Impact of Redundant Data on Evolution of Neural Networks

Date of Award

2005

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

MacDonald, Patrick

Keywords

Neural Networks, Genetic Algorithm, Degenerate Code

Area of Concentration

Mathematics

Abstract

Algorithms were developed and implemented to encode a population of neural networks as 'digital genomes'. After training, the population is tested, a high performance subpopulation is selected, and individuals of the selected subpopulation are bred. Subject to this routine, neural networks were observed to develop topologies specific to the problem under evaluation. Evolutionary outcomes were studied as a function of the degree of degeneracy in the genetic coding, by statistical analysis of over 4,000 evolutionary experiments.

Rights

This bibliographic record is available under the Creative Commons CC0 public domain dedication. The New College of Florida, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.

This document is currently not available here.

Share

COinS