Table 13 2D River problem: hyperparameters for the CNN model
From: Deep convolutional architectures for extrapolative forecasts in time-dependent flow problems
Hyperparameters | Values |
|---|---|
Sequence length (\(n_t\)) | 20, 30 |
CNN block channels | [200, 200, 200], [250, 250, 250], [300, 300, 300] |
Latent dimension (m) | 32 \( \times \) 32 |
Kernel size (k) | 3 |
Activation | tanh |
Loss function | MSE |
Learning rate | \(3 \times 10^{-4}\) |
Max model parameters (\(n_t = 30\)) | 4,144,204 |
Min model parameters (\(n_t = 30\)) | 1,862,804 |