Fixed the eigendecomposition error generated in the search covariance matrix. In high dimension problems, it was failing due to positive definite covariance matrix. Switched to scipy.eigenh eigen hermittian decomposition
Rollbacked the number of initial function evaluations to D for deterministic cases.
changed the interface of the init Sobol function, not it returns also the number of samples size (could be helpful).
scipy.eigenheigen hermittian decomposition