Bandit-based Communication-Efficient Client Selection Strategies for\n Federated Learning

2020/12/14 by Yae Jee Cho, Samarth Gupta, Cho, Yae Jee +5 · 1 citation
Computer Science · #Artificial Intelligence (cs.AI) #Cryptography and Data Security #FOS: Computer and information sciences #Machine Learning (cs.LG) #Privacy-Preserving Technologies in Data #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.2012.08009

openalex publication_date 2020/12/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Due to communication constraints and intermittent client availability in\nfederated learning, only a subset of clients can participate in each training\nround. While most prior works assume uniform and unbiased client selection,\nrecent work on biased client selection has shown that selecting clients with\nhigher local losses can improve error convergence speed. However, previously\nproposed biased selection strategies either require additional communication\ncost for evaluating the exact local loss or utilize stale local loss, which can\neven make the model diverge. In this paper, we present a bandit-based\ncommunication-efficient client selection strategy UCB-CS that achieves faster\nconvergence with lower communication overhead. We also demonstrate how client\nselection can be used to improve fairness.\n

Cited by

• Proxy Model-Guided Reinforcement Learning for Client Selection in Federated Recommendation

Related

• Client Selection in Federated Learning: Convergence Analysis and Power-of-Choice Selection Strategies
• Client Selection for Generalization in Accelerated Federated Learning: A Multi-Armed Bandit Approach
• FedSS: Federated Learning with Smart Selection of clients
• Emulating Full Participation: An Effective and Fair Client Selection Strategy for Federated Learning
• Client Selection in Federated Learning: Principles, Challenges, and Opportunities