Table 6 The preset value of the hyperparameter.
From: Channel equalization in ultraviolet communication based on LSTM-DNN hybrid model
Parameter | Value |
|---|---|
Batch size | 64 |
Time step | 10 |
Maximum Epoch iterations | 500 |
Number of input layer neurons | 128 |
Number of hidden layer neurons in a single-layer DNN | 60 |
Number of hidden layers in DNN (H) | 4 |
Number of single-layer LSTM units | 128, 256 |
LSTM layers | 2 |
Initial learning rate | 0.00002 |
Ratio of training test set (training: test) | 7:3 |
Data volume | 12,000 |
