Decentralized Federated Learning

Decentralized Federated Learning (DFL) is a collaborative machine learning framework that trains models across distributed devices without a central server, enhancing privacy and robustness. Current research focuses on addressing challenges like data heterogeneity and Byzantine attacks through novel aggregation algorithms, optimized communication strategies (including model caching and adaptive sampling), and the exploration of various network topologies. This serverless approach improves scalability and resilience compared to centralized federated learning, with significant implications for applications requiring privacy-preserving distributed training, such as healthcare and IoT.