Escaping Local Optima Using Crossover With Emergent Diversity

被引:108
作者
Dang, Duc-Cuong [1 ]
Friedrich, Tobias [2 ]
Koetzing, Timo [2 ]
Krejca, Martin S. [2 ]
Lehre, Per Kristian [3 ]
Oliveto, Pietro S. [4 ]
Sudholt, Dirk [4 ]
Sutton, Andrew M. [2 ]
机构
[1] Univ Nottingham, Sch Comp Sci, ASAP Res Grp, Nottingham NG8 1BB, England
[2] Hasso Plattner Inst, Chair Algorithm Engn, D-14482 Potsdam, Germany
[3] Univ Birmingham, Sch Comp Sci, Birmingham B15 2TT, W Midlands, England
[4] Univ Sheffield, Dept Comp Sci, Sheffield S1 4DP, S Yorkshire, England
基金
英国工程与自然科学研究理事会;
关键词
Diversity; genetic algorithms (GAs); recombination; runtime analysis; theory; SIMPLE GENETIC ALGORITHM; EVOLUTIONARY ALGORITHMS; RUNTIME ANALYSIS; OPTIMIZATION; MECHANISMS; COMPLEXITY;
D O I
10.1109/TEVC.2017.2724201
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Population diversity is essential for avoiding premature convergence in genetic algorithms (GAs) and for the effective use of crossover. Yet the dynamics of how diversity emerges in populations are not well understood. We use rigorous runtime analysis to gain insight into population dynamics and GA performance for the (mu + 1) GA and the Jump test function. We show that the interplay of crossover followed by mutation may serve as a catalyst leading to a sudden burst of diversity. This leads to significant improvements of the expected optimization time compared to mutation-only algorithms like the (1 + 1) evolutionary algorithm. Moreover, increasing the mutation rate by an arbitrarily small constant factor can facilitate the generation of diversity, leading to even larger speedups. Experiments were conducted to complement our theoretical findings and further highlight the benefits of crossover on the function class.
引用
收藏
页码:484 / 497
页数:14
相关论文
共 35 条
  • [21] Lehre P. K., 2015, P FDN GEN ALG FOG 15, P3
  • [22] Crossover can be constructive when computing unique input-output sequences
    Lehre, Per Kristian
    Yao, Xin
    [J]. SOFT COMPUTING, 2011, 15 (09) : 1675 - 1687
  • [23] Genetic algorithms: Concepts and applications
    Man, KF
    Tang, KS
    Kwong, S
    [J]. IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, 1996, 43 (05) : 519 - 534
  • [24] Neumann F, 2011, GECCO-2011: PROCEEDINGS OF THE 13TH ANNUAL GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE, P1587
  • [25] Neumann F, 2010, NAT COMPUT SER, P3, DOI 10.1007/978-3-642-16544-3
  • [26] Improved time complexity analysis of the Simple Genetic Algorithm
    Oliveto, Pietro S.
    Witt, Carsten
    [J]. THEORETICAL COMPUTER SCIENCE, 2015, 605 : 21 - 41
  • [27] Analysis of diversity mechanisms for optimisation in dynamic environments with low frequencies of change
    Oliveto, Pietro S.
    Zarges, Christine
    [J]. THEORETICAL COMPUTER SCIENCE, 2015, 561 : 37 - 56
  • [28] On the runtime analysis of the Simple Genetic Algorithm
    Oliveto, Pietro S.
    Witt, Carsten
    [J]. THEORETICAL COMPUTER SCIENCE, 2014, 545 : 2 - 19
  • [29] Analysis of Population-based Evolutionary Algorithms for the Vertex Cover Problem
    Oliveto, Pietro S.
    He, Jun
    Yao, Xin
    [J]. 2008 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION, VOLS 1-8, 2008, : 1563 - 1570
  • [30] Benefits of a Population: Five Mechanisms That Advantage Population-Based Algorithms
    Pruegel-Bennett, Adam
    [J]. IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION, 2010, 14 (04) : 500 - 517