Improving the Variational Learning of Physics-Driven Neural Generative Models

Explore recent advances in learning probabilistic emulators for parametric PDEs in this 25-minute talk by Arnaud Vadeboncoeur from the University of Cambridge. Delve into the importance of efficiently solving parametric PDEs for engineering applications such as model calibration, iterative design, and uncertainty quantification. Examine the work conducted by the TMCF on boundary condition enforcement for physics-informed neural network (PINN) based physics-driven deep latent variable models. Discover insights from the TMCF research that led to the paper "Random Grid Neural Processes for parametric differential equations." Gain valuable knowledge about improving variational learning techniques for physics-driven neural generative models, essential for engineers and researchers working with complex physical systems.

Improving the variational learning of physics driven neural generative models