The introduction of almost every lecture for Decision Trees is a decision tree, where one node might have more than two children. However, as I understand the decision tree is always a binary tree, right? During the implementation I one-hot encode all my categorical variables, meaning the split on such a feature will always be 0 or 1 (two children).
So, if I have a feature with normal, high values after a one-hot encoding I will get two features humidity.normal and humidity.high. And the split will be first performed on either humidity.normal and then humidity.high or vice versa.
Are the RandomForestClassifier/DescitionTree in python and r being constructed via binary trees (binary splits)?