Table 4 The proposed InfoGAN structure.
From: Estimating the phase volume fraction of multi-phase steel via unsupervised deep learning
Generator network (\(G\)) | Discriminator network (\(D\)) | Auxiliary network (\(Q\)) | ||||||
|---|---|---|---|---|---|---|---|---|
Layer | Type | Dimension | Layer | Type | Dimension | Layer | Type | Dimension |
Input | Latent (\(z\)) + Code (\(c\)) | 4 + 4 | Input | Feature | 19 | Input | Hidden layer | 50 |
Hidden 1 | Dense layer | 50 | Hidden 1 | Dense layer | 100 | Hidden 1 | Dense layer | 50 |
ReLU activation | – | ReLU activation | – | ReLU activation | – | |||
Hidden 2 | Dense layer | 50 | Hidden 2 | Dense layer | 50 | Hidden 2 | Dense layer | 20 |
ReLU activation | – | ReLU activation | – | ReLU activation | – | |||
Output | Dense layer | 19 | Output | Dense layer | 1 | Output | Dense layer | 4 |
Linear activation | – | Sigmoid activation |  | Softmax activation | – | |||
