learned-simulation

[ICLR23] First deep learning-based surrogate model that jointly learns the evolution model and optimizes computational cost via remeshing

Learned crash deformation prediction with graph neural networks: an independent, open-source PyTorch research workflow from FE simulation data to autoregressive rollout. MeshGraphNets and MeshGraphNet-Transformer (MGN-T), contact edges, fair architecture comparison. Validated on DeepMind's deforming_plate.