GAC-PINN: Geometry-Adaptive and Constraint-Enhanced Physics-Informed Neural Networks

GAC-PINN is a physics-informed neural network framework for physical systems featuring steep gradients, sharp interfaces, or severe spatio-temporal coupling. The proposed approach includes a gradient-driven adaptive grid mapping designed for diffeomorphic point concentration, together with constraint-enhancement mechanisms. It is intended to improve the accuracy and convergence of PINNs in challenging spatial and temporal physical regimes.