Table B.1
Network architectures of the best-performing DNNs used as a forward model in the plaNETic code.
| Model | HL1 | HL2 | HL3 | HL4 | HL5 | HL6 | |
|---|---|---|---|---|---|---|---|
| PREV | 2048 | 2048 | 2048 | 2048 | 2048 | 2048 | |
| A | Ml | 2048 | 2048 | 2048 | 2048 | 2048 | 2048 |
| A | M2 | 2048 | 2048 | 2048 | 2048 | 2048 | 2048 |
| B | Ml | 512 | 1024 | 2048 | 1024 | 512 | 256 |
| B | M2 | 2048 | 2048 | 2048 | 2048 | 2048 | 2048 |
Notes. Network architectures are specified for both the old version of plaNETic (denoted as PREV) and the version that is newly introduced in this work. For the new version, separate DNNs were trained for different planetary mass ranges (M1 for planets from 0.5 to 6 M⊕, M2 for planets from 6 to 15 M⊕) and water prior options (A assuming a formation outside the iceline, B inside). The table shows the number of units in each of the hidden layers (HL) of the DNN in question.
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.