Log-linear Convergence of the Scale-invariant $(\mu/\mu_w,\lambda)$-{ES} and Optimal $\mu$ for Intermediate Recombination for Large Population Sizes - Archive ouverte HAL Access content directly
Conference Papers Year : 2010

Log-linear Convergence of the Scale-invariant $(\mu/\mu_w,\lambda)$-{ES} and Optimal $\mu$ for Intermediate Recombination for Large Population Sizes

(1) , (1)
1
Mohamed Jebalia
• Function : Author
• PersonId : 842777
Anne Auger
• Function : Author
• PersonId : 751513
• IdHAL : anne-auger

Abstract

Evolution Strategies (ESs) are population-based methods well suited for parallelization. In this paper, we study the convergence of the (mu/mu_w,lambda)-ES, an ES with weighted recombination, and derive its optimal convergence rate and optimal mu especially for large population sizes. First, we theoretically prove the log-linear convergence of the algorithm using a scale-invariant adaptation rule for the step-size and minimizing spherical objective functions and identify its convergence rate as the expectation of an underlying random variable. Then, using Monte-Carlo computations of the convergence rate in the case of equal weights, we derive optimal values for mu that we compare with previously proposed rules. Our numerical computations show also a dependency of the optimal convergence rate in ln(lambda) in agreement with previous theoretical results.

Dates and versions

inria-00494478 , version 1 (24-06-2010)

Identifiers

• HAL Id : inria-00494478 , version 1

Cite

Mohamed Jebalia, Anne Auger. Log-linear Convergence of the Scale-invariant $(\mu/\mu_w,\lambda)$-{ES} and Optimal $\mu$ for Intermediate Recombination for Large Population Sizes. Parallel Problem Solving From Nature (PPSN2010), Sep 2010, Krakow, Poland. pp.xxxx-xxx. ⟨inria-00494478⟩

261 View