Neuro-Symbolic AI for Analytical Solutions of Differential Equations
SIGS is a neuro-symbolic framework that uses a context-free grammar to generate mathematical building blocks for differential equations. It combines this symbolic generation with numerical optimization in a latent space search to discover closed-form analytical solutions from governing equations, achieving high accuracy on various PDE benchmarks.