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

Table 1

Simulation parameters for DQN-based contribution scores update for FL.

Parameter	Value
Replay memory size	10000
Batch size	64
Optimizer	Adam
Activation function	ReLu
Learning rate of DQN	0.01
Discount factor	0.99
Initial contribution score	5
Number of workers	8 to 32
Ratio of high-contributing clients	0.25
Ratio of ordinary clients	0.5
Ratio of free-riders	0.25
Number of FL iterations	100