Evolving neural networks in compressed weight space

Jan Koutník, Faustino Gomez, Jürgen Schmidhuber

Research output: Chapter in Book/Report/Conference proceedingConference contribution

62 Scopus citations

Abstract

We propose a new indirect encoding scheme for neural networks in which the weight matrices are represented in the frequency domain by sets of Fourier coefficients. This scheme exploits spatial regularities in the matrix to reduce the dimensionality of the representation by ignoring high-frequency coefficients, as is done in lossy image compression. We compare the efficiency of searching in this "compressed" network space to searching in the space of directly encoded networks, using the CoSyNE neuroevolution algorithm on three benchmark problems: pole-balancing, ball throwing and octopusarm control. The results show that this encoding can dramatically reduce the search space dimensionality such that solutions can be found in significantly fewer evaluations. Copyright 2010 ACM.
Original languageEnglish (US)
Title of host publicationProceedings of the 12th Annual Genetic and Evolutionary Computation Conference, GECCO '10
Pages619-625
Number of pages7
DOIs
StatePublished - Aug 27 2010
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2022-09-14

Fingerprint

Dive into the research topics of 'Evolving neural networks in compressed weight space'. Together they form a unique fingerprint.

Cite this