Cma-es algorithm

Cma-es algorithm

first, the algorithm introduction

Cma-es is the covariance matrix adaptation evolutionary strategies abbreviation, the Chinese name is the covariance matrix adaptive Evolutionary strategy, mainly used to solve the continuous optimization problem, especially in the condition of the continuous optimization problem. Evolutionary strategy algorithm is mainly used as a method to solve parameter optimization problems, imitating the principle of biological evolution, assuming that regardless of the genetic changes, the resulting results (traits) always follow the 0 mean, one side of the poor Gaussian distribution. Note that evolutionary strategies differ from genetic algorithms, but are important variants of the evolutionary algorithm (EAs).

Second, the implementation of the algorithm

Three, the main features1. Using a multivariate normal distribution to generate new search points

-follow the maximum entropy principle
X? i ~ m? +σ n ( 0,c) For i = 1, 2, ...,λ

2. Sorting-based selection process

- implicit invariance, with the same performance for G (f (x)), G is the increment function

3. Step control makes fast convergence more convenient

- based on evolutionary path

4. Covariance matrix Adaptive algorithm increases the likelihood of success step, and can improve performance based on the order of magnitude of the problem.

-C∝H?1?AdjustableWholeChangeAmountOfClassType，BySeaYeung sumMomentArraysuch asPriceYesMoreForJane single Xie party difference moment matrix
-?ProductionHealthOneTwoNew Q topic Fang method
-?F(X)=G(XTHX) ask Jane g ( xtx ) /span>

Cma-es algorithm