Optimal Bias-variance Tradeoff in Matrix and Tensor Estimation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kumar, Shivam, Luo, Xiaokai, Xu, Haotian, Padilla, Carlos Misael Madrid, Padilla, Oscar Hernan Madrid, Wang, Daren
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908816561405952
author Kumar, Shivam
Luo, Xiaokai
Xu, Haotian
Padilla, Carlos Misael Madrid
Padilla, Oscar Hernan Madrid
Wang, Daren
author_facet Kumar, Shivam
Luo, Xiaokai
Xu, Haotian
Padilla, Carlos Misael Madrid
Padilla, Oscar Hernan Madrid
Wang, Daren
contents We study matrix and tensor denoising when the underlying signal is \textbf{not} necessarily low-rank. In the tensor setting, we observe \[ Y = X^\ast + Z \in \mathbb{R}^{p_1 \times p_2 \times p_3}, \] where $X^\ast$ is an unknown signal tensor and $Z$ is a noise tensor. We propose a one-step variant of the higher-order SVD (HOSVD) estimator, denoted $\widetilde X$, and show that, uniformly over any user-specified Tucker ranks $(r_1,r_2,r_3)$, with high probability, \[ \|\widetilde X - X^\ast\|_{\mathrm F}^2 = O\Big( κ^2\Big\{r_1r_2r_3 + \sum_{k=1}^3 p_k r_k\Big\} + ξ_{(r_1,r_2,r_3)}^2 \Big). \] Here, $ξ_{(r_1,r_2,r_3)}$ is the best achievable Tucker rank-$(r_1,r_2,r_3)$ approximation error of $X^\ast$ (bias), $κ^2$ quantifies the noise level, and $κ^2\{r_1r_2r_3+\sum_{k=1}^3 p_k r_k\}$ is the variance term scaling with the effective degrees of freedom of $\widetilde X$. This yields a rank-adaptive bias-variance tradeoff: increasing $(r_1,r_2,r_3)$ decreases the bias $ξ_{(r_1,r_2,r_3)}$ while increasing variance. In the matrix setting, we show that truncated SVD achieves an analogous bias-variance tradeoff for arbitrary signal matrices. Notably, our matrix result requires \textbf{no} assumptions on the signal matrix, such as finite rank or spectral gaps. Finally, we complement our upper bounds with matching information-theoretic lower bounds, showing that the resulting bias-variance tradeoff is minimax optimal up to universal constants in both the matrix and tensor settings.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17382
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimal Bias-variance Tradeoff in Matrix and Tensor Estimation
Kumar, Shivam
Luo, Xiaokai
Xu, Haotian
Padilla, Carlos Misael Madrid
Padilla, Oscar Hernan Madrid
Wang, Daren
Machine Learning
Statistics Theory
Methodology
We study matrix and tensor denoising when the underlying signal is \textbf{not} necessarily low-rank. In the tensor setting, we observe \[ Y = X^\ast + Z \in \mathbb{R}^{p_1 \times p_2 \times p_3}, \] where $X^\ast$ is an unknown signal tensor and $Z$ is a noise tensor. We propose a one-step variant of the higher-order SVD (HOSVD) estimator, denoted $\widetilde X$, and show that, uniformly over any user-specified Tucker ranks $(r_1,r_2,r_3)$, with high probability, \[ \|\widetilde X - X^\ast\|_{\mathrm F}^2 = O\Big( κ^2\Big\{r_1r_2r_3 + \sum_{k=1}^3 p_k r_k\Big\} + ξ_{(r_1,r_2,r_3)}^2 \Big). \] Here, $ξ_{(r_1,r_2,r_3)}$ is the best achievable Tucker rank-$(r_1,r_2,r_3)$ approximation error of $X^\ast$ (bias), $κ^2$ quantifies the noise level, and $κ^2\{r_1r_2r_3+\sum_{k=1}^3 p_k r_k\}$ is the variance term scaling with the effective degrees of freedom of $\widetilde X$. This yields a rank-adaptive bias-variance tradeoff: increasing $(r_1,r_2,r_3)$ decreases the bias $ξ_{(r_1,r_2,r_3)}$ while increasing variance. In the matrix setting, we show that truncated SVD achieves an analogous bias-variance tradeoff for arbitrary signal matrices. Notably, our matrix result requires \textbf{no} assumptions on the signal matrix, such as finite rank or spectral gaps. Finally, we complement our upper bounds with matching information-theoretic lower bounds, showing that the resulting bias-variance tradeoff is minimax optimal up to universal constants in both the matrix and tensor settings.
title Optimal Bias-variance Tradeoff in Matrix and Tensor Estimation
topic Machine Learning
Statistics Theory
Methodology
url https://arxiv.org/abs/2509.17382