Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper analyzes denoising score matching (DSM) when the latent data lies on a manifold, revealing a singular noise channel in the tangent denoising target under Gaussian corruption. By conditioning on the nearest-point projection, the authors derive a Rao-Blackwellized predictor that removes this singularity and provides the L^2-optimal estimate. The small-noise expansion of this target decomposes into the intrinsic Riemannian score plus an explicit order-$σ^2$ correction involving intrinsic Tweedie and extrinsic curvature terms, offering insights into manifold-aware score estimation.
Denoising score matching on manifolds suffers from a singularity that can be elegantly resolved by Rao-Blackwellization, leading to a novel decomposition of the score into intrinsic Riemannian and extrinsic curvature components.
We study denoising score matching (DSM) when the latent distribution is supported on a smooth embedded manifold $M \subset \mathbb{R}^D$. Under ambient Gaussian corruption, the tangent denoising target contains a singular normal-fiber noise channel whose variance diverges as $d/σ^2$ as $σ\to 0^+$. We show that conditioning on the nearest-point projection $π(X)$ canonically removes this singularity: the resulting conditional expectation is the unique $L^2$-optimal Rao-Blackwellized predictor of the tangent DSM target among all estimators depending only on the projected observation $π(X)$. We then compute the small-noise expansion of this canonical target and show that it equals the intrinsic Riemannian score up to an explicit order-$σ^2$ correction that decomposes into an intrinsic Tweedie term and an extrinsic curvature term involving the Weingarten and Ricci operators. In the flat case, the construction reduces exactly to ordinary lower-dimensional Gaussian DSM, while on $S^d$ the extrinsic correction simplifies to the scalar factor $(1-d/2)\nabla_M \log q$; this extrinsic $σ^2$ correction cancels identically on $S^2$, though the intrinsic Tweedie term remains.