J-PLUS: Identifying Outlier Spectral Energy Distributions in the Fourth Data Release Using Unsupervised Machine Learning

概要

The selection of candidates belonging to rare or underrepresented stellar populations, as well as novel objects, is essential for spectroscopic follow-up. In this work, we demonstrate how combining broad- and narrow-band photometric information with machine-learning techniques enables this selection. Using photometric data in twelve filters from the Javalambre Photometric Local Universe Survey (J-PLUS) as input, we applied an anomaly detection technique based on autoencoders to identify spectral…