Reproducing Results of Galaxy Classification Paper

Overview

For a final-year undergraduate physics class, PHYS 449: Machine Learning in Physics, myself and four of my peers teamed up to work on classifying images of galaxies with ML. This is a well worn path as ML has time and time again showed to be an excellent tool at classifying galaxies based on their morphologies.

Specifically, part of the project was to successfully reproduce the results in a published paper. Our group chose Morphological classification of galaxies with deep learning: comparing 3-way and 4-way CNNs by Mitchell K. Cavanagh, Kenji Bekki and Brent A. Groves.

We were able to roughly reproduce the results, and even improve them in some areas, as well as further exploring their application through the use of the Galaxy10 SDSS dataset. I received a grade of 100% for my work on this final project and every assignment in the course.