Explorative learning of right inverse functions: theoretical implications of redundancy