Free Access

Table E.1.

Training machine learning methods on tabular data of Fornax and evaluating it on Virgo data.

averaged per galaxy average over all sources

Method TPR FPR FDR AUC ROC TPR FPR FDR AUC ROC # TPs # FPs
Logistic Regression 0.73 0.15 0.58 0.88 0.83 0.16 0.44 0.91 10241 8079
Support Vector Machine (linear) 0.76 0.14 0.56 0.84 0.14 0.41 10405 7293
Support Vector Machine (radial) 0.83 0.05 0.33 0.90 0.04 0.16 11213 2165
Nearest Neighbour 0.85 0.04 0.28 0.89 0.04 0.15 11031 1882
12 Nearest Neighbours 0.87 0.03 0.22 0.98 0.91 0.03 0.10 0.99 11338 1300
Decision Tree 0.84 0.03 0.26 0.95 0.88 0.03 0.12 0.97 10889 1520
Random Forest 0.84 0.02 0.14 0.98 0.89 0.01 0.06 0.99 11086 698
AdaBoost 0.84 0.02 0.14 0.98 0.89 0.01 0.06 0.99 11082 700
CatBoost 0.85 0.02 0.17 0.98 0.90 0.02 0.08 0.99 11169 908
Neural Network (29-1) 0.73 0.17 0.60 0.87 0.82 0.19 0.49 0.89 10193 9746
Neural Network (29-100-100-1) 0.88 0.02 0.17 0.99 0.92 0.02 0.08 0.99 11443 937

Notes. The Virgo data contains in total 63162 sources with 12395 catalogued GCs. Results are given for a decision threshold of 0.5. Similar to Table 3.

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.