A New Method of Centers Location in Gaussian RBF Interpolation Networks

Autor:	Marek Bazan, Ewa Skubalska-Rafajłowicz
Rok vydání:	2013
Předmět:	symbols.namesake Mathematical optimization Nearest-neighbor interpolation Latin hypercube sampling Gaussian symbols Power function Algorithm Upper and lower bounds Hierarchical RBF Multivariate interpolation Mathematics Interpolation
Zdroj:	Artificial Intelligence and Soft Computing ISBN: 9783642386572 ICAISC (1)
DOI:	10.1007/978-3-642-38658-9_2
Popis:	In this paper we present a new method of obtaining near-optimal points sets for interpolation by Gaussian radial basis functions networks. The method is based on minimizing the maximal value of the power function. The power function provides an upper bound on the local RBF interpolation error. We use Latin hypercube designs and a space-filling curve based space-filling designs as starting points for the optimization procedure. We restrict our attention to 1-D and 2-D interpolation problems. Finally, we provide results of several numerical experiments. We compare the performance of this new method with the method of [6].
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::0944129dab99a8fd2560b849aba35579 https://doi.org/10.1007/978-3-642-38658-9_2 Zobrazit plný text záznamu