Learning what to ignore: Memetic climbing in topology and weight space

Julian Togelius, Faustino Gomez, Jürgen Schmidhuber

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

    Abstract

    We present the memetic climber, a simple search algorithm that learns topology and weights of neural networks on different time scales. When applied to the problem of learning control for a simulated racing task with carefully selected inputs to the neural network, the memetie climber outperforms a standard hill-climber. When inputs to the network are less carefully selected, the difference is drastic. We also present two variations of the memetie climber and discuss the generalization of the underlying principle to population-based neuroevolution algorithms.

    Original languageEnglish (US)
    Title of host publication2008 IEEE Congress on Evolutionary Computation, CEC 2008
    Pages3274-3281
    Number of pages8
    DOIs
    StatePublished - 2008
    Event2008 IEEE Congress on Evolutionary Computation, CEC 2008 - Hong Kong, China
    Duration: Jun 1 2008Jun 6 2008

    Publication series

    Name2008 IEEE Congress on Evolutionary Computation, CEC 2008

    Other

    Other2008 IEEE Congress on Evolutionary Computation, CEC 2008
    Country/TerritoryChina
    CityHong Kong
    Period6/1/086/6/08

    ASJC Scopus subject areas

    • Computational Theory and Mathematics
    • Theoretical Computer Science

    Fingerprint

    Dive into the research topics of 'Learning what to ignore: Memetic climbing in topology and weight space'. Together they form a unique fingerprint.

    Cite this