deepxde

Why we included this project

DeepXDE is a practical tool for anyone whose physics problems need a neural network that respects the governing equations. Engineers and computational scientists can define forward and inverse problems, set boundary conditions, and train on TensorFlow, PyTorch, or JAX through one consistent API. The library goes beyond the standard PINN workflow, covering DeepONet operator learning, multi-fidelity data, and fractional or stochastic PDE variants, so you are not locked into a single textbook case. Because the methods are tied to published papers and the documentation is thorough, it is a solid choice when you need to reproduce published results or defend your approach in peer review.