difference between best_estimator_ and best_params

[fcoppey]

Hello, when trying to optimise a LocalOutlierFactor() function I get differences between

gs.best_estimator_:
Best Estimator LocalOutlierFactor(algorithm='auto', contamination=0.01, leaf_size=30,
metric='minkowski', metric_params=None, n_jobs=None,
n_neighbors=3, novelty=True, p=2)

and gs.best_params:
{'contamination': 0.2, 'n_neighbors': 7, 'novelty': True}

any idea why?

Thanks for the great module...