We are happy to host the Workshop on “Data-driven simulation and PDE learning using Physics-informed neural networks” (PINN) together with the Helmholtz-Zentrum Dresden-Rossendorf which takes place from 28th September to 2nd October 2020 in Görlitz.
The simulation of complex systems as well as the reconstruction of the state of that system from experimental data is a very time consuming task. Workshop leader Dr. Nico Hoffmann from HZDR and TUD together with 13 participants from different institutions will be approaching this challenge by Physics-informed neural network allowing us to approximate forward simulations as well as identify the governing equations and parameters based on experimental measurements.
The workshop is driven by the following question: are we able to either efficiently solve or recover the governing equations of certain complex systems in terms of Ordinary or Partial Differential Equations (ODE/PDE) by Physics-informed neural networks?
PINN basically provide all means for jointly solving the following tasks
- solving 1D/2D problems provided initial/boundary conditions + governing equations (PINN can be seen as implicit solver of ODE/PDEs by optimization,
- warm starting simulations by transfer learning (i.e. recovering the solution of a changed configuration),
- learning the potentially non-linear contribution of certain parameters to the solution of a complex model,
- identification of parameters and/or governing equations provided either full simulation- or experimental data.
The workshop consists of theory as well as hackathon sessions. The theory sessions introduce basic and advanced concepts of Physics-informed neural networks while the Hackathon sessions enable the participants to adapt the PINN-framework of HZDR’s Hoffmann lab to very specific research questions ranging from quantum mechanics to medical imaging and materials science.
P. Stiller, F. Bethke, M. Böhme, R. Pausch, S. Torge, A. Debus, J. Vorberger, M.Bussmann, N. Hoffmann: Large-scale Neural Solvers for Partial Differential Equations.
M. Raissi, P.Perdikaris, G. E. Karniadakis: Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations