The entropy of the class distribution generated by the model (see class_complexity), divided by the number of instances in the input data.