In this work we take a novel view of nonlinear manifold learning. Usually, manifold learning is formulated in terms of finding an embedding or "unrolling" of a manifold into a lower dimensional space.
