FedSTaS: Client Stratification and Client Level Sampling for Efficient Federated Learning

2024/12/18 by Jordan Slessor, Slessor, Jordan, D. J. Kong +7 · 1 citation
Computer Science · Decision Sciences · #Privacy-Preserving Technologies in Data #Data Quality and Management #Internet Traffic Analysis and Secure E-voting

paper · pdf · doi:10.48550/arxiv.2412.14226

Abstract

Federated learning (FL) is a machine learning methodology that involves the collaborative training of a global model across multiple decentralized clients in a privacy-preserving way. Several FL methods are introduced to tackle communication inefficiencies but do not address how to sample participating clients in each round effectively and in a privacy-preserving manner. In this paper, we propose FedSTaS, a client and data-level sampling method inspired by FedSTS and FedSampling. In each federated learning round, FedSTaS stratifies clients based on their compressed gradients, re-allocate the number of clients to sample using an optimal Neyman allocation, and sample local data from each participating clients using a data uniform sampling strategy. Experiments on three datasets show that FedSTaS can achieve higher accuracy scores than those of FedSTS within a fixed number of training rounds.

Cited by

• HeteRo-Select: Informativeness as the Participation Driver in Heterogeneous Federated Learning

Related

• FedSampling: A Better Sampling Strategy for Federated Learning
• LEFL: Low Entropy Client Sampling in Federated Learning
• Clustered Sampling: Low-Variance and Improved Representativity for Clients Selection in Federated Learning
• Better Methods and Theory for Federated Learning: Compression, Client Selection and Heterogeneity
• FedProf: Selective Federated Learning with Representation Profiling