PyTorchKR Papers

최신 논문

총 1,324편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2107.08918공식 코드 · Papers with Code

Self-Promoted Prototype Refinement for Few-Shot Class-Incremental Learning

Few-shot class-incremental learning is to recognize the new classes given few samples and not forget the old classes. It is a challenging task since representation optimization and prototype reorganization can only be achieved under little supervision. To address this problem, we propose a novel incremental prototype learning scheme. Our scheme consists of a random episode selection strategy that adapts the feature representation to various generated incremental episodes to enhance the corresponding extensibility, and a self-promoted prototype refinement mechanism which strengthens the expression ability of the new classes by explicitly considering the dependencies among different classes. Particularly, a dynamic relation projection module is proposed to calculate the relation matrix in a shared embedding space and leverage it as the factor for bootstrapping the update of prototypes. Extensive experiments on three benchmark datasets demonstrate the above-par incremental performance, outperforming state-of-the-art methods by a margin of 13%, 17% and 11%, respectively.

arXiv:2107.08387

Train on Small, Play the Large: Scaling Up Board Games with AlphaZero and GNN

Playing board games is considered a major challenge for both humans and AI researchers. Because some complicated board games are quite hard to learn, humans usually begin with playing on smaller boards and incrementally advance to master larger board strategies. Most neural network frameworks that are currently tasked with playing board games neither perform such incremental learning nor possess capabilities to automatically scale up. In this work, we look at the board as a graph and combine a graph neural network architecture inside the AlphaZero framework, along with some other innovative improvements. Our ScalableAlphaZero is capable of learning to play incrementally on small boards, and advancing to play on large ones. Our model can be trained quickly to play different challenging board games on multiple board sizes, without using any domain knowledge. We demonstrate the effectiveness of ScalableAlphaZero and show, for example, that by training it for only three days on small Othello boards, it can defeat the AlphaZero model on a large board, which was trained to play the large board for $30$ days.

arXiv:2107.07106

Online Learning for Recommendations at Grubhub

We propose a method to easily modify existing offline Recommender Systems to run online using Transfer Learning. Online Learning for Recommender Systems has two main advantages: quality and scale. Like many Machine Learning algorithms in production if not regularly retrained will suffer from Concept Drift. A policy that is updated frequently online can adapt to drift faster than a batch system. This is especially true for user-interaction systems like recommenders where the underlying distribution can shift drastically to follow user behaviour. As a platform grows rapidly like Grubhub, the cost of running batch training jobs becomes material. A shift from stateless batch learning offline to stateful incremental learning online can recover, for example, at Grubhub, up to a 45x cost savings and a +20% metrics increase. There are a few challenges to overcome with the transition to online stateful learning, namely convergence, non-stationary embeddings and off-policy evaluation, which we explore from our experiences running this system in production.

arXiv:2107.04464

A First Look at Class Incremental Learning in Deep Learning Mobile Traffic Classification

The recent popularity growth of Deep Learning (DL) re-ignited the interest towards traffic classification, with several studies demonstrating the accuracy of DL-based classifiers to identify Internet applications' traffic. Even with the aid of hardware accelerators (GPUs, TPUs), DL model training remains expensive, and limits the ability to operate frequent model updates necessary to fit to the ever evolving nature of Internet traffic, and mobile traffic in particular. To address this pain point, in this work we explore Incremental Learning (IL) techniques to add new classes to models without a full retraining, hence speeding up model's updates cycle. We consider iCarl, a state of the art IL method, and MIRAGE-2019, a public dataset with traffic from 40 Android apps, aiming to understand "if there is a case for incremental learning in traffic classification". By dissecting iCarl internals, we discuss ways to improve its design, contributing a revised version, namely iCarl+. Despite our analysis reveals their infancy, IL techniques are a promising research area on the roadmap towards automated DL-based traffic analysis systems.

arXiv:2107.01349공식 코드 · Papers with Code

Split-and-Bridge: Adaptable Class Incremental Learning within a Single Neural Network

Continual learning has been a major problem in the deep learning community, where the main challenge is how to effectively learn a series of newly arriving tasks without forgetting the knowledge of previous tasks. Initiated by Learning without Forgetting (LwF), many of the existing works report that knowledge distillation is effective to preserve the previous knowledge, and hence they commonly use a soft label for the old task, namely a knowledge distillation (KD) loss, together with a class label for the new task, namely a cross entropy (CE) loss, to form a composite loss for a single neural network. However, this approach suffers from learning the knowledge by a CE loss as a KD loss often more strongly influences the objective function when they are in a competitive situation within a single network. This could be a critical problem particularly in a class incremental scenario, where the knowledge across tasks as well as within the new task, both of which can only be acquired by a CE loss, is essentially learned due to the existence of a unified classifier. In this paper, we propose a novel continual learning method, called Split-and-Bridge, which can successfully address the above problem by partially splitting a neural network into two partitions for training the new task separated from the old task and re-connecting them for learning the knowledge across tasks. In our thorough experimental analysis, our Split-and-Bridge method outperforms the state-of-the-art competitors in KD-based continual learning.

arXiv:2106.15827

When Video Classification Meets Incremental Classes

With the rapid development of social media, tremendous videos with new classes are generated daily, which raise an urgent demand for video classification methods that can continuously update new classes while maintaining the knowledge of old videos with limited storage and computing resources. In this paper, we summarize this task as Class-Incremental Video Classification (CIVC) and propose a novel framework to address it. As a subarea of incremental learning tasks, the challenge of catastrophic forgetting is unavoidable in CIVC. To better alleviate it, we utilize some characteristics of videos. First, we decompose the spatio-temporal knowledge before distillation rather than treating it as a whole in the knowledge transfer process; trajectory is also used to refine the decomposition. Second, we propose a dual granularity exemplar selection method to select and store representative video instances of old classes and key-frames inside videos under a tight storage budget. We benchmark our method and previous SOTA class-incremental learning methods on Something-Something V2 and Kinetics datasets, and our method outperforms previous methods significantly.

arXiv:2106.13233

Post-Selections in AI and How to Avoid Them

Neural network based Artificial Intelligence (AI) has reported increasing scales in experiments. However, this paper raises a rarely reported stage in such experiments called Post-Selection alter the reader to several possible protocol flaws that may result in misleading results. All AI methods fall into two broad schools, connectionist and symbolic. The Post-Selection fall into two kinds, Post-Selection Using Validation Sets (PSUVS) and Post-Selection Using Test Sets (PSUTS). Each kind has two types of post-selectors, machines and humans. The connectionist school received criticisms for its "black box" and now the Post-Selection; but the seemingly "clean" symbolic school seems more brittle because of its human PSUTS. This paper first presents a controversial view: all static "big data" are non-scalable. We then analyze why error-backprop from randomly initialized weights suffers from severe local minima, why PSUVS lacks cross-validation, why PSUTS violates well-established protocols, and why every paper involved should transparently report the Post-Selection stage. To avoid future pitfalls in AI competitions, this paper proposes a new AI metrics, called developmental errors for all networks trained, under Three Learning Conditions: (1) an incremental learning architecture (due to a "big data" flaw), (2) a training experience and (3) a limited amount of computational resources. Developmental Networks avoid Post-Selections because they automatically discover context-rules on the fly by generating emergent Turing machines (not black boxes) that are optimal in the sense of maximum-likelihood across lifetime, conditioned on the Three Learning Conditions.

arXiv:2106.12384공식 코드 · Papers with Code

Reinforcement Learning-based Dialogue Guided Event Extraction to Exploit Argument Relations

Event extraction is a fundamental task for natural language processing. Finding the roles of event arguments like event participants is essential for event extraction. However, doing so for real-life event descriptions is challenging because an argument's role often varies in different contexts. While the relationship and interactions between multiple arguments are useful for settling the argument roles, such information is largely ignored by existing approaches. This paper presents a better approach for event extraction by explicitly utilizing the relationships of event arguments. We achieve this through a carefully designed task-oriented dialogue system. To model the argument relation, we employ reinforcement learning and incremental learning to extract multiple arguments via a multi-turned, iterative process. Our approach leverages knowledge of the already extracted arguments of the same sentence to determine the role of arguments that would be difficult to decide individually. It then uses the newly obtained information to improve the decisions of previously extracted arguments. This two-way feedback process allows us to exploit the argument relations to effectively settle argument roles, leading to better sentence understanding and event extraction. Experimental results show that our approach consistently outperforms seven state-of-the-art event extraction methods for the classification of events and argument role and argument identification.

arXiv:2106.11930공식 코드 · Papers with Code

On the importance of cross-task features for class-incremental learning

In class-incremental learning, an agent with limited resources needs to learn a sequence of classification tasks, forming an ever growing classification problem, with the constraint of not being able to access data from previous tasks. The main difference with task-incremental learning, where a task-ID is available at inference time, is that the learner also needs to perform cross-task discrimination, i.e. distinguish between classes that have not been seen together. Approaches to tackle this problem are numerous and mostly make use of an external memory (buffer) of non-negligible size. In this paper, we ablate the learning of cross-task features and study its influence on the performance of basic replay strategies used for class-IL. We also define a new forgetting measure for class-incremental learning, and see that forgetting is not the principal cause of low performance. Our experimental results show that future algorithms for class-incremental learning should not only prevent forgetting, but also aim to improve the quality of the cross-task features, and the knowledge transfer between tasks. This is especially important when tasks contain limited amount of data.

arXiv:2106.11562공식 코드 · Papers with Code

SSUL: Semantic Segmentation with Unknown Label for Exemplar-based Class-Incremental Learning

This paper introduces a solid state-of-the-art baseline for a class-incremental semantic segmentation (CISS) problem. While the recent CISS algorithms utilize variants of the knowledge distillation (KD) technique to tackle the problem, they failed to fully address the critical challenges in CISS causing the catastrophic forgetting; the semantic drift of the background class and the multi-label prediction issue. To better address these challenges, we propose a new method, dubbed SSUL-M (Semantic Segmentation with Unknown Label with Memory), by carefully combining techniques tailored for semantic segmentation. Specifically, we claim three main contributions. (1) defining unknown classes within the background class to help to learn future classes (help plasticity), (2) freezing backbone network and past classifiers with binary cross-entropy loss and pseudo-labeling to overcome catastrophic forgetting (help stability), and (3) utilizing tiny exemplar memory for the first time in CISS to improve both plasticity and stability. The extensively conducted experiments show the effectiveness of our method, achieving significantly better performance than the recent state-of-the-art baselines on the standard benchmark datasets. Furthermore, we justify our contributions with thorough ablation analyses and discuss different natures of the CISS problem compared to the traditional class-incremental learning targeting classification. The official code is available at https://github.com/clovaai/SSUL.

arXiv:2106.11437

Incremental Deep Neural Network Learning using Classification Confidence Thresholding

Most modern neural networks for classification fail to take into account the concept of the unknown. Trained neural networks are usually tested in an unrealistic scenario with only examples from a closed set of known classes. In an attempt to develop a more realistic model, the concept of working in an open set environment has been introduced. This in turn leads to the concept of incremental learning where a model with its own architecture and initial trained set of data can identify unknown classes during the testing phase and autonomously update itself if evidence of a new class is detected. Some problems that arise in incremental learning are inefficient use of resources to retrain the classifier repeatedly and the decrease of classification accuracy as multiple classes are added over time. This process of instantiating new classes is repeated as many times as necessary, accruing errors. To address these problems, this paper proposes the Classification Confidence Threshold approach to prime neural networks for incremental learning to keep accuracies high by limiting forgetting. A lean method is also used to reduce resources used in the retraining of the neural network. The proposed method is based on the idea that a network is able to incrementally learn a new class even when exposed to a limited number samples associated with the new class. This method can be applied to most existing neural networks with minimal changes to network architecture.

arXiv:2106.09835

Dual-Teacher Class-Incremental Learning With Data-Free Generative Replay

This paper proposes two novel knowledge transfer techniques for class-incremental learning (CIL). First, we propose data-free generative replay (DF-GR) to mitigate catastrophic forgetting in CIL by using synthetic samples from a generative model. In the conventional generative replay, the generative model is pre-trained for old data and shared in extra memory for later incremental learning. In our proposed DF-GR, we train a generative model from scratch without using any training data, based on the pre-trained classification model from the past, so we curtail the cost of sharing pre-trained generative models. Second, we introduce dual-teacher information distillation (DT-ID) for knowledge distillation from two teachers to one student. In CIL, we use DT-ID to learn new classes incrementally based on the pre-trained model for old classes and another model (pre-)trained on the new data for new classes. We implemented the proposed schemes on top of one of the state-of-the-art CIL methods and showed the performance improvement on CIFAR-100 and ImageNet datasets.

arXiv:2106.09701공식 코드 · Papers with Code

Always Be Dreaming: A New Approach for Data-Free Class-Incremental Learning

Modern computer vision applications suffer from catastrophic forgetting when incrementally learning new concepts over time. The most successful approaches to alleviate this forgetting require extensive replay of previously seen data, which is problematic when memory constraints or data legality concerns exist. In this work, we consider the high-impact problem of Data-Free Class-Incremental Learning (DFCIL), where an incremental learning agent must learn new concepts over time without storing generators or training data from past tasks. One approach for DFCIL is to replay synthetic images produced by inverting a frozen copy of the learner's classification model, but we show this approach fails for common class-incremental benchmarks when using standard distillation strategies. We diagnose the cause of this failure and propose a novel incremental distillation strategy for DFCIL, contributing a modified cross-entropy training and importance-weighted feature distillation, and show that our method results in up to a 25.1% increase in final task accuracy (absolute difference) compared to SOTA DFCIL methods for common class-incremental benchmarks. Our method even outperforms several standard replay based methods which store a coreset of images.

arXiv:2106.07936

Modeling morphology with Linear Discriminative Learning: considerations and design choices

This study addresses a series of methodological questions that arise when modeling inflectional morphology with Linear Discriminative Learning. Taking the semi-productive German noun system as example, we illustrate how decisions made about the representation of form and meaning influence model performance. We clarify that for modeling frequency effects in learning, it is essential to make use of incremental learning rather than the endstate of learning. We also discuss how the model can be set up to approximate the learning of inflected words in context. In addition, we illustrate how in this approach the wug task can be modeled in considerable detail. In general, the model provides an excellent memory for known words, but appropriately shows more limited performance for unseen data, in line with the semi-productivity of German noun inflection and generalization performance of native German speakers.

arXiv:2106.07268

FastICARL: Fast Incremental Classifier and Representation Learning with Efficient Budget Allocation in Audio Sensing Applications

Various incremental learning (IL) approaches have been proposed to help deep learning models learn new tasks/classes continuously without forgetting what was learned previously (i.e., avoid catastrophic forgetting). With the growing number of deployed audio sensing applications that need to dynamically incorporate new tasks and changing input distribution from users, the ability of IL on-device becomes essential for both efficiency and user privacy. However, prior works suffer from high computational costs and storage demands which hinders the deployment of IL on-device. In this work, to overcome these limitations, we develop an end-to-end and on-device IL framework, FastICARL, that incorporates an exemplar-based IL and quantization in the context of audio-based applications. We first employ k-nearest-neighbor to reduce the latency of IL. Then, we jointly utilize a quantization technique to decrease the storage requirements of IL. We implement FastICARL on two types of mobile devices and demonstrate that FastICARL remarkably decreases the IL time up to 78-92% and the storage requirements by 2-4 times without sacrificing its performance. FastICARL enables complete on-device IL, ensuring user privacy as the user data does not need to leave the device.

arXiv:2106.07214공식 코드 · Papers with Code

Backdoor Learning Curves: Explaining Backdoor Poisoning Beyond Influence Functions

Backdoor attacks inject poisoning samples during training, with the goal of forcing a machine learning model to output an attacker-chosen class when presented a specific trigger at test time. Although backdoor attacks have been demonstrated in a variety of settings and against different models, the factors affecting their effectiveness are still not well understood. In this work, we provide a unifying framework to study the process of backdoor learning under the lens of incremental learning and influence functions. We show that the effectiveness of backdoor attacks depends on: (i) the complexity of the learning algorithm, controlled by its hyperparameters; (ii) the fraction of backdoor samples injected into the training set; and (iii) the size and visibility of the backdoor trigger. These factors affect how fast a model learns to correlate the presence of the backdoor trigger with the target class. Our analysis unveils the intriguing existence of a region in the hyperparameter space in which the accuracy on clean test samples is still high while backdoor attacks are ineffective, thereby suggesting novel criteria to improve existing defenses.

arXiv:2106.06795

Knowledge Consolidation based Class Incremental Online Learning with Limited Data

We propose a novel approach for class incremental online learning in a limited data setting. This problem setting is challenging because of the following constraints: (1) Classes are given incrementally, which necessitates a class incremental learning approach; (2) Data for each class is given in an online fashion, i.e., each training example is seen only once during training; (3) Each class has very few training examples; and (4) We do not use or assume access to any replay/memory to store data from previous classes. Therefore, in this setting, we have to handle twofold problems of catastrophic forgetting and overfitting. In our approach, we learn robust representations that are generalizable across tasks without suffering from the problems of catastrophic forgetting and overfitting to accommodate future classes with limited samples. Our proposed method leverages the meta-learning framework with knowledge consolidation. The meta-learning framework helps the model for rapid learning when samples appear in an online fashion. Simultaneously, knowledge consolidation helps to learn a robust representation against forgetting under online updates to facilitate future learning. Our approach significantly outperforms other methods on several benchmarks.

arXiv:2106.05350

Generative Feature-driven Image Replay for Continual Learning

Neural networks are prone to catastrophic forgetting when trained incrementally on different tasks. Popular incremental learning methods mitigate such forgetting by retaining a subset of previously seen samples and replaying them during the training on subsequent tasks. However, this is not always possible, e.g., due to data protection regulations. In such restricted scenarios, one can employ generative models to replay either artificial images or hidden features to a classifier. In this work, we propose Genifer (GENeratIve FEature-driven image Replay), where a generative model is trained to replay images that must induce the same hidden features as real samples when they are passed through the classifier. Our technique therefore incorporates the benefits of both image and feature replay, i.e.: (1) unlike conventional image replay, our generative model explicitly learns the distribution of features that are relevant for classification; (2) in contrast to feature replay, our entire classifier remains trainable; and (3) we can leverage image-space augmentations, which increase distillation performance while also mitigating overfitting during the training of the generative model. We show that Genifer substantially outperforms the previous state of the art for various settings on the CIFAR-100 and CUB-200 datasets.

arXiv:2106.03343

Energy Aligning for Biased Models

Training on class-imbalanced data usually results in biased models that tend to predict samples into the majority classes, which is a common and notorious problem. From the perspective of energy-based model, we demonstrate that the free energies of categories are aligned with the label distribution theoretically, thus the energies of different classes are expected to be close to each other when aiming for ``balanced'' performance. However, we discover a severe energy-bias phenomenon in the models trained on class-imbalanced dataset. To eliminate the bias, we propose a simple and effective method named Energy Aligning by merely adding the calculated shift scalars onto the output logits during inference, which does not require to (i) modify the network architectures, (ii) intervene the standard learning paradigm, (iii) perform two-stage training. The proposed algorithm is evaluated on two class imbalance-related tasks under various settings: class incremental learning and long-tailed recognition. Experimental results show that energy aligning can effectively alleviate class imbalance issue and outperform state-of-the-art methods on several benchmarks.

논문 ID: 2020.10001

OpinionMine:ABayesian-based framework for opinion mining using Twitter Data

This article studies opinion mining from social media with probabilistic logic reasoning. As it is known, Twitter is one of the most active social networks, with millions of tweets sent daily, where multiple users express their opinion about traveling, economic issues, political decisions etc. As such, it offers a valuable source of information for opinion mining. In this paper we present OpinionMine, a Bayesian-based framework for opinion mining, exploiting Twitter Data. Initially, our framework imports Tweets massively by using Twitter’s API. Next, the imported Tweets are further processed automatically for constructing a set of untrained rules and random variables. Then, a Bayesian Network is derived by using the set of untrained rules, the random variables and an evidence set. After that, the trained model can be used for the evaluation of new Tweets. Finally, the constructed model can be retrained incrementally, thus becoming more robust. As application domain for the development of our methodology we have selected tourism because it is one of the most popular topics in social media. Our framework can predict users’ intention to visit a place. Among the advantages of our framework is that it follows an incremental learning strategy. That is, the derived model can be retrained incrementally with new training sets thus becoming more robust. Further, our framework can be easily adapted to opinion mining from social media on other topics, whereas the rules of the derived model are constructed in an efficient way and automatically.