Oral comunication
A. W. Neitzel, T. L. Campante, D. Bossini, A. Miglio
Abstract
Stellar populations distinguish themselves from one another via differences in
chemical, kinematic and chronological properties, suggesting an interplay of different physical
mechanisms that led to their origin and subsequent evolution. As such, the identification of
stellar populations is key for gaining insight on the evolutionary history of the Milky Way galaxy.
This task is made complicated by the fact that stellar populations share significant overlap in
their chrono-chemo-kinematic properties, which hinders the ability to both identify and define
stellar populations. To tackle this problem, we explore use cases of manifold learning, a type of
unsupervised machine learning application which seeks to intelligently identify and disentangle
manifolds hidden in the input data. To test the method, we make use of Gaia DR3-like synthetic
stellar samples generated from the FIRE-2 cosmological simulation. We show that manifold
learning possesses promising abilities to differentiate stellar populations, even when considering
realistic observational constraints.
8th TESS/15th Kepler Asteroseismic Science Consortium Workshop
Porto, Portugal
2024 July
