Evaluation of Recent Developments in Federated Learning

Federated learning (FL) is a crucial technology for healthcare, IoT, and finance applications. This paper evaluates recent advancements in FL from 2023 to 2025, focusing on optimization algorithms, privacy-preserving techniques, communication efficiency, and real-world applications. It compares algorithms like FedAvg, FedProx, SCAFFOLD, and FedDyn, assessing their performance under data heterogeneity and communication constraints. Privacy techniques like differential privacy and secure aggregation are evaluated for accuracy and computational overhead. Communication-efficient methods and real-world deployments are also analyzed. The evaluation offers actionable insights for selecting appropriate FL methods for specific use cases.