AI and the World's Hard Problems · Physics · arXiv · 2026-10-26
Neural network enhancement of the Born approximation
The work addresses the inverse scattering problem for a penetrable medium, aiming to recover the refractive index quickly from multistatic far-field data at one frequency. The authors present a neural network that sharpens reconstructed images and two networks motivated by the corrected Born method that avoid inverting the regularized Born operator. Computational results show that these corrections to the Born approximation can give an accurate and flexible approach.