Table 7 Comparison experiment of weight selection based on deep-and-shallow loss and single loss
From: Spatial and temporal learning representation for end-to-end recording device identification
Loss function and parameters | Network model | End-to-end (ACC) |
|---|---|---|
Categorical crossentropy loss | DNN-LSTM | 96.5% |
| Â | DNN + Bi-LSTM | 96.6% |
Deep-and-shallow loss (0.25:0.5:0.25) | DNN-LSTM | 97.3% |
| Â | DNN + Bi-LSTM | 97.5% |
Deep-and-shallow loss (0.4:0.2:0.4) | DNN-LSTM | 97.2% |
| Â | DNN + Bi-LSTM | 97.2% |
Deep-and-shallow loss (0.25:0.25:0.5) | DNN-LSTM | 97.1% |
| Â | DNN + Bi-LSTM | 97.5% |
Deep-and-shallow loss (0.2:0.6:0.2) | DNN-LSTM | 97.3% |
| Â | DNN + Bi-LSTM | 97.3% |