O’TRAIN: A robust and flexible ‘real or bogus’ classifier for the study of the optical transient sky

Open Access

Table 5

Results of the different metrics used to evaluate the classification performances of our CNN models trained on 5 telescope data cubes.

Telescope name	Detection pipeline	p_t [0-1]	Acc [0-1]	Loss [0-1]	Prec [0-1]	Recall [0-1]	F1-score [0-1]	MCC [0-1]
JAST	gmadet	0.35	0.98	0.06	0.98	0.99	0.99	0.97

JAST	STDPipe	0.51	0.95	0.10	0.94	0.97	0.96	0.91

FRAM-CTA-N	gmadet	0.55	0.93	0.30	0.89	0.97	0.93	0.85

TACA	gmadet	0.52	0.93	0.20	0.91	0.96	0.95	0.86

TCA	gmadet	0.34	0.94	0.2	0.92	0.97	0.95	0.89

Notes. The values of the metrics have been computed based on the best probability threshold, p_t · p_t has been chosen to maximise the F1-MCC scores and minimise the false alarm rate (i.e. false positives).

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.