General
Spectral Stability and Convergence of the Unnormalized Graph Laplacian on Noisy Point Clouds
arXiv stat.TH0 views
arXiv:2611.23573v1 Announce Type: new Abstract: The first eigenvalues and eigenvectors of the unnormalized graph Laplacian of clean samples from a manifold, suitably scaled, approximate those of a weighted Laplace-Beltrami operator. However, the samples are usually observed with noise and therefore need not lie on the manifold. We build the clean and noisy graph Laplacians, with bandwidth $\varepsilon$, on the clean samples and on the noisy observations, where the noise is arbitrary but bounded by $\delta$. We show that, with high probability, the first eigenvalues and eigenvectors of the noisy graph Laplacian are close to those of the clean one: the eigenvalues within $O(\delta/\varepsilon)$, and the subspaces spanned by the first eigenvectors within $O(\sqrt{\delta/\varepsilon})$, whether or not the eigenvalues are simple. Consequently, if $\delta$ is small enough relative to $\varepsilon$, the noisy graph Laplacian approximates the weighted Laplace-Beltrami operator at the same rate as the clean one. We also show that, for large noise, we can get the same rate as the clean graph Laplacian by first using a denoiser, and then building the graph Laplacian on the denoised observations.
Source:https://arxiv.org/abs/2611.23573PreviousFederated Differential Privacy with Batch ContaminationNextBerry Esseen bounds for fractional Vasicek type processes: continuous and high-frequency regimes
Related posts
If nobody writes it down, it did not get decided
0dev.to Productivity
Micrograph B-factor estimation via Wilson statistics
0arXiv eess.IV
Greedy Vector Approximation Based on Rotation for Multiplierless Complex Products
0arXiv eess.SP
Toward Massive MIMO HAPS Networks: Antenna Architectures and Design
0arXiv eess.SP
Deep-Learning-Assisted Full Waveform Inversion for Attenuation Reconstruction in Ultrasound Computed Tomography
0arXiv eess.SP
Two is better than one: optimizing binary rating scales for online platforms
0arXiv cs.MA