PyTorchKR Papers

최신 논문

총 10,163편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2004.04730공식 코드 · Papers with Code

X3D: Expanding Architectures for Efficient Video Recognition

This paper presents X3D, a family of efficient video networks that progressively expand a tiny 2D image classification architecture along multiple network axes, in space, time, width and depth. Inspired by feature selection methods in machine learning, a simple stepwise network expansion approach is employed that expands a single axis in each step, such that good accuracy to complexity trade-off is achieved. To expand X3D to a specific target complexity, we perform progressive forward expansion followed by backward contraction. X3D achieves state-of-the-art performance while requiring 4.8x and 5.5x fewer multiply-adds and parameters for similar accuracy as previous work. Our most surprising finding is that networks with high spatiotemporal resolution can perform well, while being extremely light in terms of network width and parameters. We report competitive accuracy at unprecedented efficiency on video classification and detection benchmarks. Code will be available at: https://github.com/facebookresearch/SlowFast

arXiv:2004.04729

Dithered backprop: A sparse and quantized backpropagation algorithm for more efficient deep neural network training

Deep Neural Networks are successful but highly computationally expensive learning systems. One of the main sources of time and energy drains is the well known backpropagation (backprop) algorithm, which roughly accounts for 2/3 of the computational complexity of training. In this work we propose a method for reducing the computational cost of backprop, which we named dithered backprop. It consists in applying a stochastic quantization scheme to intermediate results of the method. The particular quantisation scheme, called non-subtractive dither (NSD), induces sparsity which can be exploited by computing efficient sparse matrix multiplications. Experiments on popular image classification tasks show that it induces 92% sparsity on average across a wide set of models at no or negligible accuracy drop in comparison to state-of-the-art approaches, thus significantly reducing the computational complexity of the backward pass. Moreover, we show that our method is fully compatible to state-of-the-art training methods that reduce the bit-precision of training down to 8-bits, as such being able to further reduce the computational requirements. Finally we discuss and show potential benefits of applying dithered backprop in a distributed training setting, where both communication as well as compute efficiency may increase simultaneously with the number of participant nodes.

arXiv:2004.04454공식 코드 · Papers with Code

TensorProjection Layer: A Tensor-Based Dimension Reduction Method in Deep Neural Networks

In this paper, we propose a dimension reduction method specifically designed for tensor-structured feature data in deep neural networks. The method is implemented as a hidden layer, called the TensorProjection layer, which transforms input tensors into output tensors with reduced dimensions through mode-wise projections. The projection directions are treated as model parameters of the layer and are optimized during model training. Our method can serve as an alternative to pooling layers for summarizing image data, or to convolutional layers as a technique for reducing the number of channels. We conduct experiments on tasks such as medical image classification and segmentation, integrating the TensorProjection layer into commonly used baseline architectures to evaluate its effectiveness. Numerical experiments indicate that the proposed method can outperform traditional downsampling methods, such as pooling layers, in our tasks, suggesting it as a promising alternative for feature summarization.

arXiv:2004.03669공식 코드 · Papers with Code

Radon cumulative distribution transform subspace modeling for image classification

We present a new supervised image classification method applicable to a broad class of image deformation models. The method makes use of the previously described Radon Cumulative Distribution Transform (R-CDT) for image data, whose mathematical properties are exploited to express the image data in a form that is more suitable for machine learning. While certain operations such as translation, scaling, and higher-order transformations are challenging to model in native image space, we show the R-CDT can capture some of these variations and thus render the associated image classification problems easier to solve. The method -- utilizing a nearest-subspace algorithm in R-CDT space -- is simple to implement, non-iterative, has no hyper-parameters to tune, is computationally efficient, label efficient, and provides competitive accuracies to state-of-the-art neural networks for many types of classification problems. In addition to the test accuracy performances, we show improvements (with respect to neural network-based methods) in terms of computational efficiency (it can be implemented without the use of GPUs), number of training samples needed for training, as well as out-of-distribution generalization. The Python code for reproducing our results is available at https://github.com/rohdelab/rcdt_ns_classifier.

arXiv:2004.03898

Towards Reusable Network Components by Learning Compatible Representations

This paper proposes to make a first step towards compatible and hence reusable network components. Rather than training networks for different tasks independently, we adapt the training process to produce network components that are compatible across tasks. In particular, we split a network into two components, a features extractor and a target task head, and propose various approaches to accomplish compatibility between them. We systematically analyse these approaches on the task of image classification on standard datasets. We demonstrate that we can produce components which are directly compatible without any fine-tuning or compromising accuracy on the original tasks. Afterwards, we demonstrate the use of compatible components on three applications: Unsupervised domain adaptation, transferring classifiers across feature extractors with different architectures, and increasing the computational efficiency of transfer learning.

arXiv:2004.04141

Empirical Perspectives on One-Shot Semi-supervised Learning

One of the greatest obstacles in the adoption of deep neural networks for new applications is that training the network typically requires a large number of manually labeled training samples. We empirically investigate the scenario where one has access to large amounts of unlabeled data but require labeling only a single prototypical sample per class in order to train a deep network (i.e., one-shot semi-supervised learning). Specifically, we investigate the recent results reported in FixMatch for one-shot semi-supervised learning to understand the factors that affect and impede high accuracies and reliability for one-shot semi-supervised learning of Cifar-10. For example, we discover that one barrier to one-shot semi-supervised learning for high-performance image classification is the unevenness of class accuracy during the training. These results point to solutions that might enable more widespread adoption of one-shot semi-supervised training methods for new applications.

arXiv:2004.03459

Hierarchical Image Classification using Entailment Cone Embeddings

Image classification has been studied extensively, but there has been limited work in using unconventional, external guidance other than traditional image-label pairs for training. We present a set of methods for leveraging information about the semantic hierarchy embedded in class labels. We first inject label-hierarchy knowledge into an arbitrary CNN-based classifier and empirically show that availability of such external semantic information in conjunction with the visual semantics from images boosts overall performance. Taking a step further in this direction, we model more explicitly the label-label and label-image interactions using order-preserving embeddings governed by both Euclidean and hyperbolic geometries, prevalent in natural language, and tailor them to hierarchical image classification and representation learning. We empirically validate all the models on the hierarchical ETHEC dataset.

arXiv:2004.03452

Strategies for Robust Image Classification

In this work we evaluate the impact of digitally altered images on the performance of artificial neural networks. We explore factors that negatively affect the ability of an image classification model to produce consistent and accurate results. A model's ability to classify is negatively influenced by alterations to images as a result of digital abnormalities or changes in the physical environment. The focus of this paper is to discover and replicate scenarios that modify the appearance of an image and evaluate them on state-of-the-art machine learning models. Our contributions present various training techniques that enhance a model's ability to generalize and improve robustness against these alterations.

arXiv:2004.03333공식 코드 · Papers with Code

Binary Neural Networks: A Survey

The binary neural network, largely saving the storage and computation, serves as a promising technique for deploying deep models on resource-limited devices. However, the binarization inevitably causes severe information loss, and even worse, its discontinuity brings difficulty to the optimization of the deep network. To address these issues, a variety of algorithms have been proposed, and achieved satisfying progress in recent years. In this paper, we present a comprehensive survey of these algorithms, mainly categorized into the native solutions directly conducting binarization, and the optimized ones using techniques like minimizing the quantization error, improving the network loss function, and reducing the gradient error. We also investigate other practical aspects of binary neural networks such as the hardware-friendly design and the training tricks. Then, we give the evaluation and discussions on different tasks, including image classification, object detection and semantic segmentation. Finally, the challenges that may be faced in future research are prospected.

arXiv:2004.02967

Evolving Normalization-Activation Layers

Normalization layers and activation functions are fundamental components in deep networks and typically co-locate with each other. Here we propose to design them using an automated approach. Instead of designing them separately, we unify them into a single tensor-to-tensor computation graph, and evolve its structure starting from basic mathematical functions. Examples of such mathematical functions are addition, multiplication and statistical moments. The use of low-level mathematical functions, in contrast to the use of high-level modules in mainstream NAS, leads to a highly sparse and large search space which can be challenging for search methods. To address the challenge, we develop efficient rejection protocols to quickly filter out candidate layers that do not work well. We also use multi-objective evolution to optimize each layer's performance across many architectures to prevent overfitting. Our method leads to the discovery of EvoNorms, a set of new normalization-activation layers with novel, and sometimes surprising structures that go beyond existing design patterns. For example, some EvoNorms do not assume that normalization and activation functions must be applied sequentially, nor need to center the feature maps, nor require explicit activation functions. Our experiments show that EvoNorms work well on image classification models including ResNets, MobileNets and EfficientNets but also transfer well to Mask R-CNN with FPN/SpineNet for instance segmentation and to BigGAN for image synthesis, outperforming BatchNorm and GroupNorm based layers in many cases.

arXiv:2004.03332

Two-Stage Resampling for Convolutional Neural Network Training in the Imbalanced Colorectal Cancer Image Classification

Data imbalance remains one of the open challenges in the contemporary machine learning. It is especially prevalent in case of medical data, such as histopathological images. Traditional data-level approaches for dealing with data imbalance are ill-suited for image data: oversampling methods such as SMOTE and its derivatives lead to creation of unrealistic synthetic observations, whereas undersampling reduces the amount of available data, critical for successful training of convolutional neural networks. To alleviate the problems associated with over- and undersampling we propose a novel two-stage resampling methodology, in which we initially use the oversampling techniques in the image space to leverage a large amount of data for training of a convolutional neural network, and afterwards apply undersampling in the feature space to fine-tune the last layers of the network. Experiments conducted on a colorectal cancer image dataset indicate the usefulness of the proposed approach.

arXiv:2004.03281공식 코드 · Papers with Code

Teacher-Class Network: A Neural Network Compression Mechanism

To reduce the overwhelming size of Deep Neural Networks (DNN) teacher-student methodology tries to transfer knowledge from a complex teacher network to a simple student network. We instead propose a novel method called the teacher-class network consisting of a single teacher and multiple student networks (i.e. class of students). Instead of transferring knowledge to one student only, the proposed method transfers a chunk of knowledge to each student. Our students are not trained for problem-specific logits, they are trained to mimic knowledge (dense representation) learned by the teacher network thus the combined knowledge learned by the class of students can be used to solve other problems as well. The proposed teacher-class architecture is evaluated on several benchmark datasets such as MNIST, Fashion MNIST, IMDB Movie Reviews, CAMVid, CIFAR-10 and ImageNet on multiple tasks including image classification, sentiment classification and segmentation. Our approach outperforms the state of-the-art single student approach in terms of accuracy as well as computational cost while achieving 10-30 times reduction in parameters.

arXiv:2004.03264

Inspector Gadget: A Data Programming-based Labeling System for Industrial Images

As machine learning for images becomes democratized in the Software 2.0 era, one of the serious bottlenecks is securing enough labeled data for training. This problem is especially critical in a manufacturing setting where smart factories rely on machine learning for product quality control by analyzing industrial images. Such images are typically large and may only need to be partially analyzed where only a small portion is problematic (e.g., identifying defects on a surface). Since manual labeling these images is expensive, weak supervision is an attractive alternative where the idea is to generate weak labels that are not perfect, but can be produced at scale. Data programming is a recent paradigm in this category where it uses human knowledge in the form of labeling functions and combines them into a generative model. Data programming has been successful in applications based on text or structured data and can also be applied to images usually if one can find a way to convert them into structured data. In this work, we expand the horizon of data programming by directly applying it to images without this conversion, which is a common scenario for industrial applications. We propose Inspector Gadget, an image labeling system that combines crowdsourcing, data augmentation, and data programming to produce weak labels at scale for image classification. We perform experiments on real industrial image datasets and show that Inspector Gadget obtains better performance than other weak-labeling techniques: Snuba, GOGGLES, and self-learning baselines using convolutional neural networks (CNNs) without pre-training.

arXiv:2004.03188

Increasing the Inference and Learning Speed of Tsetlin Machines with Clause Indexing

The Tsetlin Machine (TM) is a machine learning algorithm founded on the classical Tsetlin Automaton (TA) and game theory. It further leverages frequent pattern mining and resource allocation principles to extract common patterns in the data, rather than relying on minimizing output error, which is prone to overfitting. Unlike the intertwined nature of pattern representation in neural networks, a TM decomposes problems into self-contained patterns, represented as conjunctive clauses. The clause outputs, in turn, are combined into a classification decision through summation and thresholding, akin to a logistic regression function, however, with binary weights and a unit step output function. In this paper, we exploit this hierarchical structure by introducing a novel algorithm that avoids evaluating the clauses exhaustively. Instead we use a simple look-up table that indexes the clauses on the features that falsify them. In this manner, we can quickly evaluate a large number of clauses through falsification, simply by iterating through the features and using the look-up table to eliminate those clauses that are falsified. The look-up table is further structured so that it facilitates constant time updating, thus supporting use also during learning. We report up to 15 times faster classification and three times faster learning on MNIST and Fashion-MNIST image classification, and IMDb sentiment analysis.

arXiv:2004.03109

Generative Adversarial Zero-shot Learning via Knowledge Graphs

Zero-shot learning (ZSL) is to handle the prediction of those unseen classes that have no labeled training data. Recently, generative methods like Generative Adversarial Networks (GANs) are being widely investigated for ZSL due to their high accuracy, generalization capability and so on. However, the side information of classes used now is limited to text descriptions and attribute annotations, which are in short of semantics of the classes. In this paper, we introduce a new generative ZSL method named KG-GAN by incorporating rich semantics in a knowledge graph (KG) into GANs. Specifically, we build upon Graph Neural Networks and encode KG from two views: class view and attribute view considering the different semantics of KG. With well-learned semantic embeddings for each node (representing a visual category), we leverage GANs to synthesize compelling visual features for unseen classes. According to our evaluation with multiple image classification datasets, KG-GAN can achieve better performance than the state-of-the-art baselines.

arXiv:2004.02757

Efficient Deep Representation Learning by Adaptive Latent Space Sampling

Supervised deep learning requires a large amount of training samples with annotations (e.g. label class for classification task, pixel- or voxel-wised label map for segmentation tasks), which are expensive and time-consuming to obtain. During the training of a deep neural network, the annotated samples are fed into the network in a mini-batch way, where they are often regarded of equal importance. However, some of the samples may become less informative during training, as the magnitude of the gradient start to vanish for these samples. In the meantime, other samples of higher utility or hardness may be more demanded for the training process to proceed and require more exploitation. To address the challenges of expensive annotations and loss of sample informativeness, here we propose a novel training framework which adaptively selects informative samples that are fed to the training process. The adaptive selection or sampling is performed based on a hardness-aware strategy in the latent space constructed by a generative model. To evaluate the proposed training framework, we perform experiments on three different datasets, including MNIST and CIFAR-10 for image classification task and a medical image dataset IVUS for biophysical simulation task. On all three datasets, the proposed framework outperforms a random sampling method, which demonstrates the effectiveness of proposed framework.

arXiv:2004.01961공식 코드 · Papers with Code

Neural Architecture Search for Lightweight Non-Local Networks

Non-Local (NL) blocks have been widely studied in various vision tasks. However, it has been rarely explored to embed the NL blocks in mobile neural networks, mainly due to the following challenges: 1) NL blocks generally have heavy computation cost which makes it difficult to be applied in applications where computational resources are limited, and 2) it is an open problem to discover an optimal configuration to embed NL blocks into mobile neural networks. We propose AutoNL to overcome the above two obstacles. Firstly, we propose a Lightweight Non-Local (LightNL) block by squeezing the transformation operations and incorporating compact features. With the novel design choices, the proposed LightNL block is 400x computationally cheaper} than its conventional counterpart without sacrificing the performance. Secondly, by relaxing the structure of the LightNL block to be differentiable during training, we propose an efficient neural architecture search algorithm to learn an optimal configuration of LightNL blocks in an end-to-end manner. Notably, using only 32 GPU hours, the searched AutoNL model achieves 77.7% top-1 accuracy on ImageNet under a typical mobile setting (350M FLOPs), significantly outperforming previous mobile models including MobileNetV2 (+5.7%), FBNet (+2.8%) and MnasNet (+2.1%). Code and models are available at https://github.com/LiYingwei/AutoNL.

arXiv:2004.01817

Group Based Deep Shared Feature Learning for Fine-grained Image Classification

Fine-grained image classification has emerged as a significant challenge because objects in such images have small inter-class visual differences but with large variations in pose, lighting, and viewpoints, etc. Most existing work focuses on highly customized feature extraction via deep network architectures which have been shown to deliver state of the art performance. Given that images from distinct classes in fine-grained classification share significant features of interest, we present a new deep network architecture that explicitly models shared features and removes their effect to achieve enhanced classification results. Our modeling of shared features is based on a new group based learning wherein existing classes are divided into groups and multiple shared feature patterns are discovered (learned). We call this framework Group based deep Shared Feature Learning (GSFL) and the resulting learned network as GSFL-Net. Specifically, the proposed GSFL-Net develops a specially designed autoencoder which is constrained by a newly proposed Feature Expression Loss to decompose a set of features into their constituent shared and discriminative components. During inference, only the discriminative feature component is used to accomplish the classification task. A key benefit of our specialized autoencoder is that it is versatile and can be combined with state-of-the-art fine-grained feature extraction models and trained together with them to improve their performance directly. Experiments on benchmark datasets show that GSFL-Net can enhance classification accuracy over the state of the art with a more interpretable architecture.

arXiv:2004.02183공식 코드 · Papers with Code

Approximate Manifold Defense Against Multiple Adversarial Perturbations

Existing defenses against adversarial attacks are typically tailored to a specific perturbation type. Using adversarial training to defend against multiple types of perturbation requires expensive adversarial examples from different perturbation types at each training step. In contrast, manifold-based defense incorporates a generative network to project an input sample onto the clean data manifold. This approach eliminates the need to generate expensive adversarial examples while achieving robustness against multiple perturbation types. However, the success of this approach relies on whether the generative network can capture the complete clean data manifold, which remains an open problem for complex input domain. In this work, we devise an approximate manifold defense mechanism, called RBF-CNN, for image classification. Instead of capturing the complete data manifold, we use an RBF layer to learn the density of small image patches. RBF-CNN also utilizes a reconstruction layer that mitigates any minor adversarial perturbations. Further, incorporating our proposed reconstruction process for training improves the adversarial robustness of our RBF-CNN models. Experiment results on MNIST and CIFAR-10 datasets indicate that RBF-CNN offers robustness for multiple perturbations without the need for expensive adversarial training.

arXiv:2004.02168

Comparative Analysis of Multiple Deep CNN Models for Waste Classification

Waste is a wealth in a wrong place. Our research focuses on analyzing possibilities for automatic waste sorting and collecting in such a way that helps it for further recycling process. Various approaches are being practiced managing waste but not efficient and require human intervention. The automatic waste segregation would fit in to fill the gap. The project tested well known Deep Learning Network architectures for waste classification with dataset combined from own endeavors and Trash Net. The convolutional neural network is used for image classification. The hardware built in the form of dustbin is used to segregate those wastes into different compartments. Without the human exercise in segregating those waste products, the study would save the precious time and would introduce the automation in the area of waste management. Municipal solid waste is a huge, renewable source of energy. The situation is win-win for both government, society and industrialists. Because of fine-tuning of the ResNet18 Network, the best validation accuracy was found to be 87.8%.