PyTorchKR Papers

최신 논문

총 2,786편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2301.12131

Restricted Orthogonal Gradient Projection for Continual Learning

Continual learning aims to avoid catastrophic forgetting and effectively leverage learned experiences to master new knowledge. Existing gradient projection approaches impose hard constraints on the optimization space for new tasks to minimize interference, which simultaneously hinders forward knowledge transfer. To address this issue, recent methods reuse frozen parameters with a growing network, resulting in high computational costs. Thus, it remains a challenge whether we can improve forward knowledge transfer for gradient projection approaches using a fixed network architecture. In this work, we propose the Restricted Orthogonal Gradient prOjection (ROGO) framework. The basic idea is to adopt a restricted orthogonal constraint allowing parameters optimized in the direction oblique to the whole frozen space to facilitate forward knowledge transfer while consolidating previous knowledge. Our framework requires neither data buffers nor extra parameters. Extensive experiments have demonstrated the superiority of our framework over several strong baselines. We also provide theoretical guarantees for our relaxing strategy.

arXiv:2301.12054

Adversarial Learning Networks: Source-free Unsupervised Domain Incremental Learning

This work presents an approach for incrementally updating deep neural network (DNN) models in a non-stationary environment. DNN models are sensitive to changes in input data distribution, which limits their application to problem settings with stationary input datasets. In a non-stationary environment, updating a DNN model requires parameter re-training or model fine-tuning. We propose an unsupervised source-free method to update DNN classification models. The contributions of this work are two-fold. First, we use trainable Gaussian prototypes to generate representative samples for future iterations; second, using unsupervised domain adaptation, we incrementally adapt the existing model using unlabelled data. Unlike existing methods, our approach can update a DNN model incrementally for non-stationary source and target tasks without storing past training data. We evaluated our work on incremental sentiment prediction and incremental disease prediction applications and compared our approach to state-of-the-art continual learning, domain adaptation, and ensemble learning methods. Our results show that our approach achieved improved performance compared to existing incremental learning methods. We observe minimal forgetting of past knowledge over many iterations, which can help us develop unsupervised self-learning systems.

arXiv:2301.12314공식 코드 · Papers with Code

Progressive Prompts: Continual Learning for Language Models

We introduce Progressive Prompts - a simple and efficient approach for continual learning in language models. Our method allows forward transfer and resists catastrophic forgetting, without relying on data replay or a large number of task-specific parameters. Progressive Prompts learns a new soft prompt for each task and sequentially concatenates it with the previously learned prompts, while keeping the base model frozen. Experiments on standard continual learning benchmarks show that our approach outperforms state-of-the-art methods, with an improvement >20% in average test accuracy over the previous best-preforming method on T5 model. We also explore a more challenging continual learning setup with longer sequences of tasks and show that Progressive Prompts significantly outperforms prior methods.

arXiv:2301.12230

Continual Graph Learning: A Survey

Research on continual learning (CL) mainly focuses on data represented in the Euclidean space, while research on graph-structured data is scarce. Furthermore, most graph learning models are tailored for static graphs. However, graphs usually evolve continually in the real world. Catastrophic forgetting also emerges in graph learning models when being trained incrementally. This leads to the need to develop robust, effective and efficient continual graph learning approaches. Continual graph learning (CGL) is an emerging area aiming to realize continual learning on graph-structured data. This survey is written to shed light on this emerging area. It introduces the basic concepts of CGL and highlights two unique challenges brought by graphs. Then it reviews and categorizes recent state-of-the-art approaches, analyzing their strategies to tackle the unique challenges in CGL. Besides, it discusses the main concerns in each family of CGL methods, offering potential solutions. Finally, it explores the open issues and potential applications of CGL.

arXiv:2301.12193

CyclicFL: A Cyclic Model Pre-Training Approach to Efficient Federated Learning

Federated learning (FL) has been proposed to enable distributed learning on Artificial Intelligence Internet of Things (AIoT) devices with guarantees of high-level data privacy. Since random initial models in FL can easily result in unregulated Stochastic Gradient Descent (SGD) processes, existing FL methods greatly suffer from both slow convergence and poor accuracy, especially in non-IID scenarios. To address this problem, we propose a novel method named CyclicFL, which can quickly derive effective initial models to guide the SGD processes, thus improving the overall FL training performance. We formally analyze the significance of data consistency between the pre-training and training stages of CyclicFL, showing the limited Lipschitzness of loss for the pre-trained models by CyclicFL. Moreover, we systematically prove that our method can achieve faster convergence speed under various convexity assumptions. Unlike traditional centralized pre-training methods that require public proxy data, CyclicFL pre-trains initial models on selected AIoT devices cyclically without exposing their local data. Therefore, they can be easily integrated into any security-critical FL methods. Comprehensive experimental results show that CyclicFL can not only improve the maximum classification accuracy by up to $14.11\%$ but also significantly accelerate the overall FL training process.

arXiv:2301.11396공식 코드 · Papers with Code

Class-Incremental Learning with Repetition

Real-world data streams naturally include the repetition of previous concepts. From a Continual Learning (CL) perspective, repetition is a property of the environment and, unlike replay, cannot be controlled by the agent. Nowadays, the Class-Incremental (CI) scenario represents the leading test-bed for assessing and comparing CL strategies. This scenario type is very easy to use, but it never allows revisiting previously seen classes, thus completely neglecting the role of repetition. We focus on the family of Class-Incremental with Repetition (CIR) scenario, where repetition is embedded in the definition of the stream. We propose two stochastic stream generators that produce a wide range of CIR streams starting from a single dataset and a few interpretable control parameters. We conduct the first comprehensive evaluation of repetition in CL by studying the behavior of existing CL strategies under different CIR streams. We then present a novel replay strategy that exploits repetition and counteracts the natural imbalance present in the stream. On both CIFAR100 and TinyImageNet, our strategy outperforms other replay approaches, which are not designed for environments with repetition.

arXiv:2301.11892

Streaming LifeLong Learning With Any-Time Inference

Despite rapid advancements in lifelong learning (LLL) research, a large body of research mainly focuses on improving the performance in the existing \textit{static} continual learning (CL) setups. These methods lack the ability to succeed in a rapidly changing \textit{dynamic} environment, where an AI agent needs to quickly learn new instances in a `single pass' from the non-i.i.d (also possibly temporally contiguous/coherent) data streams without suffering from catastrophic forgetting. For practical applicability, we propose a novel lifelong learning approach, which is streaming, i.e., a single input sample arrives in each time step, single pass, class-incremental, and subject to be evaluated at any moment. To address this challenging setup and various evaluation protocols, we propose a Bayesian framework, that enables fast parameter update, given a single training example, and enables any-time inference. We additionally propose an implicit regularizer in the form of snap-shot self-distillation, which effectively minimizes the forgetting further. We further propose an effective method that efficiently selects a subset of samples for online memory rehearsal and employs a new replay buffer management scheme that significantly boosts the overall performance. Our empirical evaluations and ablations demonstrate that the proposed method outperforms the prior works by large margins.

arXiv:2301.10931공식 코드 · Papers with Code

Towards Continual Egocentric Activity Recognition: A Multi-modal Egocentric Activity Dataset for Continual Learning

With the rapid development of wearable cameras, a massive collection of egocentric video for first-person visual perception becomes available. Using egocentric videos to predict first-person activity faces many challenges, including limited field of view, occlusions, and unstable motions. Observing that sensor data from wearable devices facilitates human activity recognition, multi-modal activity recognition is attracting increasing attention. However, the deficiency of related dataset hinders the development of multi-modal deep learning for egocentric activity recognition. Nowadays, deep learning in real world has led to a focus on continual learning that often suffers from catastrophic forgetting. But the catastrophic forgetting problem for egocentric activity recognition, especially in the context of multiple modalities, remains unexplored due to unavailability of dataset. In order to assist this research, we present a multi-modal egocentric activity dataset for continual learning named UESTC-MMEA-CL, which is collected by self-developed glasses integrating a first-person camera and wearable sensors. It contains synchronized data of videos, accelerometers, and gyroscopes, for 32 types of daily activities, performed by 10 participants. Its class types and scale are compared with other publicly available datasets. The statistical analysis of the sensor data is given to show the auxiliary effects for different behaviors. And results of egocentric activity recognition are reported when using separately, and jointly, three modalities: RGB, acceleration, and gyroscope, on a base network architecture. To explore the catastrophic forgetting in continual learning tasks, four baseline methods are extensively evaluated with different multi-modal combinations. We hope the UESTC-MMEA-CL can promote future studies on continual learning for first-person activity recognition in wearable applications.

arXiv:2301.09880

Probabilistic Bilevel Coreset Selection

The goal of coreset selection in supervised learning is to produce a weighted subset of data, so that training only on the subset achieves similar performance as training on the entire dataset. Existing methods achieved promising results in resource-constrained scenarios such as continual learning and streaming. However, most of the existing algorithms are limited to traditional machine learning models. A few algorithms that can handle large models adopt greedy search approaches due to the difficulty in solving the discrete subset selection problem, which is computationally costly when coreset becomes larger and often produces suboptimal results. In this work, for the first time we propose a continuous probabilistic bilevel formulation of coreset selection by learning a probablistic weight for each training sample. The overall objective is posed as a bilevel optimization problem, where 1) the inner loop samples coresets and train the model to convergence and 2) the outer loop updates the sample probability progressively according to the model's performance. Importantly, we develop an efficient solver to the bilevel optimization problem via unbiased policy gradient without trouble of implicit differentiation. We provide the convergence property of our training procedure and demonstrate the superiority of our algorithm against various coreset selection methods in various tasks, especially in more challenging label-noise and class-imbalance scenarios.

arXiv:2301.08986공식 코드 · Papers with Code

Adapting a Language Model While Preserving its General Knowledge

Domain-adaptive pre-training (or DA-training for short), also known as post-training, aims to train a pre-trained general-purpose language model (LM) using an unlabeled corpus of a particular domain to adapt the LM so that end-tasks in the domain can give improved performances. However, existing DA-training methods are in some sense blind as they do not explicitly identify what knowledge in the LM should be preserved and what should be changed by the domain corpus. This paper shows that the existing methods are suboptimal and proposes a novel method to perform a more informed adaptation of the knowledge in the LM by (1) soft-masking the attention heads based on their importance to best preserve the general knowledge in the LM and (2) contrasting the representations of the general and the full (both general and domain knowledge) to learn an integrated representation with both general and domain-specific knowledge. Experimental results will demonstrate the effectiveness of the proposed approach.

논문 ID: document/9506788

Class incremental learning for video action classification

Class Incremental Learning (CIL) is a hot topic in machine learning for CNN models to learn new classes incrementally. However, most of the CIL studies are for image classification and object recognition tasks and few CIL studies are available for video action classification. To mitigate this problem, in this paper, we present a new Grow When Required network (GWR) based video CIL framework for action classification. GWR learns knowledge incrementally by modeling the manifold of video frames for each encountered action class in feature space. We also introduce a Knowledge Consolidation (KC) method to separate the feature manifolds of old class and new class and introduce an associative matrix for label prediction. Experimental results on KTH and Weizmann demonstrate the effectiveness of the framework.

arXiv:2301.07187공식 코드 · Papers with Code

Artificial Neuronal Ensembles with Learned Context Dependent Gating

Biological neural networks are capable of recruiting different sets of neurons to encode different memories. However, when training artificial neural networks on a set of tasks, typically, no mechanism is employed for selectively producing anything analogous to these neuronal ensembles. Further, artificial neural networks suffer from catastrophic forgetting, where the network's performance rapidly deteriorates as tasks are learned sequentially. By contrast, sequential learning is possible for a range of biological organisms. We introduce Learned Context Dependent Gating (LXDG), a method to flexibly allocate and recall `artificial neuronal ensembles', using a particular network structure and a new set of regularization terms. Activities in the hidden layers of the network are modulated by gates, which are dynamically produced during training. The gates are outputs of networks themselves, trained with a sigmoid output activation. The regularization terms we have introduced correspond to properties exhibited by biological neuronal ensembles. The first term penalizes low gate sparsity, ensuring that only a specified fraction of the network is used. The second term ensures that previously learned gates are recalled when the network is presented with input from previously learned tasks. Finally, there is a regularization term responsible for ensuring that new tasks are encoded in gates that are as orthogonal as possible from previously used ones. We demonstrate the ability of this method to alleviate catastrophic forgetting on continual learning benchmarks. When the new regularization terms are included in the model along with Elastic Weight Consolidation (EWC) it achieves better performance on the benchmark `permuted MNIST' than with EWC alone. The benchmark `rotated MNIST' demonstrates how similar tasks recruit similar neurons to the artificial neuronal ensemble.

arXiv:2301.07316

Adaptively Integrated Knowledge Distillation and Prediction Uncertainty for Continual Learning

Current deep learning models often suffer from catastrophic forgetting of old knowledge when continually learning new knowledge. Existing strategies to alleviate this issue often fix the trade-off between keeping old knowledge (stability) and learning new knowledge (plasticity). However, the stability-plasticity trade-off during continual learning may need to be dynamically changed for better model performance. In this paper, we propose two novel ways to adaptively balance model stability and plasticity. The first one is to adaptively integrate multiple levels of old knowledge and transfer it to each block level in the new model. The second one uses prediction uncertainty of old knowledge to naturally tune the importance of learning new knowledge during model training. To our best knowledge, this is the first time to connect model prediction uncertainty and knowledge distillation for continual learning. In addition, this paper applies a modified CutMix particularly to augment the data for old knowledge, further alleviating the catastrophic forgetting issue. Extensive evaluations on the CIFAR100 and the ImageNet datasets confirmed the effectiveness of the proposed method for continual learning.

arXiv:2301.06834

Knowledge Acquisition and Completion for Long-Term Human-Robot Interactions using Knowledge Graph Embedding

In Human-Robot Interaction (HRI) systems, a challenging task is sharing the representation of the operational environment, fusing symbolic knowledge and perceptions, between users and robots. With the existing HRI pipelines, users can teach the robots some concepts to increase their knowledge base. Unfortunately, the data coming from the users are usually not enough dense for building a consistent representation. Furthermore, the existing approaches are not able to incrementally build up their knowledge base, which is very important when robots have to deal with dynamic contexts. To this end, we propose an architecture to gather data from users and environments in long-runs of continual learning. We adopt Knowledge Graph Embedding techniques to generalize the acquired information with the goal of incrementally extending the robot's inner representation of the environment. We evaluate the performance of the overall continual learning architecture by measuring the capabilities of the robot of learning entities and relations coming from unknown contexts through a series of incremental learning sessions.

arXiv:2301.06030

Self-recovery of memory via generative replay

A remarkable capacity of the brain is its ability to autonomously reorganize memories during offline periods. Memory replay, a mechanism hypothesized to underlie biological offline learning, has inspired offline methods for reducing forgetting in artificial neural networks in continual learning settings. A memory-efficient and neurally-plausible method is generative replay, which achieves state of the art performance on continual learning benchmarks. However, unlike the brain, standard generative replay does not self-reorganize memories when trained offline on its own replay samples. We propose a novel architecture that augments generative replay with an adaptive, brain-like capacity to autonomously recover memories. We demonstrate this capacity of the architecture across several continual learning tasks and environments.

arXiv:2301.05058공식 코드 · Papers with Code

Sparse Coding in a Dual Memory System for Lifelong Learning

Efficient continual learning in humans is enabled by a rich set of neurophysiological mechanisms and interactions between multiple memory systems. The brain efficiently encodes information in non-overlapping sparse codes, which facilitates the learning of new associations faster with controlled interference with previous associations. To mimic sparse coding in DNNs, we enforce activation sparsity along with a dropout mechanism which encourages the model to activate similar units for semantically similar inputs and have less overlap with activation patterns of semantically dissimilar inputs. This provides us with an efficient mechanism for balancing the reusability and interference of features, depending on the similarity of classes across tasks. Furthermore, we employ sparse coding in a multiple-memory replay mechanism. Our method maintains an additional long-term semantic memory that aggregates and consolidates information encoded in the synaptic weights of the working model. Our extensive evaluation and characteristics analysis show that equipped with these biologically inspired mechanisms, the model can further mitigate forgetting.

arXiv:2301.04783공식 코드 · Papers with Code

Predictive World Models from Real-World Partial Observations

Cognitive scientists believe adaptable intelligent agents like humans perform reasoning through learned causal mental simulations of agents and environments. The problem of learning such simulations is called predictive world modeling. Recently, reinforcement learning (RL) agents leveraging world models have achieved SOTA performance in game environments. However, understanding how to apply the world modeling approach in complex real-world environments relevant to mobile robots remains an open question. In this paper, we present a framework for learning a probabilistic predictive world model for real-world road environments. We implement the model using a hierarchical VAE (HVAE) capable of predicting a diverse set of fully observed plausible worlds from accumulated sensor observations. While prior HVAE methods require complete states as ground truth for learning, we present a novel sequential training method to allow HVAEs to learn to predict complete states from partially observed states only. We experimentally demonstrate accurate spatial structure prediction of deterministic regions achieving 96.21 IoU, and close the gap to perfect prediction by 62% for stochastic regions using the best prediction. By extending HVAEs to cases where complete ground truth states do not exist, we facilitate continual learning of spatial prediction as a step towards realizing explainable and comprehensive predictive world models for real-world mobile robotics applications. Code is available at https://github.com/robin-karlsson0/predictive-world-models.

arXiv:2301.04584

Continual HyperTransformer: A Meta-Learner for Continual Few-Shot Learning

We focus on the problem of learning without forgetting from multiple tasks arriving sequentially, where each task is defined using a few-shot episode of novel or already seen classes. We approach this problem using the recently published HyperTransformer (HT), a Transformer-based hypernetwork that generates specialized task-specific CNN weights directly from the support set. In order to learn from a continual sequence of tasks, we propose to recursively re-use the generated weights as input to the HT for the next task. This way, the generated CNN weights themselves act as a representation of previously learned tasks, and the HT is trained to update these weights so that the new task can be learned without forgetting past tasks. This approach is different from most continual learning algorithms that typically rely on using replay buffers, weight regularization or task-dependent architectural changes. We demonstrate that our proposed Continual HyperTransformer method equipped with a prototypical loss is capable of learning and retaining knowledge about past tasks for a variety of scenarios, including learning from mini-batches, and task-incremental and class-incremental learning scenarios.

arXiv:2301.04506

A Distinct Unsupervised Reference Model From The Environment Helps Continual Learning

The existing continual learning methods are mainly focused on fully-supervised scenarios and are still not able to take advantage of unlabeled data available in the environment. Some recent works tried to investigate semi-supervised continual learning (SSCL) settings in which the unlabeled data are available, but it is only from the same distribution as the labeled data. This assumption is still not general enough for real-world applications and restricts the utilization of unsupervised data. In this work, we introduce Open-Set Semi-Supervised Continual Learning (OSSCL), a more realistic semi-supervised continual learning setting in which out-of-distribution (OoD) unlabeled samples in the environment are assumed to coexist with the in-distribution ones. Under this configuration, we present a model with two distinct parts: (i) the reference network captures general-purpose and task-agnostic knowledge in the environment by using a broad spectrum of unlabeled samples, (ii) the learner network is designed to learn task-specific representations by exploiting supervised samples. The reference model both provides a pivotal representation space and also segregates unlabeled data to exploit them more efficiently. By performing a diverse range of experiments, we show the superior performance of our model compared with other competitors and prove the effectiveness of each component of the proposed model.

arXiv:2301.03345공식 코드 · Papers with Code

Latent Spectral Regularization for Continual Learning

While biological intelligence grows organically as new knowledge is gathered throughout life, Artificial Neural Networks forget catastrophically whenever they face a changing training data distribution. Rehearsal-based Continual Learning (CL) approaches have been established as a versatile and reliable solution to overcome this limitation; however, sudden input disruptions and memory constraints are known to alter the consistency of their predictions. We study this phenomenon by investigating the geometric characteristics of the learner's latent space and find that replayed data points of different classes increasingly mix up, interfering with classification. Hence, we propose a geometric regularizer that enforces weak requirements on the Laplacian spectrum of the latent space, promoting a partitioning behavior. Our proposal, called Continual Spectral Regularizer for Incremental Learning (CaSpeR-IL), can be easily combined with any rehearsal-based CL approach and improves the performance of SOTA methods on standard benchmarks.