Tensor denoising with trend filtering

DOI10.4171/MSL/26zbMATH Open1493.62444arXiv2101.10692MaRDI QIDQ2113264FDOQ2113264

Francesco Ortelli, Sara Van De Geer

Publication date: 11 March 2022

Published in: Mathematical Statistics and Learning (Search for Journal in Brave)

Abstract: We extend the notion of trend filtering to tensors by considering the $k^{mth}$-order Vitali variation, a discretized version of the integral of the absolute value of the $k^{mth}$-order total derivative. We prove adaptive $el{l}^0$-rates and not-so-slow $el{l}^1$-rates for tensor denoising with trend filtering. For $k=1,2,3,4$ we prove that the $d$-dimensional margin of a $d$-dimensional tensor can be estimated at the $el{l}^0$-rate $n^{-1}$, up to logarithmic terms, if the underlying tensor is a product of $(k-1{)}^{mth}$-order polynomials on a constant number of hyperrectangles. For general $k$ we prove the $el{l}^1$-rate of estimation $n^{-fracH(d)+2k-12H(d)+2k-1}$, up to logarithmic terms, where $H(d)$ is the $d^{mth}$ harmonic number. Thanks to an ANOVA-type of decomposition we can apply these results to the lower dimensional margins of the tensor to prove bounds for denoising the whole tensor. Our tools are interpolating tensors to bound the effective sparsity for $el{l}^0$-rates, mesh grids for $el{l}^1$-rates and, in the background, the projection arguments by Dalalyan et al.

Full work available at URL: https://arxiv.org/abs/2101.10692

Recommendations

• Adaptive risk bounds in univariate total variation denoising and trend filtering
• Adaptive piecewise polynomial estimation via trend filtering
• Additive models with trend filtering
• Prediction bounds for higher order total variation regularized least squares
• Lower-Rank Tensor Approximation and Multiway Filtering

zbMATH Keywords

total variationtrend filteringoracle inequalitiestensor denoisingVitali variation

Mathematics Subject Classification ID

Ridge regression; shrinkage estimators (Lasso) (62J07)

Cites Work

• Pathwise coordinate optimization
• Title not available (Why is that?)
• Multivariate adaptive regression splines
• Adaptive piecewise polynomial estimation via trend filtering
• Title not available (Why is that?)
• Sparsity and Smoothness Via the Fused Lasso
• Locally adaptive regression splines
• Adaptive estimation of multivariate piecewise polynomials and bounded variation functions by optimal decision trees
• Adaptive risk bounds in univariate total variation denoising and trend filtering
• Estimation and testing under sparsity. École d'Été de Probabilités de Saint-Flour XLV -- 2015
• Multivariate extensions of isotonic regression and total variation denoising via entire monotonicity and Hardy-Krause variation
• Additive models with trend filtering
• $\ell_1$ Trend Filtering
• On the prediction performance of the Lasso
• Analysis versus synthesis in signal priors
• On tight bounds for the Lasso
• On the total variation regularized estimator over a class of tree graphs
• Title not available (Why is that?)
• Prediction bounds for higher order total variation regularized least squares
• New Risk Bounds for 2D Total Variation Denoising
• Oracle inequalities for square root analysis estimators with application to total variation penalties

Cited In (4)

• MARS via lasso
• Multivariate trend filtering for lattice data
• Variational, Geometric, and Level Set Methods in Computer Vision
• The Lasso with structured design and entropy of (absolute) convex hulls

This page was built for publication: Tensor denoising with trend filtering

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2113264)