Fast Space Mapping with Variable Weight Coefficients for Microwave Device Modeling

Abstract

We describe an improvement of a recent space mapping (SM) modeling approach that uses variable weight coefficients (SM-VWC). Our modification alleviates the main drawback of SM-VWC: the computational overhead related to a separate parameter extraction required for each evaluation of the surrogate model. In our new procedure, the output SM parameters of the surrogate model are obtained by solving a regression problem instead of being determined in the parameter extraction process. This dramatically reduces the evaluation time of the surrogate model. Moreover, the modeling accuracy of the modified technique is even better than the accuracy of the original SM-VWC approach. Examples demonstrate the robustness of our approach.