On the Representation of Solutions to Elliptic PDEs in Barron Spaces

Neural Information Processing Systems 

Numerical solutions to high-dimensional partial differential equations (PDEs) based on neural networks have seen exciting developments.