5 years ago

# Confidence interval of singular vectors for high-dimensional and low-rank matrix regression.

Dong Xia

Let ${\bf M}\in\mathbb{R}^{m_1\times m_2}$ be an unknown matrix with $r={\rm rank}({\bf M})\ll \min(m_1,m_2)$ whose thin singular value decomposition is denoted by ${\bf M}={\bf U}{\bf \Lambda}{\bf V}^{\top}$ where ${\bf \Lambda}={\rm diag}(\lambda_1,\cdots,\lambda_r)$ contains its non-increasing singular values. Low rank matrix regression refers to instances of estimating ${\bf M}$ from $n$ i.i.d. copies of random pair $\{({\bf X}, y)\}$ where ${\bf X}\in\mathbb{R}^{m_1\times m_2}$ is a random measurement matrix and $y\in\mathbb{R}$ is a noisy output satisfying $y={\rm tr}({\bf M}^{\top}{\bf X})+\xi$ with $\xi$ being stochastic error independent of ${\bf X}$. The goal of this paper is to construct efficient estimator (denoted by $\hat{\bf U}$ and $\hat{\bf V}$) and confidence interval of ${\bf U}$ and ${\bf V}$. In particular, we characterize the distribution of ${\rm dist}^2\big[(\hat{\bf U},\hat{\bf V}), ({\bf U},{\bf V})\big]=\|\hat{\bf U}\hat{\bf U}^{\top}-{\bf U}{\bf U}^{\top}\|_{\rm F}^2+\|\hat{\bf V}\hat{\bf V}^{\top}-{\bf V}{\bf V}^{\top}\|_{\rm F}^2.$ We prove the asymptotical normality of properly centered and normalized ${\rm dist}^2\big[(\hat{\bf U},\hat{\bf V}), ({\bf U},{\bf V})\big]$ with data-dependent centering and normalization when $r^{5/2}(m_1+m_2)^{3/2}=o(n/\log n)$, based on which confidence interval of ${\bf U}$ and ${\bf V}$ is constructed achieving any pre-determined confidence level asymptotically.

Publisher URL: http://arxiv.org/abs/1805.09871

DOI: arXiv:1805.09871v1

You might also like
Discover & Discuss Important Research

Keeping up-to-date with research can feel impossible, with papers being published faster than you'll ever be able to read them. That's where Researcher comes in: we're simplifying discovery and making important discussions happen. With over 19,000 sources, including peer-reviewed journals, preprints, blogs, universities, podcasts and Live events across 10 research areas, you'll never miss what's important to you. It's like social media, but better. Oh, and we should mention - it's free.

Researcher displays publicly available abstracts and doesnâ€™t host any full article content. If the content is open access, we will direct clicks from the abstracts to the publisher website and display the PDF copy on our platform. Clicks to view the full text will be directed to the publisher website, where only users with subscriptions or access through their institution are able to view the full article.