. . "of a problem such as hunting of the sub-feedback learning value." . .