Machine Learning G2 Geometry

Explore a seminar talk from the New Technologies in Mathematics series where Elli Heyes from Imperial College discusses the application of machine learning techniques to G2 geometry. Learn how neural networks can be used to understand topological properties of holonomy G2 manifolds, which are important in M-theory as descriptions of extra spatial dimensions. Discover the recent developments in applying machine learning to these manifolds, which began in 2024, compared to similar techniques that have been used for Calabi-Yau manifolds since 2017. The presentation covers both the theoretical foundations and potential applications of numerically approximating metrics on compact holonomy G2 manifolds using machine learning approaches, with particular emphasis on their usefulness in M-theory.