Convergence results for the (1,λ)-SA-ES using the theory of φ-irreducible Markov chains

被引:60
作者
Auger, A [1 ]
机构
[1] Univ Paris 11, Equipe TAO, INRIA Futurs, LRI, F-91405 Orsay, France
关键词
evolution strategies; convergence; Markov chains; Foster-Lyapunov drift conditions;
D O I
10.1016/j.tcs.2004.11.017
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
This paper investigates theoretically the (1, lambda)-SA-ES on the well known sphere function. We prove sufficient conditions on the parameters of the algorithm ensuring the convergence of 1/n ln(parallel to Xn parallel to), where X-n is the parent at generation n. This in turn guarantees the asymptotic log-linear convergence or divergence of the algorithm. The technique used for this analysis calls upon the theory of Markov chains on a continuous state space and on the so-called Foster-Lyapunov drift conditions. Those conditions enable us to derive practical conditions that prove stability properties of Markov chains. (c) 2005 Elsevier B.V. All rights reserved.
引用
收藏
页码:35 / 69
页数:35
相关论文
共 22 条
[1]   Convergence of simulated annealing using Foster-Lyapunov criteria [J].
Andrieu, C ;
Breyer, LA ;
Doucet, A .
JOURNAL OF APPLIED PROBABILITY, 2001, 38 (04) :975-994
[2]  
[Anonymous], 1973, EVOLUTIONSTRATEGIE O
[3]  
BENVENISTE A, 1987, ALGORITHMES ADAPTATI
[4]  
Beyer H.-G., 2001, NAT COMP SER
[5]   Toward a Theory of Evolution Strategies: Self-Adaptation [J].
Beyer, Hans-Georg .
EVOLUTIONARY COMPUTATION, 1995, 3 (03) :311-347
[6]   Global convergence for evolution strategies in spherical problems:: some simple proofs and difficulties [J].
Bienvenüe, A ;
François, O .
THEORETICAL COMPUTER SCIENCE, 2003, 306 (1-3) :269-289
[7]  
Billingsley P., 1986, PROBABILITY MEASURE
[8]  
DELAURENTIS JM, 2002, P GEN EV C, P229
[9]  
DUFLO M, 1996, ALGORITHMES STOCHAST
[10]  
Geyer CJ, 1992, STAT SCI, V7, P473, DOI [DOI 10.1214/SS/1177011137, 10.1214/ss/1177011137]