CMA-ES
Evolutionary.CMAES
— TypeCovariance Matrix Adaptation Evolution Strategy Implementation: (μ/μ_{I,W},λ)-CMA-ES
The constructor takes following keyword arguments:
μ
/mu
is the number of parentsλ
/lambda
is the number of offspringc_1
is a learning rate for the rank-one update of the covariance matrix updatec_c
is a learning rate for cumulation for the rank-one update of the covariance matrixc_mu
is a learning rate for the rank-$\mu$ update of the covariance matrix updatec_sigma
is a learning rate for the cumulation for the step-size controlc_m
is the learning rate for the mean update, $c_m \leq 1$σ0
/sigma0
is the initial step sizeσ
weights
are recombination weights, if the weights are set to $1/\mu$ then the intermediate recombination is activated.metrics
is a collection of convergence metrics.
Description
The Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is a stochastic derivative-free numerical optimization algorithm for difficult (non-convex, ill-conditioned, multi-modal, rugged, noisy) optimization problems in continuous search spaces [1].
The current CMA-ES algorithm implementation based on a simplified outline[2].
References
- 1Hansen, N. (2016), "The CMA Evolution Strategy: A Tutorial", arXiv:1604.00772
- 2http://www.scholarpedia.org/article/Evolution_strategies