Level set and density estimation on manifolds
Given an iid sample of a distribution supported on a smooth manifold M^d , which is assumed to be absolutely continuous w.r.t the Hausdorff measure inherited from the ambient space, we tackle the problem of the estimation of the level sets of the density f . A consistent estimator in both Hausdorff distance and distance in measure is proposed. The estimator is the level set of the kernel-based estimator of the density f . We prove that the kernel-based density estimator converges uniformly to the unknown density f , the consistency of the level set and the consistency of the boundary of the level set estimator. The performance of our proposal is illustrated through some simulated examples.
READ FULL TEXT