Speaker
Description
Asymmetries in a galaxy's neutral atomic hydrogen (HI) distribution and kinematics are key physical indicators of ongoing gas accretion or removal, which drives galaxy evolution. However, current classification methods - both parametric (e.g tilted-ring models) and non-parametric (e.g Asymmetry indices) often fail to reliably identify morphologically disturbed galaxies. Given the sheer number of sources detected by MeerKAT and the upcoming SKA surveys, the current reliance on time-consuming manual modeling and visual confirmation severely limits data analysis at scale, creating a challenge for modern astronomy. This project aims to overcome this limitation by developing an objective and scalable Machine Learning (ML) framework for automated morphology classification.
We will introduce our current sample of resolved HI moment maps from the 12 clusters of the MGCLS survey. Thereafter, we will show how the various non-parametric classification methods (Asymmetry, Smoothness, Concentration, Gini and M20) compare with each other, assessing how well their numerical results align with the visually inspected morphologies (the shapes) of the gas structures. This comparison will lay the foundation for reliable and high-quality training dataset for developing and optimizing machine learning models designed for the automated classification of galaxy morphologies.
| Stream | Science or Engineering |
|---|