The fl exi-PSO

Towards a more fl exible particle swarm optimizer

Muhammad Kathrada

    Research output: Contribution to journalArticle

    Abstract

    This paper presents some simple heuristics to increase fl exibility in particle swarms. The particle velocities are updated with an extended inertia weight formulation, particles are also allowed to act as hill-climbers and the population is divided into exploration particles which are allowed to roam the search space more freely and exploitation particles which try to improve the fi tness by fi ne tuning a local search, thus making the swarm much more fl exible in its behavior. These heuristics improve the traditional inertia weight PSO and show comparable performance in the computational results of benchmark functions to other state of the art techniques. The benchmark functions used are a rigorous test bed for evaluating any optimization algorithm. An introduction and general overview of approaches to particle swarm optimizers is fi rst presented, followed by a discussion of the proposed method. The result of the experiments is then discussed followed by proposals for future research directions. © Operational Research Society of India.

    Original languageEnglish
    Pages (from-to)52-68
    Number of pages17
    JournalOPSEARCH
    Volume46
    Issue number1
    DOIs
    Publication statusPublished - Mar 2009

    Fingerprint

    Particle swarm optimization (PSO)
    Tuning
    Experiments

    Keywords

    • Exploitation
    • Exploration
    • Extended particle swarm
    • Heuristics
    • Hill-climber

    Cite this

    Kathrada, Muhammad. / The fl exi-PSO : Towards a more fl exible particle swarm optimizer. In: OPSEARCH. 2009 ; Vol. 46, No. 1. pp. 52-68.
    @article{c6ec468b86b0453dafc659799de1a5fd,
    title = "The fl exi-PSO: Towards a more fl exible particle swarm optimizer",
    abstract = "This paper presents some simple heuristics to increase fl exibility in particle swarms. The particle velocities are updated with an extended inertia weight formulation, particles are also allowed to act as hill-climbers and the population is divided into exploration particles which are allowed to roam the search space more freely and exploitation particles which try to improve the fi tness by fi ne tuning a local search, thus making the swarm much more fl exible in its behavior. These heuristics improve the traditional inertia weight PSO and show comparable performance in the computational results of benchmark functions to other state of the art techniques. The benchmark functions used are a rigorous test bed for evaluating any optimization algorithm. An introduction and general overview of approaches to particle swarm optimizers is fi rst presented, followed by a discussion of the proposed method. The result of the experiments is then discussed followed by proposals for future research directions. {\circledC} Operational Research Society of India.",
    keywords = "Exploitation, Exploration, Extended particle swarm, Heuristics, Hill-climber",
    author = "Muhammad Kathrada",
    year = "2009",
    month = "3",
    doi = "10.1007/s12597-009-0004-3",
    language = "English",
    volume = "46",
    pages = "52--68",
    journal = "OPSEARCH",
    issn = "0030-3887",
    publisher = "Springer",
    number = "1",

    }

    The fl exi-PSO : Towards a more fl exible particle swarm optimizer. / Kathrada, Muhammad.

    In: OPSEARCH, Vol. 46, No. 1, 03.2009, p. 52-68.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - The fl exi-PSO

    T2 - Towards a more fl exible particle swarm optimizer

    AU - Kathrada, Muhammad

    PY - 2009/3

    Y1 - 2009/3

    N2 - This paper presents some simple heuristics to increase fl exibility in particle swarms. The particle velocities are updated with an extended inertia weight formulation, particles are also allowed to act as hill-climbers and the population is divided into exploration particles which are allowed to roam the search space more freely and exploitation particles which try to improve the fi tness by fi ne tuning a local search, thus making the swarm much more fl exible in its behavior. These heuristics improve the traditional inertia weight PSO and show comparable performance in the computational results of benchmark functions to other state of the art techniques. The benchmark functions used are a rigorous test bed for evaluating any optimization algorithm. An introduction and general overview of approaches to particle swarm optimizers is fi rst presented, followed by a discussion of the proposed method. The result of the experiments is then discussed followed by proposals for future research directions. © Operational Research Society of India.

    AB - This paper presents some simple heuristics to increase fl exibility in particle swarms. The particle velocities are updated with an extended inertia weight formulation, particles are also allowed to act as hill-climbers and the population is divided into exploration particles which are allowed to roam the search space more freely and exploitation particles which try to improve the fi tness by fi ne tuning a local search, thus making the swarm much more fl exible in its behavior. These heuristics improve the traditional inertia weight PSO and show comparable performance in the computational results of benchmark functions to other state of the art techniques. The benchmark functions used are a rigorous test bed for evaluating any optimization algorithm. An introduction and general overview of approaches to particle swarm optimizers is fi rst presented, followed by a discussion of the proposed method. The result of the experiments is then discussed followed by proposals for future research directions. © Operational Research Society of India.

    KW - Exploitation

    KW - Exploration

    KW - Extended particle swarm

    KW - Heuristics

    KW - Hill-climber

    U2 - 10.1007/s12597-009-0004-3

    DO - 10.1007/s12597-009-0004-3

    M3 - Article

    VL - 46

    SP - 52

    EP - 68

    JO - OPSEARCH

    JF - OPSEARCH

    SN - 0030-3887

    IS - 1

    ER -