Dimension-Independent Convergence Rate for Non-isotropic (1, λ) — ES
作者： Anne Auger , Claude Le Bris , Marc Schoenauer
关键词:
摘要: Based on the theory of non-negative super martingales, convergence results are proven for adaptive (1, λ) - ES (i.e. with Gaussian mutations), and geometrical rates derived. In d-dimensional case (d > 1), algorithm studied here uses a different step-size update in each direction. However, critical value step-size, resulting rate do not depend dimension. Those discussed respect to previous works. Rigorous numerical investigations some 1-dimensional functions validate theoretical results. Trends future research indicated.
springer.com
core.ac.uk
sci-hub.st 参考文章(10)
John DeLaurentis, William E. Hart, Lauren Ferguson, On The Convergence Properties Of A Simple Self-adaptive Evolutionary Algorithm genetic and evolutionary computation conference. pp. 229- 237 ,(2002)
Hans Georg Beyer, The Theory of Evolution Strategies ,(2001)
Mikhail A. Semenov, Convergence velocity of an evolutionary algorithm with self-adaptation genetic and evolutionary computation conference. pp. 210- 213 ,(2002)
David Williams, Probability with martingales ,(1991)
G. Yin, G. Rudolph, H.-P, Schwefel, Analyzing the (1, λ) evolution strategy via stochastic approximation methods Evolutionary Computation. ,vol. 3, pp. 473- 489 ,(1995) , 10.1162/EVCO.1995.3.4.473
Alexis Bienvenüe, Olivier François, Global convergence for evolution strategies in spherical problems: some simple proofs and difficulties Theoretical Computer Science. ,vol. 306, pp. 269- 289 ,(2003) , 10.1016/S0304-3975(03)00284-6
A.E. Eiben, R. Hinterding, Z. Michalewicz, Parameter control in evolutionary algorithms IEEE Transactions on Evolutionary Computation. ,vol. 3, pp. 124- 141 ,(1999) , 10.1109/4235.771166
G. Rudolph, Convergence of non-elitist strategies world congress on computational intelligence. pp. 63- 66 ,(1994) , 10.1109/ICEC.1994.350041
Hans-Paul Schwefel, Numerical optimization of computer models ,(1981)
Hans-Georg Beyer, The Theory of Evolution Strategies Natural Computing Series. ,(2001) , 10.1007/978-3-662-04378-3