Testing the Manifold Hypothesis

The hypothesis that high dimensional data tend to lie in the vicinity of a low dimensional manifold is the basis of manifold learning. The goal of this paper is to develop an algorithm (with accompanying complexity guarantees) for fitting a manifold to an unknown probability distribution supported in a separable Hilbert space, only using i.i.d samples from that distribution.

More precisely, our setting is the following. Suppose that data are drawn independently at random from a probability distribution $P$ supported on the unit ball of a separable Hilbert space $H$.

Let $G(d, V, τ)$ be the set of submanifolds of the unit ball of $H$ whose volume is at most $V$ and reach (which is the supremum of all $r$ such that any point at a distance less than $r$ has a unique nearest point on the manifold) is at least $τ$. Let $L(M, P)$ denote mean-squared distance of a random point from the probability distribution $P$ to $M$.

We obtain an algorithm that tests the manifold hypothesis in the following sense. The algorithm takes i.i.d random samples from $P$ as input, and determines which of the following two is true (at least one must be): (a) There exists $M \in G(d, CV, \fracτ{C})$ such that $L(M, P) \leq C ε.$ (b) There exists no $M \in G(d, V/C, Cτ)$ such that $L(M, P) \leq \fracε{C}.$ The answer is correct with probability at least $1-δ$.