Learning invariant object recognition in the visual system with continuous transformations