Hyperparameter optimization with approximate gradient

• scikit-learn 0.16

Standardize features of the input data such that each feature has unit variance. This makes the Hessian better conditioned. This can be done using e.g. scikit-learn's StandardScaler.