Open Access
Table 3.
Model and training hyperparameters with the range within which they are selected.
| Hyperparameter | Allowed range |
|---|---|
| Hidden layers | [11, 14] |
| Neurons per layer | [110, 135] |
| Dropout | 0.0 |
| Number of Gaussians | [2, 4, 6, 8] |
| Type embedding dimension | 10 |
| Gain(model initialization) | [0.85, 1.05] |
| Learning rate | [0.00005, 0.005] |
| Weight decay | [0.0000005, 0.003] |
| Step size (StepLR Pytorch’s scheduler) | [20] |
| Gamma (StepLR Pytorch’s scheduler) | [0.1, 0.5] |
Notes. These specific ranges are selected by observing model performance with hyperparameter values in wider ranges.
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.