Table 3
Error statistics of the FFNN in predicting abundance (A(Na)) in the leave-one-out cross validation on the 3D NLTE grid.
| λ[Å] | RMS(∆A(Na)) [dex] | MAD(∆A(Na)) [dex] | nl | n | α |
|---|---|---|---|---|---|
| 5682 | 0.027 | 0.008 | 3 | 400 | 0.001 |
| 5688 | 0.028 | 0.007 | 2 | 400 | 0.001 |
| 5889 | 0.032 | 0.016 | 2 | 500 | 0.01 |
| 5896 | 0.036 | 0.015 | 3 | 200 | 0.0001 |
| 6154 | 0.023 | 0.008 | 2 | 500 | 0.001 |
| 6160 | 0.026 | 0.008 | 3 | 400 | 0.001 |
| 8183 | 0.038 | 0.009 | 3 | 200 | 0.0001 |
| 8194 | 0.037 | 0.011 | 3 | 100 | 0.001 |
| 10 747 | 0.020 | 0.005 | 2 | 500 | 0.0001 |
Notes. The columns show the root-mean-square (RMS) and the median absolute deviation (MAD) of the corresponding FFNN with optimised hyperparameters (number of layers, nl, neurons per layer, n, and L2 penalty, α).
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.