Euclid preparation - XXXIII. Characterization of convolutional neural networks for the identification of galaxy-galaxy strong-lensing events

Open Access

Table B.5

Results of testing our best-performing networks, trained on S1, on a test set with 200 lenses and 80 000 nonlenses.

	VGG-like network		Inception Network		Residual Network		Ensemble Network
Class	0	1	0	1	0	1	0	1

Precision	1.0	0.15	1.0	0.45	1.0	0.13	1.0	0.46

Recall	0.98	0.94	0.99	0.96	0.98	0.92	1.0	0.97

F1-score	0.99	0.26	0.99	0.61	0.99	0.23	1.0	0.63

Accuracy	0.98		0.99		0.98		1.0

AUC	0.76		0.83		0.81		0.99

Notes. Class 0 refers to the nonlenses, while class 1 refers to the lenses. Ensemble network refers to the combination of the predictions of the three networks.

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.