Vine creeping algorithm for global optimisation

Christopher Neil Young, Ju Jia Zou, Chin Jian Leo

    Research output: Chapter in Book / Conference PaperConference Paperpeer-review

    2 Citations (Scopus)

    Abstract

    ![CDATA[This paper presents a novel vine creeping optimisation algorithm based on the integration of the Levenberg-Marquardt algorithm into a revised non-revisiting genetic algorithm. The global search of the genetic algorithm is enhanced in efficiency and accuracy by incorporating the Levenberg-Marquardt algorithm into the selection and mutation process. The term revisit is redefined as a local region of convergence by the Levenberg-Marquardt algorithm, rather than a particular point selected. The redefinition of a revisit allows a larger step size in mutation hence reducing the number of evaluations in order to flag the current space as saturated. The effect of the revisited regions filling out the current local minimum regions and branching into unvisited space results in the vine creeping effect. The proposed algorithm was tested against three well known benchmark functions, and was able to converge upon the global optimum within an average of 63.91 generations, with a success rate ranging between 96-100%.]]
    Original languageEnglish
    Title of host publicationProceedings of the 2010 Second World Congress on Nature and Biologically Inspired Computing, NaBIC 2010, Kitakyushu International Conference Center, Kitakyushu, Japan, December 15-17, 2010
    PublisherIEEE
    Pages455-459
    Number of pages5
    ISBN (Print)9781424473762
    DOIs
    Publication statusPublished - 2010
    EventWorld Congress on Nature & Biologically Inspired Computing -
    Duration: 15 Dec 2010 → …

    Conference

    ConferenceWorld Congress on Nature & Biologically Inspired Computing
    Period15/12/10 → …

    Fingerprint

    Dive into the research topics of 'Vine creeping algorithm for global optimisation'. Together they form a unique fingerprint.

    Cite this