Table 4.
Hyperparameters used during training.
| Parameter | Value |
|---|---|
| Learning rate | 2 × 10−5 |
| Batch size | 256 (128) |
| Iterations | 1 × 105 (4 × 104) |
| EMA rate | 0.9999 (–) |
| PMean | −2.5 |
| PStd | 1.8 |
| Dropout rate | 0.1 |
| Context dropout rate | 0.1 |
Notes. If present, numbers in brackets indicate the values used for training the FIRST model. Otherwise, the same values were used for both the training runs. EMA is only used for the LOFAR model.
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.