Table E.2.
Training machine learning methods on image data of Fornax and evaluating it on Virgo image data.
| averaged per galaxy | average over all sources | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Method | TPR | FPR | FDR | AUC ROC | TPR | FPR | FDR | AUC ROC | # TPs | # FPs |
| Nearest Neighbour | 0.81 | 0.04 | 0.31 | − | 0.86 | 0.04 | 0.15 | − | 10613 | 1926 |
| 12 Nearest Neighbours | 0.80 | 0.04 | 0.31 | 0.96 | 0.86 | 0.04 | 0.15 | 0.97 | 10676 | 1815 |
| Random Forest | 0.77 | 0.04 | 0.32 | 0.95 | 0.85 | 0.04 | 0.15 | 0.97 | 10468 | 1823 |
| Convolutional Neural Network (CNN) | 0.90 | 0.03 | 0.20 | 0.99 | 0.93 | 0.02 | 0.09 | 0.99 | 11560 | 1194 |
| CNN + Nearest Neighbour | 0.84 | 0.03 | 0.20 | − | 0.88 | 0.02 | 0.09 | − | 10920 | 1132 |
| CNN + 12 Nearest Neighbours | 0.86 | 0.02 | 0.15 | 0.98 | 0.91 | 0.02 | 0.06 | 0.99 | 11239 | 770 |
Notes. The Virgo data contains in total 62581 sources (we excluded sources from VCC538 as it contained a lot of false positives from a background galaxy) with 12385 catalogued GCs. Results are given for a decision threshold of 0.5. Similar to Table 5.
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.