CSI-Based Physics-Informed Neural Network for Solving Inverse Scattering Problems

In this study, a contrast source inversion (CSI) based physics-informed neural network method is proposed for solving inverse scattering problems. The proposed method constructs two sets of neural networks to predict the total fields and relative permittivity distribution of the scatterer. Furthermore, state and data equations of CSI are embedded into the loss function as physical constraints to guide network training. Numerical experiments demonstrate that the proposed method achieves higher accuracy compared to traditional numerical approaches, all without requiring a large dataset.