Table 6 Federated learning hyperparameter configuration.
Parameter | Description/value |
|---|---|
Number of clients | 20 (simulated learners) |
Client sampling ratio | 0.5 per round |
Local epochs per round | 5 |
Communication rounds | 20 |
Client optimizer | Adam |
Local learning rate | 0.005 |
Aggregation method | FedAvg (weighted by data volume) |
Gradient clipping threshold | 1.0 |
Proximal regularization coefficient (µ) | 0.01 (for non-IID data) |
Batch size (local) | 64 |
Differential privacy | Disabled in baseline; considered for future work |
