A Comprehensive Literature Survey on Federated Learning

The federated learning approach (FLA) has been recognized as a novel methodology for training machine intelligence models on numerous devices or servers, while ensuring that the data remains localized. In the current era of digitization, the significance of data has been increasingly recognized. Consequently, the issues pertaining to data privacy, security, and accessibility have emerged as critical difficulties. This paper provides a comprehensive analysis of the fundamental components of Federated Learning Aggregation (FLA) [1]. It explores various model aggregation methodologies, the complexities associated with data partitioning, the range of open-source frameworks available, and the intricacy of both centralized and decentralized network topologies [2]. Furthermore, the document highlights the difficulties and novel approaches associated with the accessibility of data in remote settings. The study also adopts a rigorous methodology in comprehending the diverse optimization approaches that may be applied to federated learning (FL) models, as well as the significance of secure network protocols in guaranteeing resilient and secure communication inside federated systems [3].