Recent Advance in Neuro- Transfer Function-Assisted Yield-Driven EM Optimization

Abstract

Yield-driven electromagnetic (EM) optimization is a key component in microwave design due to fabrication tolerances and manufacturing uncertainties. Neuro-transfer function (neuro-TF) methods hold great potential to accelerate the overall yield optimization process by replacing the EM-based fine models with well-trained neuro- TF surrogates. This paper provides an overview of recent advance in neuro- TF -assisted yield-driven EM optimization, with a focus on the recently reported adaptively weighted yield-driven EM optimization incorporating neuro-TF surrogate. A four-pole waveguide filter example is used to demonstrate the advantages of this advanced approach.