The Optimal Hard Threshold for Singular Values is 4 / √ 3

نویسندگان

  • Matan Gavish
  • David L. Donoho
چکیده

We consider recovery of low-rank matrices from noisy data by hard thresholding of singular values, in which empirical singular values below a threshold λ are set to 0. We study the asymptotic MSE (AMSE) in a framework where the matrix size is large compared to the rank of the matrix to be recovered, and the signal-to-noise ratio of the low-rank piece stays constant. The AMSE-optimal choice of hard threshold, in the case of n-by-n matrix in white noise of level σ, is simply (4/ √ 3) √ nσ ≈ 2.309 √ nσ when σ is known, or simply 2.858·ymed when σ is unknown, where ymed is the median empirical singular value. For nonsquare m by n matrices with m 6= n the thresholding coefficients 4/ √ 3 and 2.858 are replaced with different provided constants that depend on m/n. Asymptotically, this thresholding rule adapts to unknown rank and unknown noise level in an optimal manner: it is always better than hard thresholding at any other value, and is always better than ideal Truncated SVD (TSVD), which truncates at the true rank of the low-rank matrix we are trying to recover. Hard thresholding at the recommended value to recover an n-by-n matrix of rank r guarantees an AMSE at most 3nrσ. In comparison, the guarantees provided by TSVD, optimally tuned singular value soft thresholding and the best guarantee achievable by any shrinkage of the data singular values are 5nrσ, 6nrσ, and 2nrσ, respectively. The recommended value for hard threshold also offers, among hard thresholds, the best possible AMSE guarantees for recovering matrices with bounded nuclear norm. Empirical evidence suggests that performance improvement over TSVD and other popular shrinkage rules can be substantial, for different noise distributions, even in relatively small n.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

THE OPTIMAL HARD THRESHOLD FOR SINGULAR VALUES IS 4/√3 By

We consider recovery of low-rank matrices from noisy data by hard thresholding of singular values, in which empirical singular values below a prescribed threshold λ are set to 0. We study the asymptotic MSE (AMSE) in a framework where the matrix size is large compared to the rank of the matrix to be recovered, and the signal-to-noise ratio of the low-rank piece stays constant. The AMSE-optimal ...

متن کامل

Numerical Solution of Optimal Control of Time-varying Singular Systems via Operational Matrices

In this paper, a numerical method for solving the constrained optimal control of time-varying singular systems with quadratic performance index is presented. Presented method is based on Bernste in polynomials. Operational matrices of integration, differentiation and product are introduced and utilized to reduce the optimal control of time-varying singular problems to the solution of algebraic ...

متن کامل

A Multi-Stage Single-Machine Replacement Strategy Using Stochastic Dynamic Programming

In this paper, the single machine replacement problem is being modeled into the frameworks of stochastic dynamic programming and control threshold policy, where some properties of the optimal values of the control thresholds are derived. Using these properties and by minimizing a cost function, the optimal values of two control thresholds for the time between productions of two successive nonco...

متن کامل

A Two-Threshold Guard Channel Scheme for Minimizing Blocking Probability in Communication Networks

In this paper, we consider the call admission problem in cellular network with two classes of voice users. In the first part of paper, we introduce a two-threshold guard channel policy and study its limiting behavior under the stationary traffic. Then we give an algorithm for finding the optimal number of guard channels. In the second part of this paper, we give an algorithm, which minimizes th...

متن کامل

Near-Optimal Controls of a Fuel Cell Coupled with Reformer using Singular Perturbation methods

A singularly perturbed model is proposed for a system comprised of a PEM Fuel Cell(PEM-FC) with Natural Gas Hydrogen Reformer (NG-HR). This eighteenth order system is decomposedinto slow and fast lower order subsystems using singular perturbation techniques that provides tools forseparation and order reduction. Then, three different types of controllers, namely an optimal full-order,a near-opti...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2014