mlsanityData to 14 Oct 2026

From Selection to Scheduling: Federated Geometry-Aware Correction Makes Exemplar Replay Work Better under Continual Dynamic Heterogeneity

arXiv 2604.08617 ↗First posted 9 Apr 2026cs.LG cs.AI cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Exemplar replay has become an effective strategy for mitigating catastrophic forgetting in federated continual learning (FCL) by retaining representative samples from past tasks. Existing studies focus on designing sample-importance estimation mechanisms to identify information-rich samples. However, they typically overlook strategies for effectively utilizing the selected exemplars, which limits their performance under continual dynamic heterogeneity across clients and tasks. To address this issue, this paper proposes a Federated gEometry-Aware correcTion method, termed FEAT, which alleviates imbalance-induced representation collapse that drags rare-class features toward frequent classes across clients. Specifically, it consists of two key modules: 1) the Geometric Structure Alignment module performs structural knowledge distillation by aligning the pairwise angular similarities between feature representations and their corresponding Equiangular Tight Frame prototypes, which are fixed and shared across clients to serve as a class-discriminative reference structure. This encourages geometric consistency across tasks and helps mitigate representation drift; 2) the Energy-based Geometric Correction module removes task-irrelevant directional components from feature embeddings, which reduces prediction bias toward majority classes. This improves sensitivity to minority classes and enhances the model's robustness under class-imbalanced distributions.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 50 of its 80 references that are in the map. First posted on arXiv 9 Apr 2026.

Builds on

50 of 80 references.

References · 50

  1. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  2. 2016Learning without Forgettingcited by 2,496
  3. 2021Model-Contrastive Federated Learningcited by 766
  4. 2020Prevalence of Neural Collapse during the terminal phase of deep learning trainingcited by 719
  5. 2020Dynamical Variational Autoencoders: A Comprehensive Reviewcited by 128
  6. 2022Federated Class-Incremental Learningcited by 127
  7. 2023TARGET: Federated Class-Continual Learning via Exemplar-Free Distillationcited by 65
  8. 2023FedGH: Heterogeneous Federated Learning with Generalized Global Headercited by 56
  9. 2023Better Generative Replay for Continual Federated Learningcited by 55
  10. 2023No Fear of Classifier Biases: Neural Collapse Inspired Federated Learning with Synthetic and Fixed Classifiercited by 55
  11. 2023Federated Continual Learning via Knowledge Fusion: A Surveycited by 53
  12. 2020Variational Autoencoders for Learning Nonlinear Dynamics of Physical Systemscited by 50
38 more
  1. 2023Federated Incremental Semantic Segmentationcited by 36
  2. 2023Cross-Silo Prototypical Calibration for Federated Learning with Non-IID Datacited by 35
  3. 2024Towards Efficient Replay in Federated Incremental Learningcited by 33
  4. 2024Relaxed Contrastive Learning for Federated Learningcited by 32
  5. 2025Towards Universal Modal Tracking with Online Dense Temporal Token Learningcited by 30
  6. 2023Asynchronous Federated Continual Learningcited by 29
  7. 2023Few-shot Class-incremental Learning for Classification and Object Detection: A Surveycited by 25
  8. 2024Diffusion-Driven Data Replay: A Novel Approach to Combat Forgetting in Federated Class Continual Learningcited by 23
  9. 2024Unleashing the Power of Continual Learning on Non-Centralized Devices: A Surveycited by 23
  10. 2025Exploring Decoupled Spatio-Temporal Consistency Learning and Self-Prompting Evolution for Self-Supervised Trackingcited by 23
  11. 2022FedCross: Towards Accurate Federated Learning via Multi-Model Cross-Aggregationcited by 21
  12. 2024Federated Model Heterogeneous Matryoshka Representation Learningcited by 21
  13. 2023Exploiting Multi-Label Correlation in Label Distribution Learningcited by 20
  14. 2024Personalized Federated Continual Learning via Multi-granularity Promptcited by 20
  15. 2024Text-Enhanced Data-free Approach for Federated Class-Incremental Learningcited by 16
  16. 2024Learning Equi-angular Representations for Online Continual Learningcited by 15
  17. 2024Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligencecited by 15
  18. 2025FedAdamW: A Communication-Efficient Optimizer with Convergence and Generalization Guarantees for Federated Large Modelscited by 15
  19. 2023Ten Years of Generative Adversarial Nets (GANs): A survey of the state-of-the-artcited by 14
  20. 2023Is Aggregation the Only Choice? Federated Learning via Layer-wise Model Recombinationcited by 13
  21. 2024Federated Learning with Bilateral Curation for Partially Class-Disjoint Datacited by 13
  22. 2025Geometric Knowledge-Guided Localized Global Distribution Alignment for Federated Learningcited by 12
  23. 2021Generative Adversarial Networks (GANs) in Networking: A Comprehensive Survey & Evaluationcited by 11
  24. 2023Class-level Structural Relation Modelling and Smoothing for Visual Representation Learningcited by 9
  25. 2024Handling Spatial-Temporal Data Heterogeneity for Federated Continual Learning via Tail Anchorcited by 9
  26. 2025Federated Class-Incremental Learning: A Hybrid Approach Using Latent Exemplars and Data-Free Techniques to Address Local and Global Forgettingcited by 9
  27. 2025Class-wise Balancing Data Replay for Federated Class-Incremental Learningcited by 9
  28. 2025VT-FSL: Bridging Vision and Text with LLMs for Few-Shot Learningcited by 7
  29. 2026FedBCD:Communication-Efficient Accelerated Block Coordinate Gradient Descent for Federated Learningcited by 7
  30. 2023Don't Memorize; Mimic The Past: Federated Class Incremental Learning Without Episodic Memorycited by 4
  31. 2024Overcoming Catastrophic Forgetting in Federated Class-Incremental Learning via Federated Global Twin Generatorcited by 4
  32. 2024Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memorycited by 3
  33. 2025ProtoConNet: Prototypical Augmentation and Alignment for Open-Set Few-Shot Image Classificationcited by 3
  34. 2024Exemplar-condensed Federated Class-incremental Learningcited by 2
  35. 2025Enhancing Visual Representation with Textual Semantics: Textual Semantics-Powered Prototypes for Heterogeneous Federated Learningcited by 2
  36. 2025Empowering Vision Transformers with Multi-Scale Causal Intervention for Long-Tailed Image Classificationcited by 2
  37. 2025Global Prompt Refinement with Non-Interfering Attention Masking for One-Shot Federated Learningcited by 2
  38. 2026DVLA-RL: Dual-Level Vision-Language Alignment with Reinforcement Learning Gating for Few-Shot Learningcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Continual Learning with Elastic Regularization and Synthetic Replay for Federated MLLM Fine-Tuning
  2. 2026Do Vision Models Truly Forget? New Findings from Representation-Level Certification of Visual Unlearning in Vertical Federated Learning
  3. 2026PRISM: Exposing and Resolving Spurious Isolation in Federated Multimodal Continual Learning
  4. 2026Repurposing Adversarial Perturbations for Continual Learning: From Defense to Active Alignment
  5. 2025Federated Continual Instruction Tuning
  6. 2026MLFCIL: A Multi-Level Forgetting Mitigation Framework for Federated Class-Incremental Learning in LEO Satellites
  7. 2026Towards Adaptive Continual Model Merging via Manifold-Aware Expert Evolution
  8. 2026Exact Federated Continual Unlearning for Ridge Heads on Frozen Foundation Models