Open Access
Table 4.
Different network architectures loss.
| Training loss | Validation loss | ||
|---|---|---|---|
| One 2D convolution layer 34 filters | Keff | 0.045E−3 | 0.41E−3 |
| Pexp | 0.19E−3 | 2.0E−3 | |
| F3D | 0.34E−3 | 2.6E−3 | |
| Chosen network configuration | Keff | 0.055E−3 | 0.26E−3 |
| Two 2D convolution layers | Pexp | 0.19E−3 | 2.3E−3 |
| 34 and 68 filters | F3D | 0.42E−3 | 1.7E−3 |
| Six 2D convolution layers | Keff | 0.34E−3 | 1.3E−3 |
| First layer: 34 filters | Pexp | 0.15E−3 | 4.4E−3 |
| Others: 68 filters | F3D | 0.33E−3 | 1.5E−3 |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.