Table 3 Hyper-parameters used for the neural network
From: Temporal indexing of medical entity in Chinese clinical notes
Hyper-parameter | Value |
|---|---|
Dimension of word representation | 50 |
Dimension of position representation | 30 |
Dimension of feature representation | 20 |
Size of LSTM unit | 50 |
Size of convolution filter | 3/5/7 |
Number of convolution filters | 100 |
Max length of the sentence | 100 |
Dropout probability | 0.5 |
Batch size | 32 |
Training epochs | 20 |