Injecting Physics Knowledge in Neural Networks for a Wide Range of Photonic Devices: QNM-Net

Paper i proceeding, 2025

We will present our work on QNM-Net, a physics-informed neural network for electromagnetic scattering. Our network in corporates physics knowledge through the quasi-normal-mode framework, which means it can model a vast range of (nano) photonic devices. By incorporating physics knowledge, we can achieve a tenfold reduction of training data needed.