Extend Grid Search to Stats/Norm

[zhangpengshan]

Like number of binning: 20 or 30, missing value processing type, these should impact model performance in triaining,

how to add such parameters to grid search?

[zhangpengshan]

Like grid search in h2o:

hyper_parameters = {'ntrees':[50],

                'max_depth':list(range(5,10,2)),

                'sample_rate':[x * 0.1 for x in range(8, 11, 2)],

                'nbins':list(range(20,50,10)),

                'min_rows':list(range(10,40,10)),

                'histogram_type': ["UniformAdaptive","QuantilesGlobal","RoundRobin"]

               }