Machine-learned Calabi-Yau metrics and curvature (Q6572091)

The authors apply machine learning techniques to investigate numerical Ricci-flat metrics on smooth and singular K3 surfaces and Calabi-Yau threefolds, specifically the Cefalú family of quartic twofolds and the Dwork family of quintic threefolds.\N\NThey study the numerical stability of the numerically computed topological invariants such as the Euler characteristics and the Chern classes, suggesting a different neural network model, called spectral networks. They also apply persistent homology methods.