Machine learning the universe

Using machine learning methods to search the vast space of Calabi-Yau manifolds for ones that predict the Standard Model from string theory.

String theory is the most promising candidate for a description of our universe that combines gravity with the Standard Model. It predicts 10 dimensions, four of which are space-time. The remaining six are folded up at a tiny scale, and their geometry determines the properties of our universe. A natural candidate for these extra dimensions are Calabi-Yau manifolds—compact, Ricci flat Kähler geometries. However, there are a vast number of such manifolds, a challenge known as the string vacuum degeneracy problem. Selecting the right one of the key puzzles to fundamental physics today.

In this project, we seek the Calabi-Yau geometry that generates the Standard Model. An exhaustive search of Calabi-Yau manifolds is computationally infeasible. What’s more, there is no theoretical principle to hone in on the correct geometry. To circumvent this, we use machine-learning methods, informed by data compiled by practitioners over recent years.

This programme, in which London Institute scientists are pioneers, will help find a selection principle for the right geometry, speed up the computation, and uncover new patterns in the space of geometries. It will also raise new conjectures in pure mathematics and provide new bench-marks for machine learning.

Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe
Machine learning the universe