New AI Applications to Large Astronomical Data Sets

One of the pivotal challenges in astrophysics for the next decade and onward will be deciphering the enormous influx of data from new space and ground-based telescopes, surpassing the complexity of all existing astronomical data several times over. For scientific projects that produce huge data volumes, often referred to as "big science" (e.g., SKA, CERN), AI-driven decisions are increasingly necessary to replace human decisions at multiple points within scientific analyses.
The Euclid Space Telescope (launched in July 2023) is set to obtain imaging for hundreds of millions of objects and is the most recent telescope which underscores this challenge and one of many that leads to the SKA. Data from Euclid is now being taken and soon will exceed the ability of nearly all current methods of analysis. Thus, new machine learning tools within the decision-making framework, classifying objects-galaxies, stars, quasar, image defects, among many others will soon be needed.

Our objectives in this PhD project within the AI CDT extends beyond just classification. We aim to leverage this information to unravel galaxy formation processes and predict the features of real objects - stars, galaxies, etc.. The proposed interdisciplinary project merges Manchester's expertise in astronomy and machine learning in a new way. As part of this, we will develop leading-edge probabilistic machine learning methods, leveraging uncertainty in a statistical manner to drive the exploration of new parameter spaces and promote scientific discovery.

Previously, supervised and unsupervised ML on galaxy images has led to the discovery of unique galaxy classes previously unidentified. We are delving into the optimal utilization of methods like VQ-VAE for both training and discovering the composition of galaxies. However, this is just the start as other ML tools need to be applied to this problem, for which a PhD working in this CDT will do. While ML approaches in astronomy have...