Table 1 Hyperparameters and tuned values for Deep-CTGAN and ResNet.
Model component | Parameter | Tuned value |
|---|---|---|
Deep-CTGAN | Learning rate | 0.0002 |
Batch size | 128 | |
Embedding dimension | 128–256 (dataset-specific) | |
Number of epochs | 300 | |
Generator activation | LeakyReLU | |
Discriminator activation | LeakyReLU | |
Optimizer | Adam | |
Adam betas | (0.5, 0.999) | |
Mode-specific normalization | Enabled | |
Gradient penalty | Applied | |
Conditional vector encoding | Used for class labels | |
ResNet (integrated) | Network depth | ResNet-18 |
Residual block type | Basic Block | |
Dropout rate | 0.3 | |
Weight decay | 1e-5 | |
Activation function | ReLU | |
Optimizer | Adam | |
Learning rate | 0.0001 | |
Batch normalization | Enabled |
