Speaker
Description
Understanding the structural diversity of galaxies is essential for investigating their evolution and addressing the impact of source structure on celestial reference frames, where intrinsic variability and extended structures can significantly influence positional accuracy and stability.
This study uses a Convolutional Neural Network (CNN) to automate the classification of galaxy morphologies using high-resolution K-band (24 GHz) imaging data. With a threefold improvement in interferometer resolution compared to the standard S/X bands, K-band observations enable more accurate and stable reference frames, reducing astrophysical systematics such as core shifts and extended emission, while supporting precise morphological analysis.
Our CNN architecture integrates hierarchical layers that detect patterns, reduce data complexity, and classify morphologies, such as compact, weak extended emissions, elongated, weak second component, strong second component, weak third component and complex morphologies. This program addresses challenges like source variability and extended structure effects, which significantly impact celestial reference frames (CRFs).
This project represents a step forward in applying machine learning to astronomical imaging, providing an automated and scalable solution for analyzing large datasets. Preliminary results indicate effective classification performance, reducing manual effort and paving the way for more precise astrophysical analyses. The findings have implications for both galaxy evolution studies and improving the stability of celestial reference frames through accurate source modeling and classification.
| Stream | Science |
|---|