Search ranges and hyperparameters explored during architecture tuning of hybrid BiLSTM-FNN-based surrogate model
| Hyperparameter | Search range/options | Description |
|---|---|---|
| Dense activation function | ReLU, tanh, ELU, softplus | Activation function for hidden dense layers |
| Number of dense layers | 1–3 | Number of dense hidden layers |
| Dense neurons per layer | 16–256 (step 16) | Number of neurons in each dense layer |
| Activation function for static branch | ReLU, tanh, ELU, softplus | Activation function for static input in the layer merging recurrent and static branches |
| BiLSTM activation function | tanh, ReLU, ELU, sigmoid, softplus | Activation function for shaping unit output |
| BiLSTM recurrent activation function | tanh, ELU, sigmoid, softplus | Activation function used in gating mechanism |
| Number of BiLSTM layers | 1–3 | Number of BiLSTM layers |
| BILSTM units per layer | 16–256 (step 16) | Number of units in each BiLSTM layer |
| Learning rate | 1 × 10−5 to 1 × 10−1 | Learning rate for Adam optimiser |
| Batch size | {16, 32, 64, 128} | Number of samples processed together in one training step |
| Hyperparameter | Search range/options | Description |
|---|---|---|
| Dense activation function | ReLU, tanh, ELU, softplus | Activation function for hidden dense layers |
| Number of dense layers | 1–3 | Number of dense hidden layers |
| Dense neurons per layer | 16–256 (step 16) | Number of neurons in each dense layer |
| Activation function for static branch | ReLU, tanh, ELU, softplus | Activation function for static input in the layer merging recurrent and static branches |
| BiLSTM activation function | tanh, ReLU, ELU, sigmoid, softplus | Activation function for shaping unit output |
| BiLSTM recurrent activation function | tanh, ELU, sigmoid, softplus | Activation function used in gating mechanism |
| Number of BiLSTM layers | 1–3 | Number of BiLSTM layers |
| 16–256 (step 16) | Number of units in each BiLSTM layer | |
| Learning rate | 1 × 10−5 to 1 × 10−1 | Learning rate for Adam optimiser |
| Batch size | {16, 32, 64, 128} | Number of samples processed together in one training step |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.