The entropy of the class distribution of the prior (see prior_class_complexity), divided by the number of instances in the input data.