AFCL: Analytic Federated Continual Learning for Spatio-Temporal Invariance of Non-IID Data

TL;DR

This paper introduces AFCL, a gradient-free federated continual learning method that achieves spatio-temporal invariance for non-IID data, effectively mitigating catastrophic forgetting in dynamic, heterogeneous environments.

Contribution

AFCL provides a novel analytical approach that enables single-epoch local training and efficient global aggregation, addressing data heterogeneity challenges in federated continual learning.

Findings

AFCL outperforms state-of-the-art methods on multiple benchmarks.

Theoretical analysis confirms spatio-temporal invariance of the global model.

AFCL reduces training complexity with single-round aggregation.

Abstract

Federated Continual Learning (FCL) enables distributed clients to collaboratively train a global model from online task streams in dynamic real-world scenarios. However, existing FCL methods face challenges of both spatial data heterogeneity among distributed clients and temporal data heterogeneity across online tasks. Such data heterogeneity significantly degrades the model performance with severe spatial-temporal catastrophic forgetting of local and past knowledge. In this paper, we identify that the root cause of this issue lies in the inherent vulnerability and sensitivity of gradients to non-IID data. To fundamentally address this issue, we propose a gradient-free method, named Analytic Federated Continual Learning (AFCL), by deriving analytical (i.e., closed-form) solutions from frozen extracted features. In local training, our AFCL enables single-epoch learning with only a lightweight forward-propagation process for each client. In global aggregation, the server can recursively and efficiently update the global model with single-round aggregation. Theoretical analyses validate that our AFCL achieves spatio-temporal invariance of non-IID data. This ideal property implies that, regardless of how heterogeneous the data are distributed across local clients and online tasks, the aggregated model of our AFCL remains invariant and identical to that of centralized joint learning. Extensive experiments show the consistent superiority of our AFCL over state-of-the-art baselines across various benchmark datasets and settings.