Client and Training Data Selection for Computationally Efficient Synchronized Federated Learning

2026-09-30

AI Summary

This study addresses the slow convergence and degraded accuracy in federated learning caused by straggler-induced latency and non-independent and identically distributed (Non-IID) data. To mitigate these challenges, we propose a joint client and training data selection algorithm that transcends the limitations of conventional probabilistic selection strategies. By integrating asynchronous and synchronous aggregation mechanisms with distributed privacy-preserving techniques, the proposed method effectively optimizes computational efficiency and client participation rates, enabling efficient synchronous federated learning under few-shot Non-IID scenarios. Experimental evaluations on the CIFAR-100 dataset demonstrate that the algorithm significantly accelerates model convergence and improves classification accuracy. Furthermore, it exhibits strong robustness in heterogeneous computing environments, successfully balancing performance gains with resilience to diverse hardware capabilities across participating clients.

Abstract

Federated learning (FL) is a promising paradigm of machine learning, which preserves user privacy by enabling learning without sharing raw data with a cloud server. Straggling clients have been a problem for FL as they introduce delays in aggregating the local models and hence, the convergence of the global model. Therefore, it is important to have a mechanism that ensures fast convergence of the global model as well as good FL participation rate. Another issue for the convergence of a model in FL is the non-independent and identically distributed (non-iid) data across the clients. Prior approaches based on probabilistic client selection do not work well under non-iid data especially when the number of clients is small. We show scenarios where such approaches fail and propose a joint client-training data selection algorithm for fast convergence of FL models. Our experiments on CIFAR-100 dataset show that convergence of the FL model can be significantly improved over prior works that can consider non-iid data and heterogeneous computation and higher model accuracy.

Problem

• Federated Learning
• Straggling Clients
• Non-IID Data
• Client Selection
• Model Convergence

Innovation

• Federated Learning
• Client Selection
• Training Data Selection
• Non-IID Data
• Synchronized Federated Learning