Genetic algorithms as a tool for the analysis of adaptive resonance theory network training sets

Author

Caudell, Thomas P.

Author_Institution

Boeing Comput. Services, Res. & Technol., Seattle, WA, USA

fYear

1992

fDate

6/6/1992 12:00:00 AM

Firstpage

184

Lastpage

200

Abstract

Genetic algorithms (GAs) are used to study the basins of attraction for adaptive resonances theory (ART) neural networks. Two GAs are used to study attractors formed by the `bottom-up´ and the `top-down´ subsystems of a single binary input/binary output ART (ART1) neural network. The author presents results based on a simple training data set consisting of two distributions of centered rectangles. These results will ultimately aid in the refinement of training sets being used in neural information retrieval applications

Keywords

genetic algorithms; learning (artificial intelligence); neural nets; ART1; adaptive resonance theory network training sets; attractors; basins of attraction; bottom-up; centered rectangles; genetic algorithms; neural information retrieval; single binary input/binary output ART neural net; top-down; Adaptive systems; Algorithm design and analysis; Costs; Design automation; Design engineering; Genetic algorithms; Group technology; Neural networks; Resonance; Subspace constraints;

fLanguage

English

Publisher

ieee

Conference_Titel

Combinations of Genetic Algorithms and Neural Networks, 1992., COGANN-92. International Workshop on

Conference_Location

Baltimore, MD

Print_ISBN

0-8186-2787-5

Type

conf

DOI

10.1109/COGANN.1992.273939

Filename

273939