Abstract

In this talk, we propose a multi-frequency adaptive Bayesian inference framework for the inverse biharmonic scattering problem using limited-aperture far-field data. A Sobolev prior with a power-law decay of the Fourier coefficient variances is introduced to enforce boundary smoothness, thereby effectively suppressing high-frequency oscillations in reconstructions under limited-aperture conditions. To eliminate the reliance on manual parameter tuning and initial guesses, we employ a Robbins-Monro adaptive mechanism that dynamically adjusts the proposal step size to attain a target acceptance rate during the burn-in phase. We further establish the well-posedness of the Bayesian posterior distribution under the Hellinger metric. Numerical experiments demonstrate the effectiveness and feasibility of the proposed method for the limited-aperture inverse problem.