Prof. DENG Linhua from Yunnan Observatories of the Chinese Academy of Sciences and his collaborators proposed a robust identification method to identify hot subdwarfs based on a convolutional neural network.

The method can help to identify specific spectra with robust results and high accuracy, and can be further applied to the classification of large-scale spectra and the search for specific targets.

Related results were published in The Astrophysical Journal Supplement Series on Feb. 17.

Prof. WANG Feng and TAN Lei from Guangzhou University, and researchers from National Astronomical Observatories of the Chinese Academy of Sciences and China West Normal University were also involved in the study.

The researchers first constructed the data set using the spectral data of LAMOST DR7-V1, and then constructed a hybrid recognition model including an eight-class classification model and a binary classification model. They selected 835 hot subdwarfs that were not involved in the training process from the identified LAMOST catalog as the validation set.

Hot subdwarfs are core helium-burning stars located below the upper main sequence of the Hertzsprung-Russell diagram and are referred to as extreme horizontal branch stars because of their evolution stage.

The traditional method searching for subdwarfs is mainly based on the basic characteristics of hot subdwarfs. However, it is mainly dependent on manual processing, which is laborious and difficult to use to meet the demands of handling large-scale spectral data.

"Identifying hot subdwarfs from the LAMOST catalog has essential research value because LAMOST can reveal the spectral characteristics of hot subdwarfs that show details of their formation and evolution," said Prof. DENG.

Identifying hot subdwarfs based on deep learning can reduce the difficulty of manual identification. In the study, a hybrid model with an eight-class classification model and a binary classification model was proposed and applied to search for hot subdwarfs in the LAMOST catalog.

Based on the model, the researchers filtered and classified all 10,640,255 spectra of LAMOST DR7-V1, and obtained a catalog of 2,393 hot subdwarf candidates, of which 2,067 have been confirmed. They also found 25 new hot subdwarfs among the remaining candidates by manual validation.