Table 4 Federated reinforcement learning parameter configuration specifications.
Parameter name | Value range | Functional description |
|---|---|---|
Learning rate (α) | 0.001–0.1.001.1 | Control the model parameter update step size |
Discount factor (γ) | 0.9–0.99 | Balancing immediate rewards with long-term benefits |
Exploration rate (ε) | 0.05–0.3 | Adjusting the exploration-exploitation balance |
Batch size | 32–256 | Optimize training efficiency and stability |
Aggregation round interval | 10–100 | Controlling global model update frequency |
Privacy Budget (δ) | 0.01–0.1 | Differential privacy protection strength |
Network Layers | 3–8 | Complexity of Deep Neural Networks |
Participant Weight | 0.1–1.0.1.0 | Contribution weight based on data quality |
