation, the proposed method outperformed recent convolutional neural network and graph neural network approaches for breast cancer multi-class subtyping. The proposed entity-based topological analysis is more inline with the pathological diagnostic procedure of the tissue. It provides more command over the tissue modelling, therefore encourages the further inclusion of pathological priors into task-specific tissue representation.

arXiv:2007.00334

Estimation with Uncertainty via Conditional Generative Adversarial Networks

Conventional predictive Artificial Neural Networks (ANNs) commonly employ deterministic weight matrices; therefore, their prediction is a point estimate. Such a deterministic nature in ANNs causes the limitations of using ANNs for medical diagnosis, law problems, and portfolio management, in which discovering not only the prediction but also the uncertainty of the prediction is essentially required. To address such a problem, we propose a predictive probabilistic neural network model, which corresponds to a different manner of using the generator in conditional Generative Adversarial Network (cGAN) that has been routinely used for conditional sample generation. By reversing the input and output of ordinary cGAN, the model can be successfully used as a predictive model; besides, the model is robust against noises since adversarial training is employed. In addition, to measure the uncertainty of predictions, we introduce the entropy and relative entropy for regression problems and classification problems, respectively. The proposed framework is applied to stock market data and an image classification task. As a result, the proposed framework shows superior estimation performance, especially on noisy data; moreover, it is demonstrated that the proposed framework can properly estimate the uncertainty of predictions.

arXiv:2007.00302

Robustifying the Deployment of tinyML Models for Autonomous mini-vehicles

Standard-size autonomous navigation vehicles have rapidly improved thanks to the breakthroughs of deep learning. However, scaling autonomous driving to low-power systems deployed on dynamic environments poses several challenges that prevent their adoption. To address them, we propose a closed-loop learning flow for autonomous driving mini-vehicles that includes the target environment in-the-loop. We leverage a family of compact and high-throughput tinyCNNs to control the mini-vehicle, which learn in the target environment by imitating a computer vision algorithm, i.e., the expert. Thus, the tinyCNNs, having only access to an on-board fast-rate linear camera, gain robustness to lighting conditions and improve over time. Further, we leverage GAP8, a parallel ultra-low-power RISC-V SoC, to meet the inference requirements. When running the family of CNNs, our GAP8's solution outperforms any other implementation on the STM32L4 and NXP k64f (Cortex-M4), reducing the latency by over 13x and the energy consummation by 92%.

arXiv:2006.16627공식 코드 · Papers with Code

Training highly effective connectivities within neural networks with randomly initialized, fixed weights

We present some novel, straightforward methods for training the connection graph of a randomly initialized neural network without training the weights. These methods do not use hyperparameters defining cutoff thresholds and therefore remove the need for iteratively searching optimal values of such hyperparameters. We can achieve similar or higher performances than in the case of training all weights, with a similar computational cost as for standard training techniques. Besides switching connections on and off, we introduce a novel way of training a network by flipping the signs of the weights. If we try to minimize the number of changed connections, by changing less than 10% of the total it is already possible to reach more than 90% of the accuracy achieved by standard training. We obtain good results even with weights of constant magnitude or even when weights are drawn from highly asymmetric distributions. These results shed light on the over-parameterization of neural networks and on how they may be reduced to their effective size.

arXiv:2006.16981공식 코드 · Papers with Code

Learning to Combine Top-Down and Bottom-Up Signals in Recurrent Neural Networks with Attention over Modules

Robust perception relies on both bottom-up and top-down signals. Bottom-up signals consist of what's directly observed through sensation. Top-down signals consist of beliefs and expectations based on past experience and short-term memory, such as how the phrase `peanut butter and~...' will be completed. The optimal combination of bottom-up and top-down information remains an open question, but the manner of combination must be dynamic and both context and task dependent. To effectively utilize the wealth of potential top-down information available, and to prevent the cacophony of intermixed signals in a bidirectional architecture, mechanisms are needed to restrict information flow. We explore deep recurrent neural net architectures in which bottom-up and top-down signals are dynamically combined using attention. Modularity of the architecture further restricts the sharing and communication of information. Together, attention and modularity direct information flow, which leads to reliable performance improvements in perceptual and language tasks, and in particular improves robustness to distractions and noisy data. We demonstrate on a variety of benchmarks in language modeling, sequential image classification, video prediction and reinforcement learning that the \emph{bidirectional} information flow can improve results over strong baselines.

arXiv:2006.16971공식 코드 · Papers with Code

Improving robustness against common corruptions by covariate shift adaptation

Today's state-of-the-art machine vision models are vulnerable to image corruptions like blurring or compression artefacts, limiting their performance in many real-world applications. We here argue that popular benchmarks to measure model robustness against common corruptions (like ImageNet-C) underestimate model robustness in many (but not all) application scenarios. The key insight is that in many scenarios, multiple unlabeled examples of the corruptions are available and can be used for unsupervised online adaptation. Replacing the activation statistics estimated by batch normalization on the training set with the statistics of the corrupted images consistently improves the robustness across 25 different popular computer vision models. Using the corrected statistics, ResNet-50 reaches 62.2% mCE on ImageNet-C compared to 76.7% without adaptation. With the more robust DeepAugment+AugMix model, we improve the state of the art achieved by a ResNet50 model up to date from 53.6% mCE to 45.4% mCE. Even adapting to a single sample improves robustness for the ResNet-50 and AugMix models, and 32 samples are sufficient to improve the current state of the art for a ResNet-50 architecture. We argue that results with adapted statistics should be included whenever reporting scores in corruption benchmarks and other out-of-distribution generalization settings.

arXiv:2006.16917

Ontology-guided Semantic Composition for Zero-Shot Learning

Zero-shot learning (ZSL) is a popular research problem that aims at predicting for those classes that have never appeared in the training stage by utilizing the inter-class relationship with some side information. In this study, we propose to model the compositional and expressive semantics of class labels by an OWL (Web Ontology Language) ontology, and further develop a new ZSL framework with ontology embedding. The effectiveness has been verified by some primary experiments on animal image classification and visual question answering.

arXiv:2006.16537

Theory-Inspired Path-Regularized Differential Network Architecture Search

Despite its high search efficiency, differential architecture search (DARTS) often selects network architectures with dominated skip connections which lead to performance degradation. However, theoretical understandings on this issue remain absent, hindering the development of more advanced methods in a principled way. In this work, we solve this problem by theoretically analyzing the effects of various types of operations, e.g. convolution, skip connection and zero operation, to the network optimization. We prove that the architectures with more skip connections can converge faster than the other candidates, and thus are selected by DARTS. This result, for the first time, theoretically and explicitly reveals the impact of skip connections to fast network optimization and its competitive advantage over other types of operations in DARTS. Then we propose a theory-inspired path-regularized DARTS that consists of two key modules: (i) a differential group-structured sparse binary gate introduced for each operation to avoid unfair competition among operations, and (ii) a path-depth-wise regularization used to incite search exploration for deep architectures that often converge slower than shallow ones as shown in our theory and are not well explored during the search. Experimental results on image classification tasks validate its advantages.

arXiv:2006.15538

Compositional Convolutional Neural Networks: A Robust and Interpretable Model for Object Recognition under Occlusion

Computer vision systems in real-world applications need to be robust to partial occlusion while also being explainable. In this work, we show that black-box deep convolutional neural networks (DCNNs) have only limited robustness to partial occlusion. We overcome these limitations by unifying DCNNs with part-based models into Compositional Convolutional Neural Networks (CompositionalNets) - an interpretable deep architecture with innate robustness to partial occlusion. Specifically, we propose to replace the fully connected classification head of DCNNs with a differentiable compositional model that can be trained end-to-end. The structure of the compositional model enables CompositionalNets to decompose images into objects and context, as well as to further decompose object representations in terms of individual parts and the objects' pose. The generative nature of our compositional model enables it to localize occluders and to recognize objects based on their non-occluded parts. We conduct extensive experiments in terms of image classification and object detection on images of artificially occluded objects from the PASCAL3D+ and ImageNet dataset, and real images of partially occluded vehicles from the MS-COCO dataset. Our experiments show that CompositionalNets made from several popular DCNN backbones (VGG-16, ResNet50, ResNext) improve by a large margin over their non-compositional counterparts at classifying and detecting partially occluded objects. Furthermore, they can localize occluders accurately despite being trained with class-level supervision only. Finally, we demonstrate that CompositionalNets provide human interpretable predictions as their individual components can be understood as detecting parts and estimating an objects' viewpoint.

arXiv:2006.15476

Frequency learning for image classification

Machine learning applied to computer vision and signal processing is achieving results comparable to the human brain on specific tasks due to the great improvements brought by the deep neural networks (DNN). The majority of state-of-the-art architectures nowadays are DNN related, but only a few explore the frequency domain to extract useful information and improve the results, like in the image processing field. In this context, this paper presents a new approach for exploring the Fourier transform of the input images, which is composed of trainable frequency filters that boost discriminative components in the spectrum. Additionally, we propose a slicing procedure to allow the network to learn both global and local features from the frequency-domain representations of the image blocks. The proposed method proved to be competitive with respect to well-known DNN architectures in the selected experiments, with the advantage of being a simpler and lightweight model. This work also raises the discussion on how the state-of-the-art DNNs architectures can exploit not only spatial features, but also the frequency, in order to improve its performance when solving real world problems.

arXiv:2006.15366공식 코드 · Papers with Code

ReMarNet: Conjoint Relation and Margin Learning for Small-Sample Image Classification

Despite achieving state-of-the-art performance, deep learning methods generally require a large amount of labeled data during training and may suffer from overfitting when the sample size is small. To ensure good generalizability of deep networks under small sample sizes, learning discriminative features is crucial. To this end, several loss functions have been proposed to encourage large intra-class compactness and inter-class separability. In this paper, we propose to enhance the discriminative power of features from a new perspective by introducing a novel neural network termed Relation-and-Margin learning Network (ReMarNet). Our method assembles two networks of different backbones so as to learn the features that can perform excellently in both of the aforementioned two classification mechanisms. Specifically, a relation network is used to learn the features that can support classification based on the similarity between a sample and a class prototype; at the meantime, a fully connected network with the cross entropy loss is used for classification via the decision boundary. Experiments on four image datasets demonstrate that our approach is effective in learning discriminative features from a small set of labeled samples and achieves competitive performance against state-of-the-art methods. Codes are available at https://github.com/liyunyu08/ReMarNet.

arXiv:2006.15357

An Evoked Potential-Guided Deep Learning Brain Representation For Visual Classification

The new perspective in visual classification aims to decode the feature representation of visual objects from human brain activities. Recording electroencephalogram (EEG) from the brain cortex has been seen as a prevalent approach to understand the cognition process of an image classification task. In this study, we proposed a deep learning framework guided by the visual evoked potentials, called the Event-Related Potential (ERP)-Long short-term memory (LSTM) framework, extracted by EEG signals for visual classification. In specific, we first extracted the ERP sequences from multiple EEG channels to response image stimuli-related information. Then, we trained an LSTM network to learn the feature representation space of visual objects for classification. In the experiment, 10 subjects were recorded by over 50,000 EEG trials from an image dataset with 6 categories, including a total of 72 exemplars. Our results showed that our proposed ERP-LSTM framework could achieve classification accuracies of cross-subject of 66.81% and 27.08% for categories (6 classes) and exemplars (72 classes), respectively. Our results outperformed that of using the existing visual classification frameworks, by improving classification accuracies in the range of 12.62% - 53.99%. Our findings suggested that decoding visual evoked potentials from EEG signals is an effective strategy to learn discriminative brain representations for visual classification.

arXiv:2006.15351

Unsupervised Deep Representation Learning and Few-Shot Classification of PolSAR Images

Deep learning and convolutional neural networks (CNNs) have made progress in polarimetric synthetic aperture radar (PolSAR) image classification over the past few years. However, a crucial issue has not been addressed, i.e., the requirement of CNNs for abundant labeled samples versus the insufficient human annotations of PolSAR images. It is well-known that following the supervised learning paradigm may lead to the overfitting of training data, and the lack of supervision information of PolSAR images undoubtedly aggravates this problem, which greatly affects the generalization performance of CNN-based classifiers in large-scale applications. To handle this problem, in this paper, learning transferrable representations from unlabeled PolSAR data through convolutional architectures is explored for the first time. Specifically, a PolSAR-tailored contrastive learning network (PCLNet) is proposed for unsupervised deep PolSAR representation learning and few-shot classification. Different from the utilization of optical processing methods, a diversity stimulation mechanism is constructed to narrow the application gap between optics and PolSAR. Beyond the conventional supervised methods, PCLNet develops an unsupervised pre-training phase based on the proxy objective of instance discrimination to learn useful representations from unlabeled PolSAR data. The acquired representations are transferred to the downstream task, i.e., few-shot PolSAR classification. Experiments on two widely-used PolSAR benchmark datasets confirm the validity of PCLNet. Besides, this work may enlighten how to efficiently utilize the massive unlabeled PolSAR data to alleviate the greedy demands of CNN-based methods for human annotations.

arXiv:2006.15983

Explainable 3D Convolutional Neural Networks by Learning Temporal Transformations

In this paper we introduce the temporally factorized 3D convolution (3TConv) as an interpretable alternative to the regular 3D convolution (3DConv). In a 3TConv the 3D convolutional filter is obtained by learning a 2D filter and a set of temporal transformation parameters, resulting in a sparse filter where the 2D slices are sequentially dependent on each other in the temporal dimension. We demonstrate that 3TConv learns temporal transformations that afford a direct interpretation. The temporal parameters can be used in combination with various existing 2D visualization methods. We also show that insight about what the model learns can be achieved by analyzing the transformation parameter statistics on a layer and model level. Finally, we implicitly demonstrate that, in popular ConvNets, the 2DConv can be replaced with a 3TConv and that the weights can be transferred to yield pretrained 3TConvs. pretrained 3TConvnets leverage more than a decade of work on traditional 2DConvNets by being able to make use of features that have been proven to deliver excellent results on image classification benchmarks.

arXiv:2006.15892공식 코드 · Papers with Code

Matrix Shuffle-Exchange Networks for Hard 2D Tasks

Convolutional neural networks have become the main tools for processing two-dimensional data. They work well for images, yet convolutions have a limited receptive field that prevents its applications to more complex 2D tasks. We propose a new neural model, called Matrix Shuffle-Exchange network, that can efficiently exploit long-range dependencies in 2D data and has comparable speed to a convolutional neural network. It is derived from Neural Shuffle-Exchange network and has $\mathcal{O}( \log{n})$ layers and $\mathcal{O}( n^2 \log{n})$ total time and space complexity for processing a $n \times n$ data matrix. We show that the Matrix Shuffle-Exchange network is well-suited for algorithmic and logical reasoning tasks on matrices and dense graphs, exceeding convolutional and graph neural network baselines. Its distinct advantage is the capability of retaining full long-range dependency modelling when generalizing to larger instances - much larger than could be processed with models equipped with a dense attention mechanism.

arXiv:2006.15799

Conditional Classification: A Solution for Computational Energy Reduction

Deep convolutional neural networks have shown high efficiency in computer visions and other applications. However, with the increase in the depth of the networks, the computational complexity is growing exponentially. In this paper, we propose a novel solution to reduce the computational complexity of convolutional neural network models used for many class image classification. Our proposed technique breaks the classification task into two steps: 1) coarse-grain classification, in which the input samples are classified among a set of hyper-classes, 2) fine-grain classification, in which the final labels are predicted among those hyper-classes detected at the first step. We illustrate that our proposed classifier can reach the level of accuracy reported by the best in class classification models with less computational complexity (Flop Count) by only activating parts of the model that are needed for the image classification.

arXiv:2006.15771

Active Ensemble Deep Learning for Polarimetric Synthetic Aperture Radar Image Classification

Although deep learning has achieved great success in image classification tasks, its performance is subject to the quantity and quality of training samples. For classification of polarimetric synthetic aperture radar (PolSAR) images, it is nearly impossible to annotate the images from visual interpretation. Therefore, it is urgent for remote sensing scientists to develop new techniques for PolSAR image classification under the condition of very few training samples. In this letter, we take the advantage of active learning and propose active ensemble deep learning (AEDL) for PolSAR image classification. We first show that only 35\% of the predicted labels of a deep learning model's snapshots near its convergence were exactly the same. The disagreement between snapshots is non-negligible. From the perspective of multiview learning, the snapshots together serve as a good committee to evaluate the importance of unlabeled instances. Using the snapshots committee to give out the informativeness of unlabeled data, the proposed AEDL achieved better performance on two real PolSAR images compared with standard active learning strategies. It achieved the same classification accuracy with only 86% and 55% of the training samples compared with breaking ties active learning and random selection for the Flevoland dataset.