Phil Breen

School of Mathematics, University of Edinburgh

Research

My research focuses on using machine learning techniques to tackle problems in theoretical astrophysics. Specifically exploiting the power of deep neural networks to accerelate expensive numerical simulations and to provide insights into the data produced by astrophysical simulations. I am also in the process of developing a Monte Carlo scheme powered by generative adversarial networks. I have also contributed to the statistical development of a pattern recognition algorithm in the field of statistical genetics in a collaborative project with researches at MRC Biostatistics Unit, University of Cambridge.

I have proposed a novel scenario to address what is arguably the greatest unsolved open problem concerning globular star clusters, that is the mysterious presence of multiple stellar populations. The origin of the stars enriched in light elements has many puzzling observational constraints and proves a major challenge for traditionally proposed enrichment mechanisms. I have suggested that light element enrichment could have been produced in accretion discs around stellar mass black holes (ironically usually seen as the sight of destruction rather than creation), the model has the potential to address some of the most difficult observational constraints (see the paper).

I am also involved in a collobrative project which focuses on the dynamical evolution of collisional stellar systems (e.g. globular and nuclear star clusters). I am particularly interested in the effect of a retained population of stellar mass black holes, internal rotation and velocity anisotropy, and dark matter.

For a full list of publications see ADS query or Google Scholar (or for my industry experience see my LinkedIn profile)