Extracting Milky Way stellar populations using t-SNE

Abstract

t-SNE is a clustering algorithm that can excel at grouping large, multidimensional datasets such as that produced by a spectroscopic survey. We apply t-SNE to the GALAH+ DR3 high-resolution spectroscopic survey and successfully extract populations that are well studied within the Milky Way. t-SNE successfully separated a halo population from the main disc and produced a clustering representative of the thick and thin disc subpopulations. These results are consistent with both chemical and dynamical definitions of Milky Way populations and show t-SNE can reveal potential new populations that are within these large datasets.