Might be worthwhile to add conditional risk to it, to generalize it to the minimum risk classifier. That way you'd also distinguish it from, say the scikit-learn implementation.