Continual Learning

In Continual learning (CL) balancing effective adaptation while combating catastrophic forgetting is a central challenge. Many of the recent best-performing methods utilize various forms of prior task data, e.g. a replay buffer, to tackle the catastrophic forgetting problem. Having access to previous task data can be restrictive in many real-world scenarios, for example when task data is sensitive or proprietary. To overcome the necessity of using previous tasks' data, in this work, we start with strong representation learning methods that have been shown to be less prone to forgetting. We propose a holistic approach to jointly learn the representation and class prototypes while maintaining the relevance of old class prototypes and their embedded similarities. Specifically, samples are mapped to an embedding space where the representations are learned using a supervised contrastive loss. Class prototypes are evolved continually in the same latent space, enabling learning and prediction at any point. To continually adapt the prototypes without keeping any prior task data, we propose a novel distillation loss that constrains class prototypes to maintain relative similarities as compared to new task data. This method yields state-of-the-art performance in the task-incremental setting, outperforming methods relying on large amounts of data, and provides strong performance in the class-incremental setting without using any stored data points.

Motivated by the efficiency and rapid convergence of pre-trained models for solving downstream tasks, this paper extensively studies the impact of Continual Learning (CL) models as pre-trainers. We find that, in both supervised and unsupervised CL, the transfer quality of representations does not show a noticeable degradation of fine-tuning performance but rather increases gradually. This is because CL models can learn improved task-general features when easily forgetting task-specific knowledge. Based on this observation, we suggest a new unsupervised CL framework with masked modeling, which aims to capture fluent task-generic representation during training. Furthermore, we propose a new fine-tuning scheme, GLobal Attention Discretization (GLAD), that preserves rich task-generic representation during solving downstream tasks. The model fine-tuned with GLAD achieves competitive performance and can also be used as a good pre-trained model itself. We believe this paper breaks the barriers between pre-training and fine-tuning steps and leads to a sustainable learning framework in which the continual learner incrementally improves model generalization, yielding better transfer to unseen tasks.

Existing research on task incremental learning in continual learning has primarily focused on preventing catastrophic forgetting (CF). Although several techniques have achieved learning with no CF, they attain it by letting each task monopolize a sub-network in a shared network, which seriously limits knowledge transfer (KT) and causes over-consumption of the network capacity, i.e., as more tasks are learned, the performance deteriorates. The goal of this paper is threefold: (1) overcoming CF, (2) encouraging KT, and (3) tackling the capacity problem. A novel technique (called SPG) is proposed that soft-masks (partially blocks) parameter updating in training based on the importance of each parameter to old tasks. Each task still uses the full network, i.e., no monopoly of any part of the network by any task, which enables maximum KT and reduction in capacity usage. To our knowledge, this is the first work that soft-masks a model at the parameter-level for continual learning. Extensive experiments demonstrate the effectiveness of SPG in achieving all three objectives. More notably, it attains significant transfer of knowledge not only among similar tasks (with shared knowledge) but also among dissimilar tasks (with little shared knowledge) while mitigating CF.

Continual learning (CL), which aims to learn a sequence of tasks, has attracted significant recent attention. However, most work has focused on the experimental performance of CL, and theoretical studies of CL are still limited. In particular, there is a lack of understanding on what factors are important and how they affect "catastrophic forgetting" and generalization performance. To fill this gap, our theoretical analysis, under overparameterized linear models, provides the first-known explicit form of the expected forgetting and generalization error for a general CL setup with an arbitrary number of tasks. Further analysis of such a key result yields a number of theoretical explanations about how overparameterization, task similarity, and task ordering affect both forgetting and generalization error of CL. More interestingly, by conducting experiments on real datasets using deep neural networks (DNNs), we show that some of these insights even go beyond the linear models and can be carried over to practical setups. In particular, we use concrete examples to show that our results not only explain some interesting empirical observations in recent studies, but also motivate better practical algorithm designs of CL.

This paper focuses on continual meta-learning, where few-shot tasks are heterogeneous and sequentially available. Recent works use a mixture model for meta-knowledge to deal with the heterogeneity. However, these methods suffer from parameter inefficiency caused by two reasons: (1) the underlying assumption of mutual exclusiveness among mixture components hinders sharing meta-knowledge across heterogeneous tasks. (2) they only allow increasing mixture components and cannot adaptively filter out redundant components. In this paper, we propose an Adaptive Compositional Continual Meta-Learning (ACML) algorithm, which employs a compositional premise to associate a task with a subset of mixture components, allowing meta-knowledge sharing among heterogeneous tasks. Moreover, to adaptively adjust the number of mixture components, we propose a component sparsification method based on evidential theory to filter out redundant components. Experimental results show ACML outperforms strong baselines, showing the effectiveness of our compositional meta-knowledge, and confirming that ACML can adaptively learn meta-knowledge.

The goal of continual learning is to find a model that solves multiple learning tasks which are presented sequentially to the learner. A key challenge in this setting is that the learner may "forget" how to solve a previous task when learning a new task, a phenomenon known as catastrophic forgetting. To address this challenge, many practical methods have been proposed, including memory-based, regularization-based and expansion-based methods. However, a rigorous theoretical understanding of these methods remains elusive. This paper aims to bridge this gap between theory and practice by proposing a new continual learning framework called "Ideal Continual Learner" (ICL), which is guaranteed to avoid catastrophic forgetting by construction. We show that ICL unifies multiple well-established continual learning methods and gives new theoretical insights into the strengths and weaknesses of these methods. We also derive generalization bounds for ICL which allow us to theoretically quantify "how rehearsal affects generalization". Finally, we connect ICL to several classic subjects and research topics of modern interest, which allows us to make historical remarks and inspire future directions.

Generative models have grown into the workhorse of many state-of-the-art machine learning methods. However, their vulnerability under poisoning attacks has been largely understudied. In this work, we investigate this issue in the context of continual learning, where generative replayers are utilized to tackle catastrophic forgetting. By developing a novel customization of dirty-label input-aware backdoors to the online setting, our attacker manages to stealthily promote forgetting while retaining high accuracy at the current task and sustaining strong defenders. Our approach taps into an intriguing property of generative models, namely that they cannot well capture input-dependent triggers. Experiments on four standard datasets corroborate the poisoner's effectiveness.

We analyze continual learning on a sequence of separable linear classification tasks with binary labels. We show theoretically that learning with weak regularization reduces to solving a sequential max-margin problem, corresponding to a special case of the Projection Onto Convex Sets (POCS) framework. We then develop upper bounds on the forgetting and other quantities of interest under various settings with recurring tasks, including cyclic and random orderings of tasks. We discuss several practical implications to popular training practices like regularization scheduling and weighting. We point out several theoretical differences between our continual classification setting and a recently studied continual regression setting.

Large-scale multi-modal contrastive learning frameworks like CLIP typically require a large amount of image-text samples for training. However, these samples are always collected continuously in real scenarios. This paper discusses the feasibility of continual CLIP training using streaming data. Unlike continual learning based on self-supervised learning methods for pure images, which is empirically robust against catastrophic forgetting, CLIP's performance degeneration in the continual setting is significant and non-neglectable. By analyzing the changes in the model's representation space during continual CLIP training from a spatial geometry perspective, we explore and summarize these spatial variations as **Spatial Disorder (SD)**, which can be divided into **Intra-modal Rotation** and **Inter-modal Deviation**. Moreover, we empirically and theoretically demonstrate how SD leads to a performance decline for CLIP on cross-modal retrieval tasks. To alleviate SD, we propose a new continual vision-language representation learning framework **Mod-X**: **M**aintain **o**ff-**d**iagonal information-matri**X**. By selectively aligning the off-diagonal information distribution of contrastive matrices, the Mod-X improves the capability of the multi-modal model by maintaining the multi-modal representation space alignment on the old data domain during continuously fitting the new training data domain. Experiments on commonly used datasets with different scales and scopes have demonstrated the effectiveness of our method.

We study fine-grained error bounds for differentially private algorithms for counting under continual observation. Our main insight is that the matrix mechanism when using lower-triangular matrices can be used in the continual observation model. More specifically, we give an explicit factorization for the counting matrix $M_\mathsf{count}$ and upper bound the error explicitly. We also give a fine-grained analysis, specifying the exact constant in the upper bound. Our analysis is based on upper and lower bounds of the *completely bounded norm* (cb-norm) of $M_\mathsf{count}$. Along the way, we improve the best-known bound of 28 years by Mathias (SIAM Journal on Matrix Analysis and Applications, 1993) on the cb-norm of $M_\mathsf{count}$ for a large range of the dimension of $M_\mathsf{count}$. Furthermore, we are the first to give concrete error bounds for various problems under continual observation such as binary counting, maintaining a histogram, releasing an approximately cut-preserving synthetic graph, many graph-based statistics, and substring and episode counting. Finally, we note that our result can be used to get a fine-grained error bound for non-interactive local learning and the first lower bounds on the additive error for $(\epsilon,\delta)$-differentially-private counting under continual observation. Subsequent to this work, Henzinger et al. (SODA, 2023) showed that our factorization also achieves fine-grained mean-squared error.

Popular deep-learning models in the field of image classification suffer from catastrophic forgetting---models will forget previously acquired skills when learning new ones. Generative replay (GR), which typically consists of a generator and a classifier, is an efficient way to mitigate catastrophic forgetting. However, conventional GR methods only focus on a single instruction relationship (generator-to-classifier), where the generator synthesizes samples for previous tasks to instruct the training of the classifier, while ignoring the ways in which the classifier can benefit the generator. In addition, most generative replay methods typically reuse the generated samples to update the generator, which causes the samples regenerated by the generator deviating from the distribution of previous tasks. To overcome these two issues, we propose a novel approach, called deep diffusion-based generative replay (DDGR), which adopts a diffusion model as the generator and calculates an instruction-operator through the classifier to instruct the generation of samples. Extensive experiments in class incremental (CI) and class incremental with repetition (CIR) settings demonstrate the advantages of DDGR. Our code is available at https://github.com/xiaocangshengGR/DDGR.

Deep neural networks perform well on classification tasks where data streams are i.i.d. and labeled data is abundant. Challenges emerge with non-stationary training data streams such as continual learning. One powerful approach that has addressed this challenge involves pre-training of large encoders on volumes of readily available data, followed by task-specific tuning. Given a new task, however, updating the weights of these encoders is challenging as a large number of weights needs to be fine-tuned, and as a result, they forget information about the previous tasks. In the present work, we propose a model architecture to address this issue, building upon a discrete bottleneck containing pairs of separate and learnable key-value codes. Our paradigm will be to encode; process the representation via a discrete bottleneck; and decode. Here, the input is fed to the pre-trained encoder, the output of the encoder is used to select the nearest keys, and the corresponding values are fed to the decoder to solve the current task. The model can only fetch and re-use a sparse number of these key-value pairs during inference, enabling *localized and context-dependent model updates*. We theoretically investigate the ability of the discrete key-value bottleneck to minimize the effect of learning under distribution shifts and show that it reduces the complexity of the hypothesis class. We empirically verify the proposed method under challenging class-incremental learning scenarios and show that the proposed model --- without any task boundaries --- reduces catastrophic forgetting across a wide variety of pre-trained models, outperforming relevant baselines on this task.

Current fake audio detection algorithms have achieved promising performances on most datasets. However, their performance may be significantly degraded when dealing with audio of a different dataset. The orthogonal weight modification to overcome catastrophic forgetting does not consider the similarity of genuine audio across different datasets. To overcome this limitation, we propose a continual learning algorithm for fake audio detection to overcome catastrophic forgetting, called Regularized Adaptive Weight Modification (RAWM). When fine-tuning a detection network, our approach adaptively computes the direction of weight modification according to the ratio of genuine utterances and fake utterances. The adaptive modification direction ensures the network can effectively detect fake audio on the new dataset while preserving its knowledge of old model, thus mitigating catastrophic forgetting. In addition, genuine audio collected from quite different acoustic conditions may skew their feature distribution, so we introduce a regularization constraint to force the network to remember the old distribution in this regard. Our method can easily be generalized to related fields, like speech emotion recognition. We also evaluate our approach across multiple datasets and obtain a significant performance improvement on cross-dataset experiments.

Rehearsal-based approaches are a mainstay of continual learning (CL). They mitigate the catastrophic forgetting problem by maintaining a small fixed-size buffer with a subset of data from past tasks. While most rehearsal-based approaches exploit the knowledge from buffered past data, little attention is paid to inter-task relationships and to critical task-specific and task-invariant knowledge. By appropriately leveraging inter-task relationships, we propose a novel CL method, named DualHSIC, to boost the performance of existing rehearsal-based methods in a simple yet effective way. DualHSIC consists of two complementary components that stem from the so-called Hilbert Schmidt independence criterion (HSIC): HSIC-Bottleneck for Rehearsal (HBR) lessens the inter-task interference and HSIC Alignment (HA) promotes task-invariant knowledge sharing. Extensive experiments show that DualHSIC can be seamlessly plugged into existing rehearsal-based methods for consistent performance improvements, outperforming recent state-of-the-art regularization-enhanced rehearsal methods.

Distribution shift (e.g., task or domain shift) in continual learning (CL) usually results in catastrophic forgetting of previously learned knowledge. Although it can be alleviated by repeatedly replaying buffered data, the every-step replay is time-consuming. In this paper, we study which modules in neural networks are more prone to forgetting by investigating their training dynamics during CL. Our proposed metrics show that only a few modules are more task-specific and sensitive to task change, while others can be shared across tasks as common knowledge. Hence, we attribute forgetting mainly to the former and find that finetuning them only on a small buffer at the end of any CL method can bring non-trivial improvement. Due to the small number of finetuned parameters, such ''Forgetting Prioritized Finetuning (FPF)'' is efficient in computation. We further propose a more efficient and simpler method that entirely removes the every-step replay and replaces them by only $k$-times of FPF periodically triggered during CL. Surprisingly, this ''$k$-FPF'' performs comparably to FPF and outperforms the SOTA CL methods but significantly reduces their computational overhead and cost. In experiments on several benchmarks of class- and domain-incremental CL, FPF consistently improves existing CL methods by a large margin, and $k$-FPF further excels in efficiency without degrading the accuracy. We also empirically studied the impact of buffer size, epochs per task, and finetuning modules on the cost and accuracy of our methods.

The ability of deep neural networks to continually learn and adapt to a sequence of tasks has remained challenging due to catastrophic forgetting of previously learned tasks. Humans, on the other hand, have a remarkable ability to acquire, assimilate, and transfer knowledge across tasks throughout their lifetime without catastrophic forgetting. The versatility of the brain can be attributed to the rehearsal of abstract experiences through a complementary learning system. However, representation rehearsal in vision transformers lacks diversity, resulting in overfitting and consequently, performance drops significantly compared to raw image rehearsal. Therefore, we propose BiRT, a novel representation rehearsal-based continual learning approach using vision transformers. Specifically, we introduce controllable noises at various stages of the vision transformer and enforce consistency in predictions with respect to an exponential moving average of the working model. Our method provides consistent performance gain over raw image and vanilla representation rehearsal on several challenging CL benchmarks while being memory efficient and robust to natural and adversarial corruptions.