view article
Figure 4
Confusion matrices for the deployed GoniOwl model (batch size 16, epoch-30 checkpoint) using deployment-style offline inference with two-sided confidence gating [confident if and only if max(s, 1 − s) > 0.85]. (a) The held-out test set and (b) the training set. Rows indicate true labels, columns predicted labels; class order is [pin_off, pin_on]. Predictions in the intermediate band were assigned to the fail-safe undetermined state and excluded from the matrices (none in the test set, eight in the training set).

Journal logoJOURNAL OF
SYNCHROTRON
RADIATION
ISSN: 1600-5775
Volume 33| Part 5| September 2026|
https://doi.org/10.1107/S1600577526007277
Follow J. Synchrotron Rad.
Sign up for e-alerts
E-alerts
Follow J. Synchrotron Rad. on Twitter
Twitter
Follow us on facebook
Facebook
Sign up for RSS feeds
RSS