22 January 2025: Andy Wilson
Speaker: Andy Wilson (Exeter)
Date: Wednesday 22 January 2025
Time: 15:00
Location: Physics 3.21
Explainable Machine Learning for Finding Young Stars
Our knowledge of star formation in the nearby Milky Way has been largely based on the concentrations of young stars found in clusters and large star forming complexes. Recent developments in instrumentation and computing allow a broader picture to be developed. Large numbers of sources can be efficiently categorised by running machine learning on new large-scale ground and space based photometric surveys. While a new generation of multi-object spectrographs will conduct surveys of the sky, recording the spectra of millions of objects. In this seminar, I shall present the results of a naïve Bayes classifier for identifying young stars in a broad region of the Galactic plane covering 1600 square degrees out to 2kpc. This form of Machine Learning is excellent at ranking objects into classifications, and its simple algorithm allows astrophysical insight into the reasons for the classifications. In the results we see halos of young stars around known young clusters, as well as hints of filamentary structures and connections between star forming regions. The WEAVE multiple object spectrograph being commissioned on the William Herschel Telescope, will obtain spectra of the candidate young stars identified by the classifier. These spectra will allow us to confirm the nature of our candidates and verify the structures we see in the spatial distribution of young stars.