Beyond Weights and Gradients: A Taxonomy of Federated Learning Messages
For researchers and practitioners in federated learning, this taxonomy provides a structured framework to understand trade-offs in computational demands, communication costs, and privacy risks across different message types.
The paper proposes a formal taxonomy of federated learning messages beyond weights and gradients, categorizing them into model structures, statistical summaries, and data-conditioned representations. A review of 202 recent publications reveals a significant shift since 2021 toward diverse messaging paradigms.
Federated Learning is rapidly evolving beyond the exchange of traditional model weights and gradients, yet existing definitions fail to capture the full scope of modern payloads like synthetic data and federated analytics. This paper addresses the gap by proposing a formal mathematical definition of a federated message that accounts for both utility and privacy. We introduce a taxonomy that organizes these exchanges into three categories: model structures, statistical summaries, and data-conditioned representations. By evaluating these groups based on computational demands, communication costs, and privacy risks, we provide a clearer understanding of the trade-offs involved in decentralized training. Our review of 202 recent publications highlights a significant shift since 2021 toward diverse messaging paradigms, signaling a move away from standard deep learning updates toward more specialized information sharing. This framework provides a structured path for future research to optimize federated systems for varying hardware and security requirements.