DIFFERENTIAL EVOLUTION-BASED FEDERATED LEARNING INCENTIVE METHOD AND SYSTEM

A differential evolution-based federated learning incentive method and system, for minimizing, along a time t during a federated learning process by using global optimization capabilities and local detection capabilities of DE, the difference and latency between a gain obtained by each participant from federation and a gain the participant should obtain, automatically balancing the difference between an actual gain and an expected resulting reward of each participant in the federated learning, effectively promoting participants to provide reliable data so that the federated learning is stably carried out for a long term, effectively realizing the dynamic adjustment of the total gain of the federated learning and the gain of each participant, maximizing a sustainable operation goal, minimizing the problem of unfairness of the participants, and avoiding manual intervention. L'invention concerne un procédé et un système d'incitation à l'apprentissage fédéré à base d'évolution différentielle, permettant de réduir