Module cma_es 

Covariance Matrix Adaptation Evolution Strategy (CMA-ES)

CMA-ES is a stochastic, derivative-free method for numerical optimization of non-linear or non-convex continuous optimization problems. It belongs to the class of evolutionary algorithms and evolution strategies.

Key features

• Invariant under order-preserving transformations of the fitness function
• Robust to scaling, rotation, and translation of the search space
• Adaptive step-size control (Cumulative Step-size Adaptation / CSA)
• Covariance matrix adaptation via rank-1 and rank-mu updates
• IPOP restart strategy for escaping local optima
• Multiple boundary handling strategies

References

• Hansen, N. (2016). The CMA Evolution Strategy: A Tutorial. arXiv:1604.00772
• Auger, A. & Hansen, N. (2005). A Restart CMA Evolution Strategy with Increasing Population Size. CEC 2005.

Structs

CmaEsOptions

Options for CMA-ES optimization

CmaEsResult

Result of CMA-ES optimization

CmaEsState

Internal state of the CMA-ES algorithm

IpopCmaEs

IPOP-CMA-ES: CMA-ES with Increasing Population Restarts

Enums

BoundaryHandling

Boundary handling strategy for CMA-ES

RestartStrategy

Restart strategy for CMA-ES

Functions

cma_es_minimize

Convenience function to minimize using CMA-ES