A systematic review of federated learning: Challenges, aggregation methods, and development tools

Since its inception in 2016, federated learning has evolved into a highly promising decentral-ized machine learning approach, facilitating collaborative model training across numerous devices while ensuring data privacy. This survey paper offers an exhaustive and systematic review of federated learning, emphasizing its categories, challenges, aggregation techniques, and associated development tools. To start, we outline our research strategy used for this survey and evaluate other existing reviews related to federated learning. We initiate the
discussion, about federated learning concepts, with a detailed examination of the primary challenges inherent in
federated learning, including communication overhead, device and data heterogeneity, and data privacy issues.
Subsequently, we scrutinize and classify various aggregation techniques designed to mitigate these challenges,
such as federated averaging, secure aggregation, and strategies leveraging clustering and optimization methodologies. Furthermore, we delve into the exploration of cutting-edge development tools and frameworks that expedite efficient implementations of federated learning. Through our review, we aspire to provide a holistic understanding of the federated learning landscape, thereby setting the stage for future investigations, advancements, and practical implementations in this prosperous field.