Abstract
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 broadband and narrowband photometric information with machine learning techniques enables this selection. Using photometric data in 12 filters from the Javalambre Photometric Local Universe Survey (J-PLUS) as input, we applied an anomaly detection technique based on autoencoders to identify spectral energy distributions (SEDs) that deviate from the dominant pattern in the sample of J-PLUS objects. Subsequently, we applied the t-distributed stochastic neighbor embedding (t-SNE) algorithm to the J-PLUS colors of the population of outlier SEDs. Combining information from the SIMBAD database and other astronomical surveys to classify the outliers identified by the autoencoder, we are able to characterize overdensities in the t-SNE map. Our study specifically focuses on well-defined islands in the t-SNE space, corresponding to specific populations. We identified a large sample of white dwarfs and low-mass stars, including a small subset of polluted white dwarfs, and report potential candidates for spectroscopic follow-up, including 19 carbon-rich stars, 11 cataclysmic variables, and 170 white dwarf–main-sequence binaries, all previously unreported in the literature.

This publication has 65 references indexed in Scilit: