Physics-Informed Machine Learning of Optical Modes in Composites

We present machine learning techniques that incorporate physics into the training process. We demonstrate, on example of predicting light propagation in multilayered composites, that physics-informed models are significantly more robust than their black box counterparts.