| Name: Subhajit Kar |
| Affiliation: Inter-University Centre for Astronomy and Astrophysics |
| Conference ID: ASI2026_792 |
| Title: Cartography of the Milky Way: Mapping Galactic Stellar Populations using Machine Learning |
| Abstract Type: Oral |
| Abstract Category: Stars, Interstellar Medium, and Astrochemistry in Milky Way |
| Author(s) and Co-Author(s) with Affiliation: Rajorshi Bhattacharya(University of New Mexico, Albuquerque - 87131, USA), Subhajit Kar(Inter-University Centre for Astronomy and Astrophysics, Pune - 411007, India), Lorant O. Sjouwerman(National Radio Astronomy Observatory, Socorro - 87801, USA) |
| Abstract: Precise distances to the stellar populations in the inner galactic region offer insights into the structure and early evolution of the Milky Way. However, direct astrometric distances remain absent or highly obscure for the majority of these sources due to severe extinction and complex variability. In this work, we construct a supervised machine-learning model employing the XGBoost regression framework to estimate reliable distances to the O-rich asymptotic giant branch (AGB) stars chosen from the AKARI mid-infrared (mid-IR) survey. The optimized model achieves R²≅0.98 with ~6% fractional accuracy and a typical absolute error of a few hundred parsecs. Applying this model to the whole AKARI AGB catalog yields distance estimates for over 36,000 AGB candidate sources, greatly expanding distance data for populations present in the dust-obscured regions of the Galaxy. The statistical distance estimates agree well with earlier reported distances to the Galactic Mira and those derived from period-luminosity relationships. Employing the updated distance data, we subsequently investigate the spatial distribution of Mira variables within the Galactic bulge and disk. Young Miras with longer periods are closer to the Galactic spiral arms and trace the bulge's barred morphology, while the older ones with shorter periods occupy the vertically extended regions. Our results imply that IR photometry-derived distances to the Mira variables provide a novel way to efficiently trace the stellar populations in heavily obscured regions and recover large-scale structures of the Milky Way. |
