Significance-Based Estimation-of-Distribution Algorithms

被引:32
作者
Doerr, Benjamin [1 ]
Krejca, Martin S. [2 ]
机构
[1] Inst Polytech Paris, Ecole Polytech, CNRS, Lab Informat, F-91120 Palaiseau, France
[2] Univ Potsdam, Hasso Plattner Inst, D-14482 Potsdam, Germany
关键词
Heuristic algorithms; Sociology; Statistics; History; Probabilistic logic; Benchmark testing; Genetic algorithms; Estimation-of-distribution algorithm (EDA); run time analysis; theory; RANDOMIZED SEARCH HEURISTICS; 1+LAMBDA EVOLUTIONARY ALGORITHM; RUNTIME ANALYSIS; LOWER BOUNDS; MUTATION; CHOICE;
D O I
10.1109/TEVC.2019.2956633
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Estimation-of-distribution algorithms (EDAs) are randomized search heuristics that create a probabilistic model of the solution space, which is updated iteratively, based on the quality of the solutions sampled according to the model. As previous works show, this iteration-based perspective can lead to erratic updates of the model, in particular, to bit-frequencies approaching a random boundary value. In order to overcome this problem, we propose a new EDA based on the classic compact genetic algorithm (cGA) that takes into account a longer history of samples and updates its model only with respect to information which it classifies as statistically significant. We prove that this significance-based cGA (sig-cGA) optimizes the commonly regarded benchmark functions OneMax (OM), LeadingOnes, and BinVal all in quasilinear time, a result shown for no other EDA or evolutionary algorithm so far. For the recently proposed stable compact genetic algorithm-an EDA that tries to prevent erratic model updates by imposing a bias to the uniformly distributed model-we prove that it optimizes OM only in a time exponential in its hypothetical population size. Similarly, we show that the convex search algorithm cannot optimize OM in polynomial time.
引用
收藏
页码:1025 / 1034
页数:10
相关论文
共 44 条
[1]  
Afshani P., 2019, DISCRETE APPL MATH, P28
[2]  
Anil G, 2009, FOGA'09: PROCEEDINGS OF THE 10TH ACM SIGRVO CONFERENCE ON FOUNDATIONS OF GENETIC ALGORITHMS, P67
[3]   A Tight Runtime Analysis for the (μ plus λ) EA [J].
Antipov, Denis ;
Doerr, Benjamin ;
Fang, Jiefeng ;
Hetet, Tangi .
GECCO'18: PROCEEDINGS OF THE 2018 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE, 2018, :1459-1466
[4]   Precise Runtime Analysis for Plateaus [J].
Antipov, Denis ;
Doerr, Benjamin .
PARALLEL PROBLEM SOLVING FROM NATURE - PPSN XV, PT II, 2018, 11102 :117-128
[5]  
Badkobeh G, 2014, LECT NOTES COMPUT SC, V8672, P892
[6]  
Doerr B., 2020, THEORY EVOLUTIONARY
[7]  
Doerr B., 2010, GENETIC EVOLUTIONARY, P759
[8]   Analyzing randomized search heuristics via stochastic domination [J].
Doerr, Benjamin .
THEORETICAL COMPUTER SCIENCE, 2019, 773 :115-137
[9]   Runtime Analysis for Self-adaptive Mutation Rates [J].
Doerr, Benjamin ;
Witt, Carsten ;
Yang, Jing .
GECCO'18: PROCEEDINGS OF THE 2018 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE, 2018, :1475-1482
[10]   On the Runtime Analysis of Selection Hyper-Heuristics with Adaptive Learning Periods [J].
Doerr, Benjamin ;
Lissovoi, Andrei ;
Oliveto, Pietro S. ;
Warwicker, John Alasdair .
GECCO'18: PROCEEDINGS OF THE 2018 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE, 2018, :1015-1022