Table 1
Metrics describing the performance of our halo star classifier for the three different cases depending on the training features available.
| Ideal GUMS sample with G < 12.5 mag | |||
|---|---|---|---|
| 6D phase-space + phot +[Fe/H] | 6D phase-space + phot | 5D phase-space + phot | |
| recall | 0.94 | 0.90 | 0.84 |
| precision | 0.99 | 0.98 | 0.98 |
| logarithmic loss | 0.01 | 0.01 | 0.01 |
| Matthews coef | 0.96 | 0.94 | 0.90 |
| GUMS sample with G < 12.5 mag convolved with median TGAS uncertainties | |||
| 6D phase-space + phot +[Fe/H] | 6D phase-space + phot | 5D phase-space + phot | |
| recall | 0.71 | 0.64 | 0.46 |
| precision | 0.90 | 0.87 | 0.82 |
| logarithmic loss | 0.01 | 0.01 | 0.01 |
| Matthews coef | 0.80 | 0.75 | 0.62 |
| Error convolved GUMS sample with Gaia DR2 uncertainties | |||
| 6D phase-space + phot +[Fe/H]a | 6D phase-space + phota | 5D phase-space + photb | |
| recall | 0.93 | 0.90 | 0.58 |
| precision | 0.98 | 0.98 | 0.92 |
| logarithmic loss | 0.01 | 0.01 | 0.02 |
| Matthews coef | 0.95 | 0.94 | 0.73 |
Notes. The upper bound of the uncertainty of these metrics, calculated as the standard deviation from the five fold cross-validation process during the training, is 0.03. The superscripts a and b indicate that the model was evaluated on GUMS selected samples with G < 12.5 mag, and G <19 mag respectively.
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.