Research Division Seminar
Revealing Galaxy Morphology with Spectral Data and Unsupervised Techniques

Abstract

Using unsupervised machine learning methods, we present a novel approach to classifying galaxies into early and late types based on their spectral characteristics. The research utilizes a balanced dataset of 2000 galaxies from the Galaxy Zoo 2 and spectral data from the Sloan Digital Sky Survey Data Release 13. The methodology involves applying an Autoencoder Neural Network for dimensionality reduction, followed by a Gaussian Mixture Model for clustering. The study demonstrates that this approach achieves an accuracy rate of approximately 86% in galaxy classification, highlighting the potential of unsupervised machine learning techniques in enhancing the precision and efficiency of morphological classification of galaxies based on spectral data.