Students from Highams Park and their Orbyts Fellow James Ray from UCL Physics and Astronomy delved into the fascinating world of galaxy redshift regression, aiming to improve upon traditional methods using machine learning techniques. The team utilised the photometry and morphological features of the galaxies to inform a decision tree to better predict the redshift (distance) of a galaxy!

The students trained a supervised machine-learning model on photometric and redshift data from the Kilo Degree Survey (KiDS). They also used elements of Explainable AI (XAI) to better understand the features that informed the system the most and found that the main features are the colour and morphological classification of the galaxy.

The results of the experiment were promising! The linear regression model successfully predicted the redshift of galaxies at lower redshifts with a high degree of accuracy, demonstrating the potential of machine learning in astronomical research. However, the model struggled at classifying galaxies at a higher redshift due to a lack of data in that region. But the students explored other regression-based models and benchmarked them against each other.

This research was motivated by the release of the next generation of telescopes, which will generate data at a rate many magnitudes greater than seen in the field. By automating the classification process, researchers can more efficiently analyse vast amounts of astronomical data, leading to a deeper understanding of the universe's structure and evolution!