Enhancing Fairness in Federated Learning: A Contribution-Based Differentiated Model Approach

Table 2

Fairness of FedAvg and our method over MNIST dataset, with different client numbers (N-k).

Method	N4	N8	N16	N32
FedAvg	−0.23	−0.46	0.13	0.04
Our method	0.90	0.77	0.68	0.73