On the search of the shape parameter in radial basis functions using univariate global optimization methods

In this paper we consider the problem of finding an optimal value of the shape parameter in radial basis function interpolation. In particular, we propose the use of a leave-one-out cross validation (LOOCV) technique combined with univariate global optimization methods, which involve strategies of g...

Full description

Saved in:
Bibliographic Details
Published inJournal of global optimization Vol. 79; no. 2; pp. 305 - 327
Main Authors Cavoretto, R., De Rossi, A., Mukhametzhanov, M. S., Sergeyev, Ya. D.
Format Journal Article
LanguageEnglish
Published New York Springer US 01.02.2021
Springer
Springer Nature B.V
Subjects
Algorithms
Analysis
Computer Science
Global optimization
Interpolation
Mathematical optimization
Mathematics
Mathematics and Statistics
Operations Research/Decision Theory
Optimization
Parameters
Radial basis function
Real Functions
Technology application
Global optimization
65K05
Shape parameter
65D05
RBF interpolation
90C26
65D15
Online AccessGet full text

Cover

More Information
Summary:In this paper we consider the problem of finding an optimal value of the shape parameter in radial basis function interpolation. In particular, we propose the use of a leave-one-out cross validation (LOOCV) technique combined with univariate global optimization methods, which involve strategies of global optimization with pessimistic improvement (GOPI) and global optimization with optimistic improvement (GOOI). This choice is carried out to overcome serious issues of commonly used optimization routines that sometimes result in shape parameter values are not globally optimal. New locally-biased versions of geometric and information Lipschitz global optimization algorithms are presented. Numerical experiments and applications to real-world problems show a promising performance and efficacy of the new algorithms, called LOOCV-GOPI and LOOCV-GOOI, in comparison with their direct competitors.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0925-5001
1573-2916
DOI:10.1007/s10898-019-00853-3