Task-Load-Aware Game-Theoretic Framework for Wireless Federated Learning

Federated learning (FL) can protect data privacy but has difficulties in motivating user equipment (UE) to engage task training. This letter proposes a Bertrand-game based framework address the incentive problem, where model owner (MO) issues an FL and employed UEs help train by using their local data. Specially, we consider impact of time-varying load channel quality on UE’s motivation task. We adopt finite-state discrete-time Markov chain (FSDT-MC) predict these parameters during Depending performance metrics set MO estimated energy cost task, each UE seeks maximize its profit. obtain Nash equilibrium (NE) game closed form, develop distributed iterative algorithm find it. Finally, simulation result verifies effectiveness proposed approach.