Enhancing Federated Learning: A Novel Approach of Shapley Value Computation in Smart Contract

In federated learning (FL), the success of model training largely hinges on the contributions from clients. Current FL frameworks encounter obstacles in pinpointing and compensating high-contribution clients effectively. This paper proposes a novel method that innovates on the computation of Data Shapley values to accurately assess client contributions, thereby streamlining the overall learning process. This system ensures that clients who significantly contribute to model training are adequately compensated, fostering a cooperative and engaged environment for FL. The integration of our innovative Data Shapley calculation method with smart contracts not only guarantees the transparency and equity of the incentive mechanism but also strengthens the security and integrity of the FL framework. By efficiently identifying and incentivizing high-contribution clients, our method markedly improves model performance and operational efficiency, enhancing the viability and attractiveness of federated learning.