Revisiting Prototype Rehearsal for Exemplar-Free Continual Learning: Manifold-Aware Boundary Sampling with Adaptive Class-Balanced Loss

• URL: http://arxiv.org/abs/2606.05695v1
• Date: Thu, 04 Jun 2026 04:27:01 GMT
• Title: Revisiting Prototype Rehearsal for Exemplar-Free Continual Learning: Manifold-Aware Boundary Sampling with Adaptive Class-Balanced Loss
• Abstract summary: Exemplar-free class-incremental learning (EFCIL) aims to acquire new classes over time without storing raw data.<n>Historically, prototype rehearsal has been a popular strategy to reduce catastrophic forgetting.<n>Recent drift-compensation methods that explicitly realign prototypes in the evolving feature space consistently outperform prototype-based rehearsal.
• Score: 6.63181149575886
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Exemplar-free class-incremental learning (EFCIL) aims to acquire new classes over time without storing raw data. Historically, prototype rehearsal, which samples around stored class prototypes and mixes them with current-task data, has been a popular strategy to reduce catastrophic forgetting. However, recent drift-compensation methods that explicitly realign prototypes in the evolving feature space consistently outperform prototype-based rehearsal, raising the question of whether rehearsal itself is fundamentally limited. We argue that the performance gap stems not from the idea of prototype rehearsal per se, but from how it is typically instantiated: existing approaches treat prototypes as isolated class summaries that ignore information from nearby enemy classes, and fail to correct the emerging class imbalance between a handful of synthetic old-class samples and hundreds of real instances from newly introduced classes. Building on this hypothesis, we revisit prototype rehearsal and propose a manifold-aware variant that restores its competitiveness in EFCIL. First, we introduce Constrained Expansive Over-Sampling, which interpolates each old-class prototype toward its nearest enemy features from new classes, generating boundary-aware rehearsal samples that better follow the underlying data manifold while preserving inter-class separation. Second, we design an Adaptive Class-Balanced loss that performs time-based class weighting, amplifying gradients from older prototypes when they are most informative and gradually annealing their influence as richer supervision from later tasks accumulates. Together, these components turn prototype rehearsal into a drift-resilient, imbalance-aware mechanism that closes, and often reverses, the gap to recent drift-compensation methods, achieving state-of-the-art performance across multiple EFCIL benchmarks.

Related papers

• Contrastive Representation-Guided Genetic Minority Oversampling for Imbalanced Time-Series Classification [56.54664402339568]
  Real-world time-series classification tasks often exhibit class imbalance, which can be extremely severe in some applications.<n>This paper proposes a Frequency-domain representation-guided Multi-tree Genetic Programming-based oversampling approach to imbalanced time-series classification.
  arXiv  Detail & Related papers  (2026-08-24T05:03:51Z)
• Unlocking Prototype Potential: An Efficient Tuning Framework for Few-Shot Class-Incremental Learning [69.28860905525057]
  Few-shot class-incremental learning (FSCIL) seeks to continuously learn new classes from very limited samples.<n>We introduce an efficient prototype fine-tuning framework that evolves static centroids into dynamic, learnable components.
  arXiv  Detail & Related papers  (2026-02-05T03:50:53Z)
• MoTiC: Momentum Tightness and Contrast for Few-Shot Class-Incremental Learning [20.22260611239967]
  Few-Shot Class-Incremental Learning (FSCIL) must contend with the dual challenge of learning new classes from scarce samples while preserving old class knowledge.<n>New-class prototypes suffer significant estimation bias due to extreme data scarcity, whereas base-class prototypes benefit from sufficient data.<n>We show that aligning the new-class priors with old-class statistics via Bayesian analysis reduces variance and improves prototype accuracy.
  arXiv  Detail & Related papers  (2025-09-24T00:41:40Z)
• Resurrecting Old Classes with New Data for Exemplar-Free Continual Learning [13.264972882846966]
  Continual learning methods are known to suffer from catastrophic forgetting. Existing exemplar-free methods are typically evaluated in settings where the first task is significantly larger than subsequent tasks. We propose to adversarially perturb the current samples such that their embeddings are close to the old class prototypes in the old model embedding space. We then estimate the drift in the embedding space from the old to the new model using the perturbed images and compensate the prototypes accordingly.
  arXiv  Detail & Related papers  (2024-05-29T13:31:42Z)
• Rethinking Classifier Re-Training in Long-Tailed Recognition: A Simple Logits Retargeting Approach [102.0769560460338]
  We develop a simple logits approach (LORT) without the requirement of prior knowledge of the number of samples per class. Our method achieves state-of-the-art performance on various imbalanced datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist 2018.
  arXiv  Detail & Related papers  (2024-03-01T03:27:08Z)
• Decoupled Prototype Learning for Reliable Test-Time Adaptation [50.779896759106784]
  Test-time adaptation (TTA) is a task that continually adapts a pre-trained source model to the target domain during inference. One popular approach involves fine-tuning model with cross-entropy loss according to estimated pseudo-labels. This study reveals that minimizing the classification error of each sample causes the cross-entropy loss's vulnerability to label noise. We propose a novel Decoupled Prototype Learning (DPL) method that features prototype-centric loss computation.
  arXiv  Detail & Related papers  (2024-01-15T03:33:39Z)
• RanPAC: Random Projections and Pre-trained Models for Continual Learning [59.07316955610658]
  Continual learning (CL) aims to learn different tasks (such as classification) in a non-stationary data stream without forgetting old ones. We propose a concise and effective approach for CL with pre-trained models.
  arXiv  Detail & Related papers  (2023-07-05T12:49:02Z)
• Prototype-Sample Relation Distillation: Towards Replay-Free Continual Learning [14.462797749666992]
  We propose a holistic approach to jointly learn the representation and class prototypes. 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.
  arXiv  Detail & Related papers  (2023-03-26T16:35:45Z)
• Incremental Prototype Prompt-tuning with Pre-trained Representation for Class Incremental Learning [4.717066668969749]
  Class incremental learning has attracted much attention, but most existing works still continually fine-tune the representation model. We take the pre-train-and-prompt-tuning paradigm to sequentially learn new visual concepts based on a fixed semantic rich pre-trained representation model. Our method consistently outperforms other state-of-the-art methods with a large margin.
  arXiv  Detail & Related papers  (2022-04-07T12:49:14Z)
• Contrastive Prototype Learning with Augmented Embeddings for Few-Shot Learning [58.2091760793799]
  We propose a novel contrastive prototype learning with augmented embeddings (CPLAE) model. With a class prototype as an anchor, CPL aims to pull the query samples of the same class closer and those of different classes further away. Extensive experiments on several benchmarks demonstrate that our proposed CPLAE achieves new state-of-the-art.
  arXiv  Detail & Related papers  (2021-01-23T13:22:44Z)
• Few-shot Action Recognition with Prototype-centered Attentive Learning [88.10852114988829]
  Prototype-centered Attentive Learning (PAL) model composed of two novel components. First, a prototype-centered contrastive learning loss is introduced to complement the conventional query-centered learning objective. Second, PAL integrates a attentive hybrid learning mechanism that can minimize the negative impacts of outliers.
  arXiv  Detail & Related papers  (2021-01-20T11:48:12Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.