Speaker
Description
Partial Differential Equations (PDEs) play an important role in describing continuous physical systems such as fluids, waves, cloth and many more. Thus, by solving these equations, one can simulate for example fluids or garment in computer graphics or analyse lift and drag coefficients in engineering applications. However, in most scenarios, analytic solutions of PDEs are not available and traditional numerical solutions are computationally expensive. In contrast, recent deep-learning based methods promise great gains in efficiency by infering solutions in a single forward pass through a neural network. Here, we want to present our physics-driven approach to train neural surrogate models for PDEs without precomputed ground truth data. This way, we achieve fast, stable and differentiable simulations that can be used for example in interactive real-time applications or in inverse problems.
