Algorithm Portfolio for Individual-based Surrogate-Assisted Evolutionary Algorithms

被引:2
作者
Tong, Hao [1 ]
Liu, Jialin [1 ]
Yao, Xin [1 ]
机构
[1] Southern Univ Sci & Technol, Dept Comp Sci & Engn, Univ Key Lab Evolving Intelligent Syst Guangdong, Shenzhen Key Lab Computat Intelligence, Shenzhen, Peoples R China
来源
PROCEEDINGS OF THE 2019 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE (GECCO'19) | 2019年
基金
国家重点研发计划;
关键词
Algorithm portfolio; surrogate model; evolutionary algorithm; computationally expensive problem;
D O I
10.1145/3321707.3321715
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Surrogate-assisted evolutionary algorithms (SAEAs) are powerful optimisation tools for computationally expensive problems (CEPs). However, a randomly selected algorithm may fail in solving unknown problems due to no free lunch theorems, and it will cause more computational resource if we re-run the algorithm or try other algorithms to get a much solution, which is more serious in CEPs. In this paper, we consider an algorithm portfolio for SAEAs to reduce the risk of choosing an inappropriate algorithm for CEPs. We propose two portfolio frameworks for very expensive problems in which the maximal number of fitness evaluations is only 5 times of the problem's dimension. One framework named Par-IBSAEA runs all algorithm candidates in parallel and a more sophisticated framework named UCB-IBSAEA employs the Upper Confidence Bound (UCB) policy from reinforcement learning to help select the most appropriate algorithm at each iteration. An effective reward definition is proposed for the UCB policy. We consider three state-of-the-art individual-based SAEAs on different problems and compare them to the portfolios built from their instances on several benchmark problems given limited computation budgets. Our experimental studies demonstrate that our proposed portfolio frameworks significantly outperform any single algorithm on the set of benchmark problems.
引用
收藏
页码:943 / 950
页数:8
相关论文
共 28 条
  • [1] Alvaro Fialho., 2010, Proceedings of the 12th annual conference on Genetic and evolutionary computation, P767, DOI DOI 10.1145/1830483.1830619
  • [2] [Anonymous], 2002, J MACHINE LEARNING
  • [3] Baudis P., 2014, CORR
  • [4] Baudis P, 2014, LECT NOTES COMPUT SC, V8672, P40
  • [5] Brochu E., 2011, UAI, P327, DOI DOI 10.5555/3020548.3020587
  • [6] Cauwet M.-L., 2014, International Conference on Learning and Intelligent Optimization, P1
  • [7] Algorithm portfolios for noisy optimization
    Cauwet, Marie-Liesse
    Liu, Jialin
    Roziere, Baptiste
    Teytaud, Olivier
    [J]. ANNALS OF MATHEMATICS AND ARTIFICIAL INTELLIGENCE, 2016, 76 (1-2) : 143 - 172
  • [8] Elsayed SM, 2014, 2014 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC), P1650, DOI 10.1109/CEC.2014.6900308
  • [9] An economics approach to hard computational problems
    Huberman, BA
    Lukose, RM
    Hogg, T
    [J]. SCIENCE, 1997, 275 (5296) : 51 - 54
  • [10] Jin Y, 2005, SOFT COMPUT, V9, P3, DOI [10.1007/s00500-003-0328-5, 10.1007/S00500-003-0328-5]