CMA-ES

from Wikipedia, the free encyclopedia

CMA-ES stands for Covariance Matrix Adaptation Evolution Strategy . The CMA ( Covariance Matrix Adaptation ) is a derandomized procedure for adapting the covariance matrix of the Gaussian mutation distribution (normal distribution) in the evolution strategy (ES). The covariance matrix of the multidimensional (multivariate) normal distribution describes the pair-wise dependencies between the variables. The adaptation of the covariance matrix in the evolution strategy is comparable to the approximation of the inverse Hessian matrix in classical optimization , e.g. B. in the quasi-Newton method .

Basic principle

The adaptation principle is based on the idea of ​​increasing the probability of previously successful steps. For this purpose, the covariance matrix of the distribution is changed in such a way that the probability of the selected step of the last generation increases.

Web links