Table 10 Hyperparameter settings for the client-end deep learning model and the control parameters for the CVXOPT solver
From: FedOcw: optimized federated learning for cross-lingual speech-based Parkinson’s disease detection
Layer/Component | Parameter Settings |
|---|---|
Time-distributed 2D-CNNs | Filters = 16, Kernel size = (3,3), Padding = (1,1), Activation = ReLU |
Time-distributed Batch Normalization | Momentum = 0.1 |
Time-distributed 2D AveragePooling | Pool size = (3,3) |
Time-distributed Dropout | Dropout rate = 0.3 |
Time-distributed Flatten | — |
1D-CNN | Filters = 8, Kernel size = 3, Padding = 1, Activation = Sigmoid |
1D AveragePooling | Pool size = 3, Stride = 1 |
Flatten | — |
Fully Connected Layer | 384 hidden units, Activation = ReLU |
Dropout | Dropout rate = 0.3 |
Output Layer | Activation = Sigmoid |
Common Settings | Optimizer = Adam, Learning rate = 0.001, Epochs per round = 10, Batch size = 8 |
CVXOPT Solver Parameters | Maxiters = 20, Abstol = 1e-3, Reltol = 1e-3, Feastol = 1e-3 |
