Tu3I: Advances in Computational Techniques and Machine Learning Methods for Microwave Applications

Presentation Abstract
This session presents recent advances in computational techniques, including machine-learning-enabled methods for microwave applications, with a focus on accelerating full-wave analysis and design. Contributions include rapid synthesis of training data for deep-learning surrogates, physics-informed neural operators for electromagnetic forward and inverse problems, and data-driven constitutive modeling within FDTD solvers. Novel numerical methods addressing ill-posed discrete Maxwell systems and tensor-train-accelerated FDTD with logarithmic computational cost are also featured. Together, these works highlight the integration of physics-based rigor, machine learning, and low-rank numerical techniques to enable fast, accurate, and scalable simulation and design of complex microwave systems.