Self-Supervised Representation Learning for Astronomical Images

TL;DR

This paper demonstrates that self-supervised learning on astronomical sky survey images can produce representations that outperform supervised methods in galaxy classification and redshift estimation, with fewer labeled data.

Contribution

The authors introduce a contrastive self-supervised learning framework for astronomical images that achieves superior performance with less labeled data compared to traditional supervised approaches.

Findings

Self-supervised representations outperform supervised models in galaxy morphology classification.

The approach achieves comparable accuracy to supervised models using 2-4 times fewer labels.

The method is effective for multiple scientific tasks, including redshift estimation.

Abstract

Sky surveys are the largest data generators in astronomy, making automated tools for extracting meaningful scientific information an absolute necessity. We show that, without the need for labels, self-supervised learning recovers representations of sky survey images that are semantically useful for a variety of scientific tasks. These representations can be directly used as features, or fine-tuned, to outperform supervised methods trained only on labeled data. We apply a contrastive learning framework on multi-band galaxy photometry from the Sloan Digital Sky Survey (SDSS) to learn image representations. We then use them for galaxy morphology classification, and fine-tune them for photometric redshift estimation, using labels from the Galaxy Zoo 2 dataset and SDSS spectroscopy. In both downstream tasks, using the same learned representations, we outperform the supervised state-of-the-art results, and we show that our approach can achieve the accuracy of supervised models while using 2-4 times fewer labels for training.