2020ApJS..249....7B

Stellar ages play a crucial role in understanding the formation and evolution of stars and Galaxies, which pose many challenges while determining in practice. In this paper, we have introduced a new machine-learning method, Gaussian process regression (GPR), to estimate the stellar ages, which is different from the traditional isochrone fitting method, which fully utilizes the information provided by previous studies. To demonstrate the performance of our method, we have applied it to the field stars of two important phases of evolution, main-sequence turn-off (MSTO) stars and giants, whose ages and masses are available in the literature. Also, GPR is applied to the red giants of open clusters (e.g., M67). Results showed that the ages given by GPR are in better agreement with those given by isochrone fitting methods. The ages are also estimated from various other machine-learning methods (e.g., support vector regression, neural networks, and random forest) and are compared with GPR, which resulted in GPR outperforming others. In addition to ages, we have applied GPR to estimate the masses of the MSTO stars and red giants and found that the masses predicted by GPR for the red giants are within acceptable uncertainties of masses derived from the asteroseismic scaling relation. We have provided the constraints on the input parameters to GPR, which decides the accuracy of the output ages and masses. Results conclude that the newly introduced GPR is promising to provide a novel approach to estimate stellar ages and masses in the era of big data sets. As a supplement, masses and ages for the MSTO stars and red giants estimated from GPR are provided as a catalog that could be used as a training set for upcoming large data sets with spectroscopic parameters.