Design of artificial neural networks with a stochastic graph-L-system

Author

Voigt, H.-M. ; Born, J. ; Santibáñez-Koref, I.

Author_Institution

Tech. Univ. Berlin, Bionics & Evolution Techniques Lab., Germany

fYear

1993

fDate

22-23 Apr 1993

Abstract

A discussion is given on the modelling of the individual development process as part of biological evolution. The purpose is to use the model to solve difficult optimization problems. The authors deal with the structuring of artificial neural networks (ANN). Among the natural models used to process information that of ANN has become one of the most used. The authors describe a model of individual development based on Lindenmayer-systems (L-systems) (A. Lindenmayer, 1968), the building blocks in cascades algorithm (BBC-algorithm). They give a detailed description of their approach and a short description of the evolutionary framework (the BBC-EVO-algorithm), a version of the BBC-algorithm embedded in an evolutionary process. Some exemplary results are given in order to describe the performance of the BBC-EVO-algorithm

Keywords

neural nets; optimisation; stochastic processes; BBC-EVO-algorithm; Lindenmayer-systems; artificial neural networks; biological evolution; building blocks; evolutionary framework; individual development process; natural models; optimization problems; stochastic graph-L-system;

fLanguage

English

Publisher

iet

Conference_Titel

Grammatical Inference: Theory, Applications and Alternatives, IEE Colloquium on

Conference_Location

Colchester

Type

conf

Filename

243136