Federated Learning as a Mean-Field Game
This foundational work aims to bridge machine learning and game theory to address large-scale distributed and privacy-preserving learning, though it is incremental in conceptual framing without new empirical results.
The paper connects federated learning to mean-field game theory by modeling local learners as players and gradient aggregation as a mean-field effect, presenting it as a differential game and discussing equilibrium properties.
We establish a connection between federated learning, a concept from machine learning, and mean-field games, a concept from game theory and control theory. In this analogy, the local federated learners are considered as the players and the aggregation of the gradients in a central server is the mean-field effect. We present federated learning as a differential game and discuss the properties of the equilibrium of this game. We hope this novel view to federated learning brings together researchers from these two distinct areas to work on fundamental problems of large-scale distributed and privacy-preserving learning algorithms.