Adaptive Basis Selection and Dimensionality Reduction with Bayesian Compressive Sensing

Abstract

Bayesian inference is implemented to obtain a polynomial chaos expansion for
the response of a complex model, given a sparse set of training runs. For
polynomial basis reduction, Bayesian compressive sensing is
employed to detect basis terms with strong impact on model output
via relevance vector machine. Furthermore, a recursive algorithm is
proposed to determine the optimal set of basis terms. The
methodology is applied to global sensitivity studies of the Community
Land Model.


Date
Apr 3, 2012
Location
Raleigh, NC