My research focuses on differentiable physical simulation and applying machine learning techniques to enhance simulation. I am also interested in incorporating scientific knowledge to improve the performance of machine learning algorithms.
Learning to Estimate and Refine Fluid Motion with Physical Dynamics
Mingrui Zhang, Jianhong Wang, James Tlhomole, Matthew D Piggott
International Conference on Machine Learning (ICML), 2022 [Paper]
M2N: Mesh movement networks for PDE solvers
Wenbin Song*, Mingrui Zhang*, Joseph G Wallwork, Junpeng Gao, Zheng Tian, Fanglei Sun, Matthew D Piggott, Junqing Chen, Zuoqiang Shi, Xiang Chen, Jun Wang (*Equal contribution)
Advances in Neural Information Processing Systems (NeurIPS), 2022 [Paper]
E2N: error estimation networks for goal-oriented mesh adaptation
Joseph G Wallwork, Jingyi Lu, Mingrui Zhang, Matthew D Piggott
arXiv preprint, 2022 [Paper]
Unsupervised learning of particle image velocimetry
Mingrui Zhang, Matthew Piggott
International Conference on High Performance Computing (ISC), 2020 [Paper]