Incentivize Contribution and Learn Parameters Too: Federated Learning with Strategic Data Owners
The paper introduces two mechanisms, UPBReD and 2P-UPBReD, to incentivize clients to contribute their maximum data while simultaneously learning the optimal model parameters in federated learning. UPBReD finds a Nash equilibrium for contribution and model training. 2P-UPBReD ensures full contribution by adding budget-balanced transfers, leading to the optimal model, and T-2P-UPBReD ensures truthful reporting.