“What-Where” sparse distributed invariant representations of visual patterns