A hybrid RBF based method for highly multidimensional response surfaces using scarce data sets Conference

Colaço, MJ, Dulikravich, GS. (2008). A hybrid RBF based method for highly multidimensional response surfaces using scarce data sets . 10.2514/6.2008-5892

cited authors

  • Colaço, MJ; Dulikravich, GS

authors

abstract

  • A number of interpolation functions, including some radial basis functions, were evaluated for accuracy and computing time requirements when fitting high dimensional response surfaces. In addition, a hybrid method was developed that uses the best radial basis function for each response surface topology. A set of non-trivial highly noisy response surfaces were fitted by each of these methods and hybrid method was found to be the most reliable and the most accurate. However, the computing cost of the hybrid algorithm rapidly increases with the dimensionality of the surface to be fitted thus limiting the practical utility of this hybrid method to response surfaces with less than approximately 100 dimensions. Copyright © 2008 by Colaco, M. J. and Dulikravich, G. S.

publication date

  • January 1, 2008

Digital Object Identifier (DOI)

International Standard Book Number (ISBN) 13