Stellar age determination using deep neural networks

Context . Recent spectroscopic surveys provide element abundances for large samples of Milky Way stars, from which stellar parameters can be inferred. Stellar ages, among them, are both a notoriously difficult parameter to estimate and a fundamental property for Galactic archaeology studies. Aims . We aim to develop a model-driven deep learning approach to age determination by training neural networks on stellar evolutionary grids. Contrary to the usual data-driven deep learning approach of using prior age estimates as training data, our method has the potential for a wider and less biased range of application. The low computational cost of deep learning methods compared to, for example, Bayesian isochrone fitting enables a broad analysis of large spectroscopic catalogues. Methods . We trained multilayer perceptrons on different stellar evolutionary grids to map M/H, M G , ( G BP − G RP ) to stellar age τ . We combined Gaia photometry and parallaxes, metallicities, and α elements from spectroscopic surveys and extinction maps, which are passed through neural networks to estimate stellar ages. Results . We applied our method to the LAMOST DR10, GALAH DR3 & DR4, and APOGEE DR17 spectroscopic surveys, estimating ages using the BaSTI tracks and other stellar evolutionary models. We leveraged this novel technique to study, for the first time, differences in age estimates from several evolutionary grids applied to very large datasets. In addition, we dated 13 open clusters and one globular cluster, finding a median absolute deviation with literature ages of 0.20 Gyr. Along with the stellar age catalogues from our estimates, we release NEST (Neural Estimator of Stellar Times), a python package to estimate stellar age based on this work, as well as a web interface. Conclusions . We show that, when using the same evolutionary grid, our method retrieves the same ages as a Bayesian approach similar to SPInS, for only a fraction of the computational cost, with a 60 000 speed-up factor for a typical star. This model-driven deep learning technique thus opens up the way for broad galactic archaeology studies on the largest datasets available today and in the near future with upcoming surveys such as 4MOST.