How to obtain the Hessian matrix after optimization?

[Landau1908]

Hi,

Is there a way to propagate uncertainties from parameters(ai) to observables(xj), like as the Hessian matrix does?
In the Hessian approach, the uncertainty on the observable is determined in terms of the uncertainties in parameter space.
The uncertainty on the observable gives the confidence level (C.L.).
In CMA, how to get the Hessian matrix or C.L.?

Regards

[nikohansen]

The sample covariance matrix is an estimator of the inverse Hessian up to a scalar factor. If isinstance(es, cma.CMAEvolutionStrategy) and es.sigma_vec.is_identity and es.gp.isidentity then es.C is the sample covariance matrix up to a scalar factor.