FHDMs achieve minimax optimal TV convergence rates for spherically supported Sobolev data distributions up to log factors, the first optimality result for random-time denoising diffusion models.
Journal of the American Mathematical Society , month =
9 Pith papers cite this work, alongside 13 external citations. Polarity classification is still indexing.
9
Pith papers citing it
13
external citations · OpenAlex
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Introduces tangential Bayes denoiser for Riemannian Gaussian mixtures on manifolds via spectral Laplace-Beltrami approximation, with nearly Bayes risk in low noise and minimax optimality on the circle.
An exact algebraic identity plus low-rank SVD and Haar-measure null-space approximation reduce per-point mean curvature cost from O(m^4) to O(k^2 m + k m p^2) with 50-300x speedups and negligible accuracy loss.
WaveDiff with wavefront feature projection recovers WFE from noisy undersampled in-focus observations at ~3% error, a tenfold improvement over the prior version.
Fixed orthonormal bottlenecks of modest dimension preserve RL expressivity and performance under linear realizability while stabilizing feature norms and effective rank.
Nonlinear dimensionality reduction on ECG signals enables unsupervised personalized arrhythmia detection with high accuracy on 2D embeddings using standard algorithms on the MIT-BIH database.
Derives an asymptotic equivalent for the Representation Gap in equivariant diffusion models, showing it depends primarily on the intrinsic dimension of the task.
Asynchronous sequential updates in KLR Hopfield networks produce statistically indistinguishable trajectories from synchronous dynamics, achieve empirical capacities near P/N=30, and converge with event counts close to initial Hamming distance.
Projects GNN embeddings of IoT traffic onto interpretable manifolds to achieve 0.83 F1-score intrusion detection and reveal concept drift.
citing papers explorer
-
Statistical Convergence of Spherical First Hitting Diffusion Models
FHDMs achieve minimax optimal TV convergence rates for spherically supported Sobolev data distributions up to log factors, the first optimality result for random-time denoising diffusion models.
-
Nonparametric Riemannian Empirical Bayes, and Denoising Measurements on Manifolds
Introduces tangential Bayes denoiser for Riemannian Gaussian mixtures on manifolds via spectral Laplace-Beltrami approximation, with nearly Bayes risk in low noise and minimax optimality on the circle.
-
Efficient Mean Curvature Computation on High-Dimensional Data Manifolds
An exact algebraic identity plus low-rank SVD and Haar-measure null-space approximation reduce per-point mean curvature cost from O(m^4) to O(k^2 m + k m p^2) with 50-300x speedups and negligible accuracy loss.
-
Point spread function wavefront recovery from in-focus stellar observations
WaveDiff with wavefront feature projection recovers WFE from noisy undersampled in-focus observations at ~3% error, a tenfold improvement over the prior version.
-
Learning in Low-Dimensional Subspaces: Orthogonal Bottlenecks for Reinforcement Learning
Fixed orthonormal bottlenecks of modest dimension preserve RL expressivity and performance under linear realizability while stabilizing feature norms and effective rank.
-
Manifold Learning for Personalized and Label-Free Detection of Cardiac Arrhythmias
Nonlinear dimensionality reduction on ECG signals enables unsupervised personalized arrhythmia detection with high accuracy on 2D embeddings using standard algorithms on the MIT-BIH database.
-
Representation Gap: Explaining the Unreasonable Effectiveness of Neural Networks from a Geometric Perspective
Derives an asymptotic equivalent for the Representation Gap in equivariant diffusion models, showing it depends primarily on the intrinsic dimension of the task.
-
Efficient event-driven retrieval in high-capacity kernel Hopfield networks
Asynchronous sequential updates in KLR Hopfield networks produce statistically indistinguishable trajectories from synchronous dynamics, achieve empirical capacities near P/N=30, and converge with event counts close to initial Hamming distance.
-
Interpreting Manifolds and Graph Neural Embeddings from Internet of Things Traffic Flows
Projects GNN embeddings of IoT traffic onto interpretable manifolds to achieve 0.83 F1-score intrusion detection and reveal concept drift.