Statistical modeling of frequency responses using linear Bayesian vector fitting

Abstract

AbstractThis article presents a Bayesian extension of the vector fitting (VF) procedure for rational approximation of frequency‐domain responses. The proposed method treats the linear part of VF in a Bayesian way, while propagating distributions through the nonlinear part by sampling. As such, it is capable of providing data‐driven uncertainty information along with the rational fit. The Bayesian VF technique is applied to two realistic design examples, a double folded stub filter and a RAM memory channel, demonstrating its validity and highlighting three potential applications of this novel framework.