* clustering all the samples into max categories clusters that is some
* categories are merged together. The clustering is applied only in n>2-class
* classification problems for categorical variables with N > max_categories
* possible values. In case of regression and 2-class classification the optimal split can be found efficiently without employing clustering, thus the parameter is not used in these cases.
* possible values. In case of regression and 2-class classification the
* optimal split can be found efficiently without employing clustering, thus