Table 2 Deep learning model hyperparameter configuration specifying the architectural design and training parameters for the LSTM-Attention hybrid neural network applied to ventilation parameter prediction.
From: Digital twin-driven deep learning prediction and adaptive control for coal mine ventilation systems
Hyperparameter | Configuration Value | Description |
|---|---|---|
LSTM layers | 3 layers | Number of stacked LSTM layers |
Hidden units per layer | 128, 64, 32 | Neuron count in each LSTM layer |
Attention heads | 4 | Number of parallel attention mechanisms |
Dropout rate | 0.3 | Probability for dropout regularization |
Learning rate | 0.001 | Initial learning rate for Adam optimizer |
Batch size | 64 | Number of samples per training batch |
Time window length | 60 steps | Historical sequence length for input |
Prediction horizon | 12 steps | Future time steps to forecast |
