Centralized vs Decentralized Federated Learning: A trade-off performance analysis

2026-05-15

AI Summary

This study addresses the lack of systematic empirical comparison among centralized (CFL), decentralized (DFL), and semi-decentralized (SDFL) federated learning architectures with respect to the trade-offs between communication overhead, privacy preservation, and model performance. Leveraging the Fedstellar simulation platform, the authors conduct the first unified experimental evaluation across these three architectural paradigms using the MNIST dataset and a multilayer perceptron (MLP) classifier. The results reveal significant differences in accuracy, communication efficiency, and robustness, thereby clarifying the distinct operational strengths and limitations of each approach. These findings offer empirical evidence and practical design guidance for selecting appropriate federated learning architectures based on specific application requirements.

Abstract

Federated Learning (FL) has emerged as a promising paradigm for collaborative model training across distributed edge devices while preserving data privacy especially with the huge increase amount of data due to the adoption of technologies which contributes to the growing number of IoT devices. Storing this amount of data centrally is challenging due to issues like limited communication, privacy, and regulations. FL can be Centralized (CFL), Decentralized (DFL), and Semi-decentralized (SDFL). Choosing the right FL architecture depends on the application's needs. However, very few research studies have experimentally compared these three types of architectures to not only understand the respective strengths and limitations, but also trade-offs between different performance indicators. This paper overcome this lack of analysis, conducting experimental analyses using the Fedstellar simulator, MNIST dataset, and MLP classifier.

Problem

• Federated Learning
• Centralized
• Decentralized
• Performance Trade-off
• Architecture Comparison

Innovation

• Federated Learning
• Centralized vs Decentralized
• Performance Trade-off
• Experimental Analysis
• Fedstellar