Estimating galaxy properties & redshifts using Machine learning

Kumar, Vikrant (2022) Estimating galaxy properties & redshifts using Machine learning. Masters thesis, Indian Institute of Science Education and Research Kolkata.

[img] Text (MS dissertation of Vikrant Kumar (17MS089))
17MS089_Thesis_file.pdf - Submitted Version
Restricted to Repository staff only

Download (2MB)
Official URL:


Star formation properties are principal agents that help us describe the formation and evolution of galaxies as a function of cosmic epoch. The redshift in their spectra is another crucial parameter that helps us determine distances. It is also routinely used to understand the expansion of the universe and the evolution of galaxies. Estimating these parameters through photometry has become an indispensable tool in extragalactic astronomy since imaging surveys are being performed faster than followup spectroscopy. Traditionally, stellar population synthesis(SPS) models have been used to obtain the best-fit parameters to characterize the redshifts & star formation in galaxies. As vast amounts of multi-temporal & multi-wavelength flux measurements are being collected every day for thousands of galaxies, alternative approaches which offer automated exploration tools to detect, classify and characterize objects are being developed. In this work, we develop machine learning and deep learning techniques to predict three important star formation properties – stellar mass, star formation rate and dust luminosity using the Galaxy & Mass Assembly(GAMA) Data Release 3(DR3). We also compute photometric redshifts(photo-z) for galaxies in GAMA DR3 and X-ray selected sources in Stripe 82X using the publicly available photometric and spectroscopic catalogs. We also apply a data interpolation technique using gradient boosted decision trees that enable us to perform accurate predictions on samples with missing information. Finally, we calculate the relative contribution of each band flux for the final predictions. We find that the visible and near-infrared bands significantly impact redshift, stellar mass and star formation rate prediction. On the other hand, the Dust Mass estimation depends on far-infrared bands and seems quite robust to other flux variations.

Item Type: Thesis (Masters)
Additional Information: Supervisor: Dr. Abhirup Datta, IIT Indore
Uncontrolled Keywords: Galaxy Properties; Machine Learning; Redshifts; Star Formation
Subjects: Q Science > QC Physics
Divisions: Department of Physical Sciences
Depositing User: IISER Kolkata Librarian
Date Deposited: 14 Sep 2023 06:15
Last Modified: 14 Sep 2023 06:15

Actions (login required)

View Item View Item