arXiv subjects

Andrzej Jarosz

Publications and source records attributed to Andrzej Jarosz.

8 recordsLinked to original sources

arXiv2022

Cleaning large-dimensional covariance matrices for correlated samples

We elucidate the problem of estimating large-dimensional covariance matrices in the presence of correlations between samples. To this end, we generalize the Marcenko-Pastur equation and the Ledoit-Peche shrinkage estimator using methods of random matrix theory and free probability. We develop an efficient algorithm that implements the corresponding analytic formulas, based on the Ledoit-Wolf kernel estimation technique. We also provide an associated open-source Python library, called "shrinkage", with a user-friendly API to assist in practical tasks of estimation of large covariance matrices. We present an example of its usage for synthetic data generated according to exponentially-decaying auto-correlations.

math-ph↗

arXiv2015

Random matrix theory of quantum transport in chaotic cavities with non-ideal leads

We determine the joint probability density function (JPDF) of reflection eigenvalues in three Dyson's ensembles of normal-conducting chaotic cavities coupled to the outside world through both ballistic and tunnel point contacts. Expressing the JPDF in terms of hypergeometric functions of matrix arguments (labeled by the Dyson index $β$), we further show that reflection eigenvalues form a determinantal ensemble at $β=2$ and a new type of a Pfaffian ensemble at $β=4$. As an application, we derive a simple analytic expression for the concurrence distribution describing production of orbitally entangled electrons in chaotic cavities with tunnel point contacts when time reversal symmetry is preserved.

cond-mat.mes-hall↗

arXiv2012

Hermitian and non-Hermitian covariance estimators for multivariate Gaussian and non-Gaussian assets from random matrix theory

The random matrix theory method of planar Gaussian diagrammatic expansion is applied to find the mean spectral density of the Hermitian equal-time and non-Hermitian time-lagged cross-covariance estimators, firstly in the form of master equations for the most general multivariate Gaussian system, secondly for seven particular toy models of the true covariance function. For the simplest one of these models, the existing result is shown to be incorrect and the right one is presented, moreover its generalizations are accomplished to the exponentially-weighted moving average estimator as well as two non-Gaussian distributions, Student t and free Levy. The paper revolves around applications to financial complex systems, and the results constitute a sensitive probe of the true correlations present there.

q-fin.ST↗

arXiv2012

Generalized Bures products from free probability

Inspired by the theory of quantum information, I use two non-Hermitian random matrix models - a weighted sum of circular unitary ensembles and a product of rectangular Ginibre unitary ensembles - as building blocks of three new products of random matrices which are generalizations of the Bures model. I apply the tools of both Hermitian and non-Hermitian free probability to calculate the mean densities of their eigenvalues and singular values in the thermodynamic limit, along with their divergences at zero; the results are supported by Monte Carlo simulations. I pose and test conjectures concerning the relationship between the two densities (exploiting the notion of the N-transform), the shape of the mean domain of the eigenvalues (an extension of the single ring theorem), and the universal behavior of the mean spectral density close to the domain's borderline (using the complementary error function).

math-ph↗

arXiv2011

Summing free unitary random matrices

I use quaternion free probability calculus - an extension of free probability to non-Hermitian matrices (which is introduced in a succinct but self-contained way) - to derive in the large-size limit the mean densities of the eigenvalues and singular values of sums of independent unitary random matrices, weighted by complex numbers. In the case of CUE summands, I write them in terms of two "master equations," which I then solve and numerically test in four specific cases. I conjecture a finite-size extension of these results, exploiting the complementary error function. I prove a central limit theorem, and its first sub-leading correction, for independent identically-distributed zero-drift unitary random matrices.

math-ph↗

arXiv2010

A Random Matrix Approach to VARMA Processes

We apply random matrix theory to derive spectral density of large sample covariance matrices generated by multivariate VMA(q), VAR(q) and VARMA(q1,q2) processes. In particular, we consider a limit where the number of random variables N and the number of consecutive time measurements T are large but the ratio N/T is fixed. In this regime the underlying random matrices are asymptotically equivalent to Free Random Variables (FRV). We apply the FRV calculus to calculate the eigenvalue density of the sample covariance for several VARMA-type processes. We explicitly solve the VARMA(1,1) case and demonstrate a perfect agreement between the analytical result and the spectra obtained by Monte Carlo simulations. The proposed method is purely algebraic and can be easily generalized to q1>1 and q2>1.

q-fin.ST↗

arXiv2006

Addition of Free Unitary Random Matrices

We consider a new class of non-Hermitian random matrices, namely the ones which have the form of sums of freely independent terms involving unitary matrices. To deal with them, we exploit the recently developed quaternion technique. After having derived some general identities describing additive properties of unitary matrices, we solve three particular models: CUE plus CUE, CUE plus... plus CUE (i. e. the sum of an arbitrary number of CUE matrices), and CUE plus GUE. By solution of a given model we mean here to calculate the borderline of the eigenvalues' two-dimensional domain, as well as the eigenvalues' density function inside the domain. We confirm numerically all the results, obtaining very good agreement.

math-ph↗

arXiv2004

A Novel Approach to Non-Hermitian Random Matrix Models

In this paper we propose a new method for studying spectral properties of the non-hermitian random matrix ensembles. Alike complex Green's function encodes, via discontinuities, the real spectrum of the hermitian ensembles, the proposed here quaternion extension of the Green's function leads directly to complex spectrum in case of non-hermitian ensembles and encodes additionally some spectral properties of the eigenvectors. The standard two-by-two matrix representation of the quaternions leads to generalization of so-called matrix-valued resolvent, proposed recently in the context of diagrammatic methods [1-6]. We argue that quaternion Green's function obeys Free Variables Calculus [7,8]. In particular, the quaternion functional inverse of the matrix Green's function, called after [9] Blue's function obeys simple addition law, as observed some time ago [1,3]. Using this law we derive new, general, algorithmic and efficient method to find the non-holomorphic Green's function for all non-hermitian ensembles of the form H+iH', where ensembles H and H' are independent (free in the sense of Voiculescu [7]) hermitian ensembles from arbitrary measure. We demonstrate the power of the method by a straightforward rederivation of spectral properties for several examples of non-hermitian random matrix models.

math-ph↗