English

Optimal query complexity for estimating the trace of a matrix

Computational Complexity 2014-05-29 v1 Data Structures and Algorithms

Abstract

Given an implicit n×nn\times n matrix AA with oracle access xTAxx^TA x for any xRnx\in \mathbb{R}^n, we study the query complexity of randomized algorithms for estimating the trace of the matrix. This problem has many applications in quantum physics, machine learning, and pattern matching. Two metrics are commonly used for evaluating the estimators: i) variance; ii) a high probability multiplicative-approximation guarantee. Almost all the known estimators are of the form 1ki=1kxiTAxi\frac{1}{k}\sum_{i=1}^k x_i^T A x_i for xiRnx_i\in \mathbb{R}^n being i.i.d. for some special distribution. Our main results are summarized as follows. We give an exact characterization of the minimum variance unbiased estimator in the broad class of linear nonadaptive estimators (which subsumes all the existing known estimators). We also consider the query complexity lower bounds for any (possibly nonlinear and adaptive) estimators: (1) We show that any estimator requires Ω(1/ϵ)\Omega(1/\epsilon) queries to have a guarantee of variance at most ϵ\epsilon. (2) We show that any estimator requires Ω(1ϵ2log1δ)\Omega(\frac{1}{\epsilon^2}\log \frac{1}{\delta}) queries to achieve a (1±ϵ)(1\pm\epsilon)-multiplicative approximation guarantee with probability at least 1δ1 - \delta. Both above lower bounds are asymptotically tight. As a corollary, we also resolve a conjecture in the seminal work of Avron and Toledo (Journal of the ACM 2011) regarding the sample complexity of the Gaussian Estimator.

Keywords

Cite

@article{arxiv.1405.7112,
  title  = {Optimal query complexity for estimating the trace of a matrix},
  author = {Karl Wimmer and Yi Wu and Peng Zhang},
  journal= {arXiv preprint arXiv:1405.7112},
  year   = {2014}
}

Comments

full version of the paper in ICALP 2014

R2 v1 2026-06-22T04:24:46.562Z