Table 2
Optimal configurations of the single-band classifiers, yielding best results on the validation sets.
| Observational setup | n PCA | n hn | n l | n c | F 1,va | F 1,tt | S va | S tt |
|
|
||||||||
| Euclid | 24 | 26 | 2 | 48 | 0.75 | 0.42 | 0.98 | 0.90 |
| HST F606W | 27 | 29 | 3 | 24 | 0.57 | 0.30 | 0.94 | 0.68 |
Notes. The hyper-parameter nPCA is the number of retained PCA components, nhn is the number nodes in the hidden layers of the ANN, nl is the number of hidden layers, and nc is the number of ANNs retained in the committee (out of the 96 trained). The F1 score and the success rate S are given for the validation (va) and test (tt) sets. If the output catalogues were randomly drawn, the metrics would be F1 ≈ 0.07 and S ≈ 0.21.
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.