Evolution of Fisher´s discriminants

Author

Sierra, A.

Author_Institution

E.T.S. de Inf., Univ. Autonoma de Madrid, Spain

Volume

2

fYear

2001

fDate

2001

Firstpage

747

Abstract

Series expansions on a fixed set of basis functions have a neat advantage over neural-like modeling: the absence of local minima. Polynomial regression constitutes the paradigm. Even when non-linear basis functions are used, the coefficients of the expansion are uniquely specified and easily calculated. However, the number of adjustable coefficients is not controlled by the complexity of the problem but by the input dimension. In this paper, an evolutionary approach is proposed as a means of alleviating this drawback. As a proof of concept, a genetic algorithm is used to evolve the inputs used to construct Fisher´s discriminants. A varied group of UCI datasets is used to show that the evolved models perform 30% better than the discriminants constructed with the whole set of inputs

Keywords

genetic algorithms; neural nets; series (mathematics); statistical analysis; Fisher discriminants; UCI datasets; adjustable coefficients; dimensionality; evolutionary approach; genetic algorithm; learning machine; neural network model; pattern classification; polynomial regression; regression techniques; series expansions; Classification algorithms; Genetic algorithms; Linearity; Machine learning; Neural networks; Pattern classification; Petroleum; Polynomials; Scalability;

fLanguage

English

Publisher

ieee

Conference_Titel

Evolutionary Computation, 2001. Proceedings of the 2001 Congress on

Conference_Location

Seoul

Print_ISBN

0-7803-6657-3

Type

conf

DOI

10.1109/CEC.2001.934264

Filename

934264