Abstract:A covariance matrix estimator using two bits per entry was recently developed by Dirksen, Maly and Rauhut [Annals of Statistics, 50(6), pp. 3538-3562]. The estimator achieves near minimax rate for general sub-Gaussian distributions, but also suffers from two downsides: theoretically, there is an essential gap on operator norm error between their estimator and sample covariance when the diagonal of the covariance matrix is dominated by only a few entries; practically, its performance heavily relies on the dithering scale, which needs to be tuned according to some unknown parameters. In this work, we propose a new 2-bit covariance matrix estimator that simultaneously addresses both issues. Unlike the sign quantizer associated with uniform dither in Dirksen et al., we adopt a triangular dither prior to a 2-bit quantizer inspired by the multi-bit uniform quantizer. By employing dithering scales varying across entries, our estimator enjoys an improved operator norm error rate that depends on the effective rank of the underlying covariance matrix rather than the ambient dimension, thus closing the theoretical gap. Moreover, our proposed method eliminates the need of any tuning parameter, as the dithering scales are entirely determined by the data. Experimental results under Gaussian samples are provided to showcase the impressive numerical performance of our estimator. Remarkably, by halving the dithering scales, our estimator oftentimes achieves operator norm errors less than twice of the errors of sample covariance.

A complete parameterization of all positive rational extensions of a covariance sequence

Multidimensional Rational Covariance Extension with Applications to Spectral Estimation and Image Compression

Multidimensional Rational Covariance Extension with Approximate Covariance Matching

Parametrizing spectral filters for nonsingular polynomial matrices

Separable Expansions for Covariance Estimation

Covariance Estimation in High Dimensions Via Kronecker Product Expansions

Role of the extended Hilbert space in the attainability of the Quantum Cramér-Rao bound for multiparameter estimation

Asymptotic analysis for covariance parameter estimation of Gaussian processes with functional inputs

Parameter Estimation on Homogeneous Spaces

A parameter ASIP for the quadratic family

Rationality and Parametrizations of Algebraic Curves under Specializations

Unified Analysis of Periodization-Based Sampling Methods for Matérn Covariances

A Spectral Algorithm for List-Decodable Covariance Estimation in Relative Frobenius Norm

Quantitatively hyper-positive real rational functions II

Towards a Unified Formalism of Multivariate Coefficients of Variation -- Application to the Analysis of Polarimetric Speckle Time Series

On the exact spectral factorization of rational matrix functions with applications to paraunitary filter banks

Malliavin-Skorohod calculus and Paley-Wiener integral for covariance singular processes

Positive Definite Kernels and Random Sequences Connected to Polynomial Hypergroups

A Parameter-Free Two-Bit Covariance Estimator with Improved Operator Norm Error Rate

Rational Points Near Manifolds, Homogeneous Dynamics, and Oscillatory Integrals