19
10

Affine Invariant Covariance Estimation for Heavy-Tailed Distributions

Abstract

In this work we provide an estimator for the covariance matrix of a heavy-tailed multivariate distributionWe prove that the proposed estimator S^\widehat{\mathbf{S}} admits an \textit{affine-invariant} bound of the form \[(1-\varepsilon) \mathbf{S} \preccurlyeq \widehat{\mathbf{S}} \preccurlyeq (1+\varepsilon) \mathbf{S}\]in high probability, where S\mathbf{S} is the unknown covariance matrix, and \preccurlyeq is the positive semidefinite order on symmetric matrices. The result only requires the existence of fourth-order moments, and allows for ε=O(κ4dlog(d/δ)/n)\varepsilon = O(\sqrt{\kappa^4 d\log(d/\delta)/n}) where κ4\kappa^4 is a measure of kurtosis of the distribution, dd is the dimensionality of the space, nn is the sample size, and 1δ1-\delta is the desired confidence level. More generally, we can allow for regularization with level λ\lambda, then dd gets replaced with the degrees of freedom number. Denoting cond(S)\text{cond}(\mathbf{S}) the condition number of S\mathbf{S}, the computational cost of the novel estimator is O(d2n+d3log(cond(S)))O(d^2 n + d^3\log(\text{cond}(\mathbf{S}))), which is comparable to the cost of the sample covariance estimator in the statistically interesing regime ndn \ge d. We consider applications of our estimator to eigenvalue estimation with relative error, and to ridge regression with heavy-tailed random design.

View on arXiv
Comments on this paper