Computation of the regression kernel matrix using semidefinite programming

Autor: Trafalis, T., Malyscheff, A.
Jazyk: ruština
Rok vydání: 2006
Zdroj: Вычислительные технологии.
ISSN: 1560-7534
Popis: In this paper we calculate the optimal kernel matrix for regression analysis problems from three basis matrices by employing a semidefinite programming formulation and thereby benefiting from the transductive nature of these techniques. Preliminary experimental results using standard benchmark data are presented for which the optimal parameters for a linear combination of three basis kernel matrices are retrieved.
Databáze: OpenAIRE