DNNsolver: Accurate and Efficient Deep Learning Modeling of Polarization-Sensitive Diffractive Neural NetworksNew

This paper presents DNNsolver, a deep learning framework for accurate and efficient modeling of polarization-sensitive diffractive neural networks (DNNs). Diffractive neural networks composed of cascaded diffractive layers provide a promising platform for high-speed optical computing. The work addresses the limitation of conventional thin element approximation (TEA) models that fail to capture electromagnetic coupling within layers, while full-wave simulations like finite-difference time-domain (FDTD) methods are computationally expensive.