Countering poisonous inputs with memetic neuroevolution

Julian Togelius, Tom Schaul, Jürgen Schmidhuber, Faustino Gomez

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

    Abstract

    Applied to certain problems, neuroevolution frequently gets stuck in local optima with very low fitness; in particular, this is true for some reinforcement learning problems where the input to the controller is a high-dimensional and/or ill-chosen state description. Evidently, some controller inputs are "poisonous", and their inclusion induce such local optima. Previously, we proposed the memetic climber, which evolves neural network topology and weights at different timescales, as a solution to this problem. In this paper, we further explore the memetic climber, and introduce its population-based counterpart: the memetic ES. We also explore which types of inputs are poisonous for two different reinforcement learning problems.

    Original languageEnglish (US)
    Title of host publicationParallel Problem Solving from Nature - PPSN X - 10th International Conference, Proceedings
    Pages610-619
    Number of pages10
    DOIs
    StatePublished - 2008
    Event10th International Conference on Parallel Problem Solving from Nature, PPSN X - Dortmund, Germany
    Duration: Sep 13 2008Sep 17 2008

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume5199 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Other

    Other10th International Conference on Parallel Problem Solving from Nature, PPSN X
    CountryGermany
    CityDortmund
    Period9/13/089/17/08

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Fingerprint Dive into the research topics of 'Countering poisonous inputs with memetic neuroevolution'. Together they form a unique fingerprint.

    Cite this