Table 4 Federated learning configuration.
Parameter | Value |
|---|---|
Number of clients | 10 |
Local epochs | 5 |
Batch size | 32 (edge) |
Learning rate | 0.001– 0.005 |
Optimizer | Adam (β1 = 0.9, β2 = 0.999) |
DP noise multiplier | σ = 1.1 |
Aggregation method | FedAvg |
Differential privacy | Gaussian mechanism (ε = 2.5 for proposed) |
Diffusion steps | 20 |
Generative AI Component | Client level data augmentation through variational autoencoder (VAE). |
