top-1

top-2 top-3

top-4 top-5

menutop

   Program

 

   Committee

 

   Author Index

 

   Search

 

   About GECCO

 

   CD Tech Support

menubot2

 

 

 

 

Session:

Workshop - Complexity through Development and Self-Organizing Representations (CODESOAR)

Title:

Effect of Encoding on the Evolvability of an Embodied Neural Network

 

 

Authors:

Chad W. Seys
Randall D. Beer

 

 

Abstract:

The evolvability of a neural network controller for a hexa- pod agent encoded directly and symmetrically is examined. The symmetric encoding imposes a structural regularity on the neural network and decreases the size of genotype space relative to the direct encoding. The symmetrically encoded neural networks are found to be more evolvable than the di- rectly encoded neural networks, but it is unknown whether structural regularity or decreased size of the genotype is more important. To test whether structural regularity is more important than genotype size, the architecture of the neural network is manipulated to increase the genotype size of the symmetric encodings so that they are larger than the directly encoded genotypes. These symmetric encodings are still found to be more evolvable than the direct encodings despite having a larger genotype. In these experiments it is the encoding which determines evolvability more than size of genotype space.

 

 

CD-ROM Produced by X-CD Technologies