Machine learning methods for constructing probabilistic Fermi-LAT catalogs

Free Access

Table 7.

Expected precision and recall for the classification of 4FGL-DR2 sources.

	Classification	AGN	PSR	OTHER
Two-class precision	all agree	0.96	0.71	–
	∑_ap_a > 7	0.97	0.78	–

Two-class recall	all agree	0.95	0.70	–
	∑_ap_a > 7	0.94	0.64	–

Three-class precision	all agree	0.98	0.89	0.80
	∑_ap_a > 7	0.99	0.94	0.85

Three-class recall	all agree	0.94	0.62	0.38
	∑_ap_a > 7	0.87	0.30	0.06

Notes. The expected precision and recall are calculated for two classification conditions: all algorithms agree (all agree) and the sum of probabilities for the eight algorithms is larger than 7 (∑_ap_a > 7) (for details on the two- and three-class classification, see Sects. 4.2 and 5, 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.