Enhancing Federated Class-Incremental Learning via Spatial-Temporal Statistics Aggregation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Guan, Zenghao, Zhu, Guojun, Zhou, Yucan, Liu, Wu, Wang, Weiping, Luo, Jiebo, Gu, Xiaoyan
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908787363807232
author Guan, Zenghao
Zhu, Guojun
Zhou, Yucan
Liu, Wu
Wang, Weiping
Luo, Jiebo
Gu, Xiaoyan
author_facet Guan, Zenghao
Zhu, Guojun
Zhou, Yucan
Liu, Wu
Wang, Weiping
Luo, Jiebo
Gu, Xiaoyan
contents Federated Class-Incremental Learning (FCIL) enables Class-Incremental Learning (CIL) from distributed data. Existing FCIL methods typically integrate old knowledge preservation into local client training. However, these methods cannot avoid spatial-temporal client drift caused by data heterogeneity and often incur significant computational and communication overhead, limiting practical deployment. To address these challenges simultaneously, we propose a novel approach, Spatial-Temporal Statistics Aggregation (STSA), which provides a unified framework to aggregate feature statistics both spatially (across clients) and temporally (across stages). The aggregated feature statistics are unaffected by data heterogeneity and can be used to update the classifier in closed form at each stage. Additionally, we introduce STSA-E, a communication-efficient variant with theoretical guarantees, achieving similar performance to STSA-E with much lower communication overhead. Extensive experiments on three widely used FCIL datasets, with varying degrees of data heterogeneity, show that our method outperforms state-of-the-art FCIL methods in terms of performance, flexibility, and both communication and computation efficiency. The code is available at https://github.com/Yuqin-G/STSA.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01327
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Enhancing Federated Class-Incremental Learning via Spatial-Temporal Statistics Aggregation
Guan, Zenghao
Zhu, Guojun
Zhou, Yucan
Liu, Wu
Wang, Weiping
Luo, Jiebo
Gu, Xiaoyan
Machine Learning
Artificial Intelligence
Federated Class-Incremental Learning (FCIL) enables Class-Incremental Learning (CIL) from distributed data. Existing FCIL methods typically integrate old knowledge preservation into local client training. However, these methods cannot avoid spatial-temporal client drift caused by data heterogeneity and often incur significant computational and communication overhead, limiting practical deployment. To address these challenges simultaneously, we propose a novel approach, Spatial-Temporal Statistics Aggregation (STSA), which provides a unified framework to aggregate feature statistics both spatially (across clients) and temporally (across stages). The aggregated feature statistics are unaffected by data heterogeneity and can be used to update the classifier in closed form at each stage. Additionally, we introduce STSA-E, a communication-efficient variant with theoretical guarantees, achieving similar performance to STSA-E with much lower communication overhead. Extensive experiments on three widely used FCIL datasets, with varying degrees of data heterogeneity, show that our method outperforms state-of-the-art FCIL methods in terms of performance, flexibility, and both communication and computation efficiency. The code is available at https://github.com/Yuqin-G/STSA.
title Enhancing Federated Class-Incremental Learning via Spatial-Temporal Statistics Aggregation
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2506.01327