Conceptual

Neural-Network Solution of Einstein Equations for Gravity Compactifications

Represents the metric of a warped gravity compactification with a neural network and trains it by minimizing the residual of the Einstein equations, turning the construction of vacua into an optimization problem. Demonstrated by building negatively curved Einstein metrics on three-manifolds obtained by Dehn-filling the cusps of hyperbolic manifolds, as a route toward higher-dimensional Einstein metrics and de Sitter compactifications of M-theory.