Open Access
Table 2
Hyperparameters and other properties of the neural network.
| Property | Value | Range |
|---|---|---|
| Network type | conv. | conv., dense |
| Number of hidden layers | 2 | 1–3 |
| Nodes in the input l. | 401 | fixed |
| Nodes/filters in hid. l. | 24 and 8 | 4–32 |
| Nodes in the output l. | 10 | fixed |
| Conv. kernel size | 5 | 3–5 |
| Hid. l. activation | ReLU | ReLU, ELU |
| Out. l. activation | sigmoid | softmax, sigmoid |
| Dropout rate (in.-hid.) | 0.0 | 0.0–0.2 |
| Dropout rate (hid.-hid.) | 0.3 | 0.0–0.3 |
| Dropout rate (hid.-out.) | 0.4 | 0.2–0.4 |
| Training algorithm | Adam | Adam, SGD |
| Max. number of epochs | 5000 | fixed |
| Batch size | 8 | 8–1024 |
| Learning rate | 0.0005 | 0.0001–0.01 |
| L1 trade-off parameter | 0.005 | 0–0.001 |
| α trade-off parameter | 0.1 | 0.01–1 |
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.