ew 18: e18

Research Article

GPGPU based Multi-hive ABC Algorithm for Constrained Global Optimization Problems

Download10 downloads
  • @ARTICLE{10.4108/eai.13-7-2018.163156,
        author={Sandeep U. Mane and Amol C. Adamuthe and Aprupa S. Pawar},
        title={GPGPU based Multi-hive ABC Algorithm for Constrained Global Optimization Problems},
        journal={EAI Endorsed Transactions on Energy Web: Online First},
        volume={},
        number={},
        publisher={EAI},
        journal_a={EW},
        year={2020},
        month={2},
        keywords={Multi-hive ABC algorithm, GPGPU, Coarse-grained model, Constrained benchmark functions, Optimization problems},
        doi={10.4108/eai.13-7-2018.163156}
    }
    
  • Sandeep U. Mane
    Amol C. Adamuthe
    Aprupa S. Pawar
    Year: 2020
    GPGPU based Multi-hive ABC Algorithm for Constrained Global Optimization Problems
    EW
    EAI
    DOI: 10.4108/eai.13-7-2018.163156
Sandeep U. Mane1,*, Amol C. Adamuthe2, Aprupa S. Pawar3
  • 1: Dept. of CSE, Rajarambapu Institute of Technology (affiliated to Shivaji University Kolhapur), Rajaramnagar, Sangli Dist., MH, India
  • 2: Dept. of CS&IT, Rajarambapu Institute of Technology (affiliated to Shivaji University Kolhapur), Rajaramnagar, Sangli Dist., MH, India
  • 3: Dept. of CSE, Walchand College of Engineering, Sangli, (affiliated to Shivaji University Kolhapur), MH, India
*Contact email: manesandip82@gmail.com

Abstract

INTRODUCTION: The artificial bee colony (ABC) algorithm is a nature-inspired technique used for solving different optimization problems. This paper presents a multi-hive ABC algorithm for solving constrained benchmark functions of CEC2006. The CEC2006 data set contains the global benchmark functions with different design variables, number and type of constraints.

OBJECTIVES: The objective of the proposed work is to design and apply the GPGPU based multi-hive ABC algorithm to solve constrained optimization problems.

METHODS: The proposed approach is a multi-population coarse-grained system implemented using General Purpose Graphics Processing Unit (GPGPU). The performance of the proposed approach is compared with the serial ABC algorithm for eleven benchmark functions and results in the literature. The multi-hive ABC algorithm has multiple hives, each running separate ABC algorithm on different cores of GPGPU.

RESULTS: The proposed approach provides global best solutions in significantly reduced time for all benchmark functions. The speed-up obtained is approximately 7X to 9X. The GPGPU device utilization is approximately 57% to 91%.

CONCLUSION: The GPGPU based multi-hive ABC algorithm is a found good with respect to best results, speed up factor and GPU utilization to solve constrained optimization problems.