The Classification of Optical Galaxy Morphology Using Unsupervised Learning Techniques

TL;DR

This paper explores unsupervised learning methods, including autoencoders and clustering algorithms, to classify galaxy morphology in large astronomical datasets, aiming to reduce reliance on manual labeling.

Contribution

It demonstrates the application of convolutional autoencoders combined with clustering techniques for galaxy classification, highlighting potential for scalable, automated morphological analysis.

Findings

Agglomerative clustering performed best among tested methods.

Unsupervised clustering showed promising results comparable to human classifications.

Optimizations could enable more accurate, automated galaxy morphology classification.

Abstract

In recent years, large scale data intensive astronomical surveys have resulted in more detailed images being produced than scientists can manually classify. Even attempts to crowd-source this work will soon be outpaced by the large amount of data generated by modern surveys. This has brought into question the viability of human-based methods for classifying galaxy morphology. While supervised learning methods require datasets with existing labels, unsupervised learning techniques do not. Therefore, this paper implements unsupervised learning techniques to classify the Galaxy Zoo DECaLS dataset. A convolutional autoencoder feature extractor was trained and implemented. The resulting features were then clustered via k-means, fuzzy c-means and agglomerative clustering. These clusters were compared against the true volunteer classifications provided by the Galaxy Zoo DECaLS project. The best results, in general, were produced by the agglomerate clustering method. However, the increase in performance compared to k-means clustering was not significant considering the increase in clustering time. After undergoing the appropriate clustering algorithm optimizations, this approach could prove useful for classifying the better performing questions and could serve as the basis for a novel approach to generating more "human-like" galaxy morphology classifications from unsupervised techniques.