I did a seminar at CCRMA, Stanford University, on [PINN (Physics Informed Neural Network) for Physical Modeling].
PINN is recently ongoing research on using a neural network as a differential equation solver by explicitly integrating the differential equation itself and the boundary conditions into the loss function.
Recently, PINN and its variants have gotten attention mostly for fluid modeling in mechanical engineering.
And this year, a research team at Postech, South Korea, published "A Physics-informed and data-driven deep learning approach for wave propagation and its scattering characteristics," which is about using PINN to model sound field propagation and scattering.
1. introduced the concept of PINN,
2. reviewed the important prior studies on PINN,
3. shared the extensive list of recent papers on various applications of PINN,
4. also briefly introduced the Fourier Neural Operator.
