The extrapolated interval global optimization algorithm

摘要

This paper presents a new approach based on extrapolation to accelerate thelinear convergence process of Vectorized Moore-Skelboe (VMS) algorithm. The VMS is amodified version of basic Moore-Skelboe (MS) algorithm, where the vectorization is used asa means to speed up the basic MS algorithm. We propose to further accelerate the converg-ing process of VMS from linear to quadratic by combining the Richardson extrapolationtechnique with VMS. The effectiveness of the proposed algorithm is tested on various mul-tivariate examples and compared with the unaccelerated conventional method, i.e., MS andwell-known optimization software GlobSol. The test results show that the proposed extrapo-lation-based VMS offer considerable speed improvements over both the existing algorithms.