Deep Learning and Morphology Across Redshifts

Abstract

This paper addresses the question, “How does galaxy morphology differ across red shifts?” Interestingly enough, astronomers can peer through time simply by searching deeper into the universe for galaxies, as the further back one looks in time, the further back they are looking back in time. We utilized this property of physics to analyze galaxies from millions of years in the past to understand how they are structured. The data collection discussed in this paper analyzes galaxies drawn from databases and the statistics collected when running a Convolutional Neural Network (CNN) that is trained on the data set. A discussion of the distribution of morphologies across redshifts is also presented, drawn from the results of the CNN model. Afterwards, an analysis of our CNN model and various distributions are mentioned with our interpretation of the results. Lastly, a reflection about our answer to the research question is put forward with possible future steps to take.