Table 4 Hyper-parameter tuning ranges and optimal values for SWAM model
From: An explainable CNN approach for medical codes prediction from clinical text
Range | Optimal value | |
|---|---|---|
\(\eta\) | 0.0001,0.0003, | 0.001 |
(learning rate) | 0.001,0.003 | |
k | 1–10 | 4 |
(filter size) | ||
\(d_{c}\) | 50–500 | 500 |
(number of filters) | ||
q | 0.2–0.8 | 0.2 |
(dropout probability) |