The method _mean_of_feature_by_class provides a means for getting information on training set feature distributions by known class, but isn't currently used to provide a visualization for (1) how much separation there is in the training set on this feature and (2) whether the current example looks more like the distribution for its predicted class.