Hands On V

TorchPhysics - A deep learning library for solving differential equations

25.09.2023, 15:30 - 17:00
90 minutes
H 0.01 + 0.02
Max Participants:
Requires own laptop with Google Account.


Many problems in application and industry are concerned with solving partial differential equations. The goal of this workshop is to give an introduction into the software library TorchPhysics, which implements several deep learning methods for solving differential equations. We start with a short overview of different approaches and then focus on the concept of physics informed neural networks (PINNs) and the functionalities of TorchPhysics. To demonstrate the advantages and possibilities of the deep learning methods, TorchPhysics will, in a joined exercise, be used to compute the solutions of some exemplary equations.

Prerequisite: Basic Python knowledge, own Laptop and a Google account, since Google-Colab will be used for the joined exercise session. No other installations on your Laptop are required.

Tom Freudenberg
University of Bremen
Center for Industrial Mathematics
Janek Gödeke
University of Bremen
Mathematics and Computer Science: WG Technomathematics