Triangulating polytopes
AI-assisted maths
Generating triangulations and fibrations with reinforcement learning
Arxiv (2024)
Reinforcement learning produces fine, regular, star triangulations of reflexive polytopes in four dimensions that result in smooth Calabi-Yau threefolds. The approach can be used to search for threefolds satisfying desirable string theory compactification conditions. More refined properties such as stable vector bundles on the threefolds, as well as fibration structure can be generated by modifying the cost function.
Arxiv (2024)