Learning illumination- and orientation-invariant representations of objects through temporal association.