This mini-engineering project implements a 2D Poisson/anisotropic diffusion forward solver on a regular grid and an adjoint-state gradient for inverse coefficient (conductivity) estimation.

This repository investigates stability, ill-conditioning, and information flow in PDE-constrained learning methods for imaging inverse problems. We study when learning-based solvers succeed or fail at ...

Abstract: We introduce a computationally efficient image reconstruction method that accelerates microwave tomographic imaging while maintaining the accuracy of traditional methods. The new method ...