Small neural networks solve physics equations better with smarter architecture

Original: Small Models, Strong Priors: Architectural Inductive Bias for Parameter-Efficient Neural PDE Solvers

Source: arxiv.org ↗