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

TL;DR

This paper introduces FEAT, a federated continual learning method that uses geometry-aware corrections to improve exemplar replay performance amid dynamic heterogeneity across clients and tasks.

Contribution

FEAT employs geometric structure alignment and energy-based correction modules to address representation imbalance and drift in federated continual learning.

Findings

Aligns feature similarities with shared prototypes to maintain geometric consistency.

Reduces bias toward majority classes by removing task-irrelevant components.

Enhances minority class sensitivity and robustness in heterogeneous settings.

Abstract

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.