Optimizing the Collaboration Structure in Cross-Silo Federated Learning

Wenxuan Bao, Haohan Wang, Jun Wu, Jingrui He
6/10/2023
cs.LG

No official code found. Best implementation: baowenxuan/FedCollab (21 stars, last commit 2023-07-25).

Abstract

In federated learning (FL), multiple clients collaborate to train machine learning models together while keeping their data decentralized. Through utilizing more training data, FL suffers from the potential negative transfer problem: the global FL model may even perform worse than the models trained with local data only. In this paper, we propose FedCollab, a novel FL framework that alleviates negative transfer by clustering clients into non-overlapping coalitions based on their distribution distances and data quantities. As a result, each client only collaborates with the clients having similar data distributions, and tends to collaborate with more clients when it has less data. We evaluate our framework with a variety of datasets, models, and types of non-IIDness. Our results demonstrate that FedCollab effectively mitigates negative transfer across a wide range of FL algorithms and consistently outperforms other clustered FL algorithms.

View on arXivView PDF

Code Implementations(1)

baowenxuan/FedCollabCommunity100%

[ICML 2023] Optimizing the Collaboration Structure in Cross-Silo Federated Learning. Wenxuan Bao, Haohan Wang, Jun Wu, Jingrui He.

212Shell, PythonMay 29, 20233 years ago
clustered-federated-learningfederated-learningpersonalized-federated-learning

Similar Papers

How to Collaborate: Towards Maximizing the Generalization Performance in Cross-Silo Federated Learning

90% match2024

Single-Round Clustered Federated Learning via Data Collaboration Analysis for Non-IID Data

89% match2026

FedClust: Tackling Data Heterogeneity in Federated Learning through Weight-Driven Client Clustering

87% match2024

Federated Clustering: An Unsupervised Cluster-Wise Training for Decentralized Data Distributions

86% match2024

Cross-Silo Federated Learning Across Divergent Domains with Iterative Parameter Alignment

86% match2023

FedDAG: Clustered Federated Learning via Global Data and Gradient Integration for Heterogeneous Environments

86% match2026

Cite this paper

@article{bao2023optimizing,
  title  = {Optimizing the Collaboration Structure in Cross-Silo Federated Learning},
  author = {Wenxuan Bao and Haohan Wang and Jun Wu and Jingrui He},
  year   = {2023},
  eprint = {2306.06508},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2306.06508v1},
  journal = {ICML 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2306.06508.svg)](https://codeofpaper.com/paper/2306.06508)

Discussion