Table 3 Hyperparameters used in our experiments
From: CapsTM: capsule network for Chinese medical text matching
Hyperparameter | Value | Hyperparameter (BERT) | Value |
|---|---|---|---|
Embedding size | 300 | Embedding size | 768 |
#Hidden states in BiLSTM | 100 | #Hidden states in BiLSTM | 384 |
#Capsules | 6 | #Capsules | 6 |
#Dimension of capsules | 50 | #Dimension of capsules | 50 |
#Iterations | 3 | #Iterations | 3 |
Learning rate | 0.001 | Learning rate | 0.001 |
Dropout rate | 0.5 | Dropout rate | 0.9 |
Activation function | ReLU | Activation function | ReLU |
Batch size | 32 | Batch size | 32 |