PyTorchKR Papers

최신 논문

총 1,324편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2202.06592공식 코드 · Papers with Code

Memory Replay with Data Compression for Continual Learning

Continual learning needs to overcome catastrophic forgetting of the past. Memory replay of representative old training samples has been shown as an effective solution, and achieves the state-of-the-art (SOTA) performance. However, existing work is mainly built on a small memory buffer containing a few original data, which cannot fully characterize the old data distribution. In this work, we propose memory replay with data compression (MRDC) to reduce the storage cost of old training samples and thus increase their amount that can be stored in the memory buffer. Observing that the trade-off between the quality and quantity of compressed data is highly nontrivial for the efficacy of memory replay, we propose a novel method based on determinantal point processes (DPPs) to efficiently determine an appropriate compression quality for currently-arrived training samples. In this way, using a naive data compression algorithm with a properly selected quality can largely boost recent strong baselines by saving more compressed data in a limited storage space. We extensively validate this across several benchmarks of class-incremental learning and in a realistic scenario of object detection for autonomous driving.

arXiv:2202.06580

Improved Aggregating and Accelerating Training Methods for Spatial Graph Neural Networks on Fraud Detection

Graph neural networks (GNNs) have been widely applied to numerous fields. A recent work which combines layered structure and residual connection proposes an improved deep architecture to extend CAmouflage-REsistant GNN (CARE-GNN) to deep models named as Residual Layered CARE-GNN (RLC-GNN), which forms self-correcting and incremental learning mechanism, and achieves significant performance improvements on fraud detection task. However, we spot three issues of RLC-GNN, which are the usage of neighboring information reaching limitation, the training difficulty which is inherent problem to deep models and lack of comprehensive consideration about node features and external patterns. In this work, we propose three approaches to solve those three problems respectively. First, we suggest conducting similarity measure via cosine distance to take both local features and external patterns into consideration. Then, we combine the similarity measure module and the idea of adjacency-wise normalization with node-wise and batch-wise normalization and then propound partial neighborhood normalization methods to overcome the training difficulty while mitigating the impact of too much noise caused by high-density of graph. Finally, we put forward intermediate information supplement to solve the information limitation. Experiments are conducted on Yelp and Amazon datasets. And the results show that our proposed methods effectively solve the three problems. After applying the three methods, we achieve 4.81%, 6.62% and 6.81% improvements in the metrics of recall, AUC and Macro-F1 respectively on the Yelp dataset. And we obtain 1.65% and 0.29% improvements in recall and AUC respectively on the Amazon datasets.

arXiv:2202.05411공식 코드 · Papers with Code

Incremental Learning of Structured Memory via Closed-Loop Transcription

This work proposes a minimal computational model for learning structured memories of multiple object classes in an incremental setting. Our approach is based on establishing a closed-loop transcription between the classes and a corresponding set of subspaces, known as a linear discriminative representation, in a low-dimensional feature space. Our method is simpler than existing approaches for incremental learning, and more efficient in terms of model size, storage, and computation: it requires only a single, fixed-capacity autoencoding network with a feature space that is used for both discriminative and generative purposes. Network parameters are optimized simultaneously without architectural manipulations, by solving a constrained minimax game between the encoding and decoding maps over a single rate reduction-based objective. Experimental results show that our method can effectively alleviate catastrophic forgetting, achieving significantly better performance than prior work of generative replay on MNIST, CIFAR-10, and ImageNet-50, despite requiring fewer resources. Source code can be found at https://github.com/tsb0601/i-CTRL

arXiv:2202.04479

False Memory Formation in Continual Learners Through Imperceptible Backdoor Trigger

In this brief, we show that sequentially learning new information presented to a continual (incremental) learning model introduces new security risks: an intelligent adversary can introduce small amount of misinformation to the model during training to cause deliberate forgetting of a specific task or class at test time, thus creating "false memory" about that task. We demonstrate such an adversary's ability to assume control of the model by injecting "backdoor" attack samples to commonly used generative replay and regularization based continual learning approaches using continual learning benchmark variants of MNIST, as well as the more challenging SVHN and CIFAR 10 datasets. Perhaps most damaging, we show this vulnerability to be very acute and exceptionally effective: the backdoor pattern in our attack model can be imperceptible to human eye, can be provided at any point in time, can be added into the training data of even a single possibly unrelated task and can be achieved with as few as just 1\% of total training dataset of a single task.

arXiv:2202.03703

Self-Paced Imbalance Rectification for Class Incremental Learning

Exemplar-based class-incremental learning is to recognize new classes while not forgetting old ones, whose samples can only be saved in limited memory. The ratio fluctuation of new samples to old exemplars, which is caused by the variation of memory capacity at different environments, will bring challenges to stabilize the incremental optimization process. To address this problem, we propose a novel self-paced imbalance rectification scheme, which dynamically maintains the incremental balance during the representation learning phase. Specifically, our proposed scheme consists of a frequency compensation strategy that adjusts the logits margin between old and new classes with the corresponding number ratio to strengthen the expression ability of the old classes, and an inheritance transfer strategy to reduce the representation confusion by estimating the similarity of different classes in the old embedding space. Furthermore, a chronological attenuation mechanism is proposed to mitigate the repetitive optimization of the older classes at multiple step-wise increments. Extensive experiments on three benchmarks demonstrate stable incremental performance, significantly outperforming the state-of-the-art methods.

arXiv:2202.02601

Exemplar-Based Contrastive Self-Supervised Learning with Few-Shot Class Incremental Learning

Humans are capable of learning new concepts from only a few (labeled) exemplars, incrementally and continually. This happens within the context that we can differentiate among the exemplars, and between the exemplars and large amounts of other data (unlabeled and labeled). This suggests, in human learning, supervised learning of concepts based on exemplars takes place within the larger context of contrastive self-supervised learning (CSSL) based on unlabeled and labeled data. We discuss extending CSSL (1) to be based mainly on exemplars and only secondly on data augmentation, and (2) to apply to both unlabeled data (a large amount is available in general) and labeled data (a few exemplars can be obtained with valuable supervised knowledge). A major benefit of the extensions is that exemplar-based CSSL, with supervised finetuning, supports few-shot class incremental learning (CIL). Specifically, we discuss exemplar-based CSSL including: nearest-neighbor CSSL, neighborhood CSSL with supervised pretraining, and exemplar CSSL with supervised finetuning. We further discuss using exemplar-based CSSL to facilitate few-shot learning and, in particular, few-shot CIL.

arXiv:2202.00432공식 코드 · Papers with Code

Continual Attentive Fusion for Incremental Learning in Semantic Segmentation

Over the past years, semantic segmentation, as many other tasks in computer vision, benefited from the progress in deep neural networks, resulting in significantly improved performance. However, deep architectures trained with gradient-based techniques suffer from catastrophic forgetting, which is the tendency to forget previously learned knowledge while learning new tasks. Aiming at devising strategies to counteract this effect, incremental learning approaches have gained popularity over the past years. However, the first incremental learning methods for semantic segmentation appeared only recently. While effective, these approaches do not account for a crucial aspect in pixel-level dense prediction problems, i.e. the role of attention mechanisms. To fill this gap, in this paper we introduce a novel attentive feature distillation approach to mitigate catastrophic forgetting while accounting for semantic spatial- and channel-level dependencies. Furthermore, we propose a {continual attentive fusion} structure, which takes advantage of the attention learned from the new and the old tasks while learning features for the new task. Finally, we also introduce a novel strategy to account for the background class in the distillation loss, thus preventing biased predictions. We demonstrate the effectiveness of our approach with an extensive evaluation on Pascal-VOC 2012 and ADE20K, setting a new state of the art.

arXiv:2202.00386

A Comparative Study of Calibration Methods for Imbalanced Class Incremental Learning

Deep learning approaches are successful in a wide range of AI problems and in particular for visual recognition tasks. However, there are still open problems among which is the capacity to handle streams of visual information and the management of class imbalance in datasets. Existing research approaches these two problems separately while they co-occur in real world applications. Here, we study the problem of learning incrementally from imbalanced datasets. We focus on algorithms which have a constant deep model complexity and use a bounded memory to store exemplars of old classes across incremental states. Since memory is bounded, old classes are learned with fewer images than new classes and an imbalance due to incremental learning is added to the initial dataset imbalance. A score prediction bias in favor of new classes appears and we evaluate a comprehensive set of score calibration methods to reduce it. Evaluation is carried with three datasets, using two dataset imbalance configurations and three bounded memory sizes. Results show that most calibration methods have beneficial effect and that they are most useful for lower bounded memory sizes, which are most interesting in practice. As a secondary contribution, we remove the usual distillation component from the loss function of incremental learning algorithms. We show that simpler vanilla fine tuning is a stronger backbone for imbalanced incremental learning algorithms.

arXiv:2201.12559공식 코드 · Papers with Code

Rebalancing Batch Normalization for Exemplar-based Class-Incremental Learning

Batch Normalization (BN) and its variants has been extensively studied for neural nets in various computer vision tasks, but relatively little work has been dedicated to studying the effect of BN in continual learning. To that end, we develop a new update patch for BN, particularly tailored for the exemplar-based class-incremental learning (CIL). The main issue of BN in CIL is the imbalance of training data between current and past tasks in a mini-batch, which makes the empirical mean and variance as well as the learnable affine transformation parameters of BN heavily biased toward the current task -- contributing to the forgetting of past tasks. While one of the recent BN variants has been developed for "online" CIL, in which the training is done with a single epoch, we show that their method does not necessarily bring gains for "offline" CIL, in which a model is trained with multiple epochs on the imbalanced training data. The main reason for the ineffectiveness of their method lies in not fully addressing the data imbalance issue, especially in computing the gradients for learning the affine transformation parameters of BN. Accordingly, our new hyperparameter-free variant, dubbed as Task-Balanced BN (TBBN), is proposed to more correctly resolve the imbalance issue by making a horizontally-concatenated task-balanced batch using both reshape and repeat operations during training. Based on our experiments on class incremental learning of CIFAR-100, ImageNet-100, and five dissimilar task datasets, we demonstrate that our TBBN, which works exactly the same as the vanilla BN in the inference time, is easily applicable to most existing exemplar-based offline CIL algorithms and consistently outperforms other BN variants.

arXiv:2201.10947

Enabling Deep Learning on Edge Devices through Filter Pruning and Knowledge Transfer

Deep learning models have introduced various intelligent applications to edge devices, such as image classification, speech recognition, and augmented reality. There is an increasing need of training such models on the devices in order to deliver personalized, responsive, and private learning. To address this need, this paper presents a new solution for deploying and training state-of-the-art models on the resource-constrained devices. First, the paper proposes a novel filter-pruning-based model compression method to create lightweight trainable models from large models trained in the cloud, without much loss of accuracy. Second, it proposes a novel knowledge transfer method to enable the on-device model to update incrementally in real time or near real time using incremental learning on new data and enable the on-device model to learn the unseen categories with the help of the in-cloud model in an unsupervised fashion. The results show that 1) our model compression method can remove up to 99.36% parameters of WRN-28-10, while preserving a Top-1 accuracy of over 90% on CIFAR-10; 2) our knowledge transfer method enables the compressed models to achieve more than 90% accuracy on CIFAR-10 and retain good accuracy on old categories; 3) it allows the compressed models to converge within real time (three to six minutes) on the edge for incremental learning tasks; 4) it enables the model to classify unseen categories of data (78.92% Top-1 accuracy) that it is never trained with.

arXiv:2201.09381

vCLIMB: A Novel Video Class Incremental Learning Benchmark

Continual learning (CL) is under-explored in the video domain. The few existing works contain splits with imbalanced class distributions over the tasks, or study the problem in unsuitable datasets. We introduce vCLIMB, a novel video continual learning benchmark. vCLIMB is a standardized test-bed to analyze catastrophic forgetting of deep models in video continual learning. In contrast to previous work, we focus on class incremental continual learning with models trained on a sequence of disjoint tasks, and distribute the number of classes uniformly across the tasks. We perform in-depth evaluations of existing CL methods in vCLIMB, and observe two unique challenges in video data. The selection of instances to store in episodic memory is performed at the frame level. Second, untrimmed training data influences the effectiveness of frame sampling strategies. We address these two challenges by proposing a temporal consistency regularization that can be applied on top of memory-based continual learning methods. Our approach significantly improves the baseline, by up to 24% on the untrimmed continual learning task.

arXiv:2201.06740

Convolutional Cobweb: A Model of Incremental Learning from 2D Images

This paper presents a new concept formation approach that supports the ability to incrementally learn and predict labels for visual images. This work integrates the idea of convolutional image processing, from computer vision research, with a concept formation approach that is based on psychological studies of how humans incrementally form and use concepts. We experimentally evaluate this new approach by applying it to an incremental variation of the MNIST digit recognition task. We compare its performance to Cobweb, a concept formation approach that does not support convolutional processing, as well as two convolutional neural networks that vary in the complexity of their convolutional processing. This work represents a first step towards unifying modern computer vision ideas with classical concept formation research.

arXiv:2201.02574

An Incremental Learning Approach to Automatically Recognize Pulmonary Diseases from the Multi-vendor Chest Radiographs

Pulmonary diseases can cause severe respiratory problems, leading to sudden death if not treated timely. Many researchers have utilized deep learning systems to diagnose pulmonary disorders using chest X-rays (CXRs). However, such systems require exhaustive training efforts on large-scale data to effectively diagnose chest abnormalities. Furthermore, procuring such large-scale data is often infeasible and impractical, especially for rare diseases. With the recent advances in incremental learning, researchers have periodically tuned deep neural networks to learn different classification tasks with few training examples. Although, such systems can resist catastrophic forgetting, they treat the knowledge representations independently of each other, and this limits their classification performance. Also, to the best of our knowledge, there is no incremental learning-driven image diagnostic framework that is specifically designed to screen pulmonary disorders from the CXRs. To address this, we present a novel framework that can learn to screen different chest abnormalities incrementally. In addition to this, the proposed framework is penalized through an incremental learning loss function that infers Bayesian theory to recognize structural and semantic inter-dependencies between incrementally learned knowledge representations to diagnose the pulmonary diseases effectively, regardless of the scanner specifications. We tested the proposed framework on five public CXR datasets containing different chest abnormalities, where it outperformed various state-of-the-art system through various metrics.

arXiv:2201.02560공식 코드 · Papers with Code

A Novel Incremental Learning Driven Instance Segmentation Framework to Recognize Highly Cluttered Instances of the Contraband Items

Screening cluttered and occluded contraband items from baggage X-ray scans is a cumbersome task even for the expert security staff. This paper presents a novel strategy that extends a conventional encoder-decoder architecture to perform instance-aware segmentation and extract merged instances of contraband items without using any additional sub-network or an object detector. The encoder-decoder network first performs conventional semantic segmentation and retrieves cluttered baggage items. The model then incrementally evolves during training to recognize individual instances using significantly reduced training batches. To avoid catastrophic forgetting, a novel objective function minimizes the network loss in each iteration by retaining the previously acquired knowledge while learning new class representations and resolving their complex structural inter-dependencies through Bayesian inference. A thorough evaluation of our framework on two publicly available X-ray datasets shows that it outperforms state-of-the-art methods, especially within the challenging cluttered scenarios, while achieving an optimal trade-off between detection accuracy and efficiency.

arXiv:2201.01633

Adaptive Online Incremental Learning for Evolving Data Streams

Recent years have witnessed growing interests in online incremental learning. However, there are three major challenges in this area. The first major difficulty is concept drift, that is, the probability distribution in the streaming data would change as the data arrives. The second major difficulty is catastrophic forgetting, that is, forgetting what we have learned before when learning new knowledge. The last one we often ignore is the learning of the latent representation. Only good latent representation can improve the prediction accuracy of the model. Our research builds on this observation and attempts to overcome these difficulties. To this end, we propose an Adaptive Online Incremental Learning for evolving data streams (AOIL). We use auto-encoder with the memory module, on the one hand, we obtained the latent features of the input, on the other hand, according to the reconstruction loss of the auto-encoder with memory module, we could successfully detect the existence of concept drift and trigger the update mechanism, adjust the model parameters in time. In addition, we divide features, which are derived from the activation of the hidden layers, into two parts, which are used to extract the common and private features respectively. By means of this approach, the model could learn the private features of the new coming instances, but do not forget what we have learned in the past (shared features), which reduces the occurrence of catastrophic forgetting. At the same time, to get the fusion feature vector we use the self-attention mechanism to effectively fuse the extracted features, which further improved the latent representation learning.

arXiv:2201.01488공식 코드 · Papers with Code

Exemplar-free Class Incremental Learning via Discriminative and Comparable One-class Classifiers

The exemplar-free class incremental learning requires classification models to learn new class knowledge incrementally without retaining any old samples. Recently, the framework based on parallel one-class classifiers (POC), which trains a one-class classifier (OCC) independently for each category, has attracted extensive attention, since it can naturally avoid catastrophic forgetting. POC, however, suffers from weak discriminability and comparability due to its independent training strategy for different OOCs. To meet this challenge, we propose a new framework, named Discriminative and Comparable One-class classifiers for Incremental Learning (DisCOIL). DisCOIL follows the basic principle of POC, but it adopts variational auto-encoders (VAE) instead of other well-established one-class classifiers (e.g. deep SVDD), because a trained VAE can not only identify the probability of an input sample belonging to a class but also generate pseudo samples of the class to assist in learning new tasks. With this advantage, DisCOIL trains a new-class VAE in contrast with the old-class VAEs, which forces the new-class VAE to reconstruct better for new-class samples but worse for the old-class pseudo samples, thus enhancing the comparability. Furthermore, DisCOIL introduces a hinge reconstruction loss to ensure the discriminability. We evaluate our method extensively on MNIST, CIFAR10, and Tiny-ImageNet. The experimental results show that DisCOIL achieves state-of-the-art performance.

arXiv:2201.01073공식 코드 · Papers with Code

Towards Unsupervised Open World Semantic Segmentation

For the semantic segmentation of images, state-of-the-art deep neural networks (DNNs) achieve high segmentation accuracy if that task is restricted to a closed set of classes. However, as of now DNNs have limited ability to operate in an open world, where they are tasked to identify pixels belonging to unknown objects and eventually to learn novel classes, incrementally. Humans have the capability to say: I don't know what that is, but I've already seen something like that. Therefore, it is desirable to perform such an incremental learning task in an unsupervised fashion. We introduce a method where unknown objects are clustered based on visual similarity. Those clusters are utilized to define new classes and serve as training data for unsupervised incremental learning. More precisely, the connected components of a predicted semantic segmentation are assessed by a segmentation quality estimate. connected components with a low estimated prediction quality are candidates for a subsequent clustering. Additionally, the component-wise quality assessment allows for obtaining predicted segmentation masks for the image regions potentially containing unknown objects. The respective pixels of such masks are pseudo-labeled and afterwards used for re-training the DNN, i.e., without the use of ground truth generated by humans. In our experiments we demonstrate that, without access to ground truth and even with few data, a DNN's class space can be extended by a novel class, achieving considerable segmentation accuracy.

arXiv:2201.01046

Sound and Visual Representation Learning with Multiple Pretraining Tasks

Different self-supervised tasks (SSL) reveal different features from the data. The learned feature representations can exhibit different performance for each downstream task. In this light, this work aims to combine Multiple SSL tasks (Multi-SSL) that generalizes well for all downstream tasks. Specifically, for this study, we investigate binaural sounds and image data in isolation. For binaural sounds, we propose three SSL tasks namely, spatial alignment, temporal synchronization of foreground objects and binaural audio and temporal gap prediction. We investigate several approaches of Multi-SSL and give insights into the downstream task performance on video retrieval, spatial sound super resolution, and semantic prediction on the OmniAudio dataset. Our experiments on binaural sound representations demonstrate that Multi-SSL via incremental learning (IL) of SSL tasks outperforms single SSL task models and fully supervised models in the downstream task performance. As a check of applicability on other modality, we also formulate our Multi-SSL models for image representation learning and we use the recently proposed SSL tasks, MoCov2 and DenseCL. Here, Multi-SSL surpasses recent methods such as MoCov2, DenseCL and DetCo by 2.06%, 3.27% and 1.19% on VOC07 classification and +2.83, +1.56 and +1.61 AP on COCO detection. Code will be made publicly available.

arXiv:2112.09834

Improving the performance of bagging ensembles for data streams through mini-batching

Often, machine learning applications have to cope with dynamic environments where data are collected in the form of continuous data streams with potentially infinite length and transient behavior. Compared to traditional (batch) data mining, stream processing algorithms have additional requirements regarding computational resources and adaptability to data evolution. They must process instances incrementally because the data's continuous flow prohibits storing data for multiple passes. Ensemble learning achieved remarkable predictive performance in this scenario. Implemented as a set of (several) individual classifiers, ensembles are naturally amendable for task parallelism. However, the incremental learning and dynamic data structures used to capture the concept drift increase the cache misses and hinder the benefit of parallelism. This paper proposes a mini-batching strategy that can improve memory access locality and performance of several ensemble algorithms for stream mining in multi-core environments. With the aid of a formal framework, we demonstrate that mini-batching can significantly decrease the reuse distance (and the number of cache misses). Experiments on six different state-of-the-art ensemble algorithms applying four benchmark datasets with varied characteristics show speedups of up to 5X on 8-core processors. These benefits come at the expense of a small reduction in predictive performance.

arXiv:2112.15022공식 코드 · Papers with Code

Continually Learning Self-Supervised Representations with Projected Functional Regularization

Recent self-supervised learning methods are able to learn high-quality image representations and are closing the gap with supervised approaches. However, these methods are unable to acquire new knowledge incrementally -- they are, in fact, mostly used only as a pre-training phase over IID data. In this work we investigate self-supervised methods in continual learning regimes without any replay mechanism. We show that naive functional regularization, also known as feature distillation, leads to lower plasticity and limits continual learning performance. Instead, we propose Projected Functional Regularization in which a separate temporal projection network ensures that the newly learned feature space preserves information of the previous one, while at the same time allowing for the learning of new features. This prevents forgetting while maintaining the plasticity of the learner. Comparison with other incremental learning approaches applied to self-supervision demonstrates that our method obtains competitive performance in different scenarios and on multiple datasets.