A Wavelet-based Encoding for Neuroevolution

S. van Steenkiste*, J. Koutnik, K. Driessens, J. Schmidhuber

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingAcademicpeer-review

6 Citations (Web of Science)

Abstract

A new indirect scheme for encoding neural network connection weights as sets of wavelet-domain coefficients is proposed in this paper. It exploits spatial regularities in the weight-space to reduce the genspace dimension by considering the low-frequency wavelet coefficients only. The wavelet-based encoding builds on top of a frequency-domain encoding, but unlike when using a Fourier-type transform, it offers gene locality while preserving continuity of the genotype-phenotype mapping. We argue that this added property allows for more efficient evolutionary search and demonstrate this on the octopus-arm control task, where superior solutions were found in fewer generations. The scalability of the wavelet-based encoding is shown by evolving networks with many parameters to control game-playing agents in the Arcade Learning Environment.
Original languageEnglish
Title of host publicationProceedings of the Genetic and Evolutionary Computation Conference
PublisherAssociation for Computing Machinery (ACM)
Pages517-524
Number of pages8
ISBN (Print)9781450342063
DOIs
Publication statusPublished - 2016
EventGenetic and Evolutionary Computation Conference (GECCO) - Denver, United States
Duration: 20 Jul 201624 Jul 2016
http://gecco-2016.sigevo.org/index.html/HomePage.html

Conference

ConferenceGenetic and Evolutionary Computation Conference (GECCO)
Abbreviated titleGECCO 2016
Country/TerritoryUnited States
CityDenver
Period20/07/1624/07/16
Internet address

Keywords

  • Neuroevolution
  • indirect encoding
  • wavelets
  • gene-locality

Cite this