An incentive mechanism design for federated learning with multiple task publishers by contract theory approach

Abstract

In the process of model training of the federated learning system, how to design an incentive mechanism to attract more high-quality worker nodes to join is a key issue. The existing researches on federated learning incentive mechanism only consider the scenario of single task publisher with multiple worker nodes. In the scenario of multi task publishers and multiple worker nodes, the competition between different task publishers makes the entire research process more complicated, and the contract design in the single task publisher scenario cannot be directly applied. To solve this problem, this paper proposes an incentive mechanism based on contract theory in the multi task publisher scenario and studies its application in federated learning. Simulation experiments show that this mechanism is effective for federated learning and can achieve the purpose of encouraging worker nodes to join and improve the efficiency of federated learning.