Table 4. Total number of parameters for four models.
Comparison of the total number of parameters for four different model configurations: DNABERT_2, DNABERT_2+CNN, DNABERT_2+BiLSTM, and DNABERT_2+CNN+BiLSTM. The table shows that the model with both CNN and BiLSTM layers (DNABERT_2+CNN+BiLSTM) has the highest parameter count, indicating increased complexity and capacity for learning more intricate patterns from the data.
| Model | Total parameters |
|---|---|
| DNABERT_2 | 117,070,082 |
| DNABERT_2+CNN | 120,615,170 |
| DNABERT_2+BiLSTM | 117,988,610 |
| DNABERT_2+CNN+BiLSTM | 122,318,594 |