TABLE 1.
Tuned parameters for each DeepMirCut model and type of input.
| Input Type | Nucleotide sequence only | Nucleotide and dot-bracket sequence | Nucleotide and bpRNA sequence |
|---|---|---|---|
| Embedding layer | 96 units | 32 units | 32 units |
| Dropout | 0.315 | 0.213 | 0.417 |
| Bi-LSTM layer 1 | 320 units | 64 units | 128 units |
| Bi-LSTM layer 2 | 192 units | 256 units | 320 units |
| Learning rate | 3.2 10–3 | 1.91 10–3 | 3.57 10–3 |
| Epsilon (10x) | −7.56 | −6.79 | −6.87 |