Genetic Algorithm with Species for Regularization Network Metalearning

Authors: Roman Neruda, Petra Vidnerová
Publication year: 2010
Publication type: Konferenční příspěvek (mezinárodní konf.)
Periodical: Advances in Information Technology
Projects: Complexity of perceptron and kernel networks
Bibliographic citation: NERUDA, Roman, VIDNEROVÁ, Petra. Genetic Algorithm with Species for Regularization Network Metalearning. In: PAPASRATORN, B.-, LAVANGNANANDA, K., CHUTIMASKUL, W., VANIJJA, V., eds. Advances in Information Technology. Berlin: Springer, 2010, S. 192-201. Communications in Computer and Information Science, 114. ISBN 978-3-642-16698-3. ISSN 1865-0929
Dostupné z: doi:10.1007/978-3-642-16699-0_21
DOI: 10.1007/978-3-642-16699-0_21
ID: 356026

Abstract

Regularization networks are one of the important methods for supervised learning. They benefit from very good theoretical background, though their drawback is the presence of metaparameters. The metaparameters are typically supposed to be given by an user. In this paper, we develop a method for finding optimal values for metaparameters, namely type of kernel function, kernel’s parameter and regularization parameter. The method is based on genetic algorithms with different species for different kinds of kernels. The method is demonstrated on experiments.