Bayesian Learning for Uncertainty Quantification, Optimization, and Inverse Design

Abstract

Design of microwave circuits require extensive simulations, which often take significant computational time due to design complexity. This can be addressed through neural networks (NNs) that provide predictive capability. Predictions often come with uncertainties that need to be quantified. Moreover, optimization and inverse designs are better done using probabilities. This article describes the use of Bayes theorem and machine learning (ML) for solving complex microwave design problems.