arXiv subjects

Mario Diaz

Publications and source records attributed to Mario Diaz.

27 records · Page 2Linked to original sources

arXiv2018

On the Contractivity of Privacy Mechanisms

We present a novel way to compare the statistical cost of privacy mechanisms using their Dobrushin coefficient. Specifically, we provide upper and lower bounds for the Dobrushin coefficient of a privacy mechanism in terms of its maximal leakage and local differential privacy guarantees. Given the geometric nature of the Dobrushin coefficient, this approach provides some insights into the general statistical cost of these privacy guarantees. We highlight the strength of this method by applying our results to develop new bounds on the $\ell_2$-minimax risk in a distribution estimation setting under maximal leakage constraints. Specifically, we find its order with respect to the sample size and privacy level, allowing a quantitative comparison with the corresponding local differential privacy $\ell_2$-minimax risk.

cs.IT↗

arXiv2018

On the Noise-Information Separation of a Private Principal Component Analysis Scheme

In a survey disclosure model, we consider an additive noise privacy mechanism and study the trade-off between privacy guarantees and statistical utility. Privacy is approached from two different but complementary viewpoints: information and estimation theoretic. Motivated by the performance of principal component analysis, statistical utility is measured via the spectral gap of a certain covariance matrix. This formulation and its motivation rely on classical results from random matrix theory. We prove some properties of this statistical utility function and discuss a simple numerical method to evaluate it.

cs.IT↗

arXiv2017

On the Global Fluctuations of Block Gaussian Matrices

In this paper we study the global fluctuations of block Gaussian matrices within the framework of second-order free probability theory. In order to compute the second-order Cauchy transform of these matrices, we introduce a matricial second-order conditional expectation and compute the matricial second-order Cauchy transform of a certain type of non-commutative random variables. As a by-product, using the linearization technique, we obtain the second-order Cauchy transform of non-commutative rational functions evaluated on selfadjoint Gaussian matrices.

math.PR↗

arXiv2017

On the Capacity of Block Multiantenna Channels

In this paper we consider point-to-point multiantenna channels with certain block distributional symmetries which do not require the entries of the channel matrix to be either Gaussian, or independent, or identically distributed. A main contribution is a capacity theorem for these channels, which might be regarded as a generalization of Telatar's theorem (1999), which reduces the numerical optimization domain in the capacity computation. With this information theoretic result and some free probability arguments, we prove an asymptotic capacity theorem that, in addition to reducing the optimization domain, does not depend on the dimension of the channel matrix. This theorem allows us to apply free probability techniques to numerically compute the asymptotic capacity of the channels under consideration. These theorems provide a very efficient method for numerically approximating both the capacity and a capacity achieving input covariance matrix of certain channels.

cs.IT↗

arXiv2017

Privacy-Aware Guessing Efficiency

We investigate the problem of guessing a discrete random variable $Y$ under a privacy constraint dictated by another correlated discrete random variable $X$, where both guessing efficiency and privacy are assessed in terms of the probability of correct guessing. We define $h(P_{XY}, ε)$ as the maximum probability of correctly guessing $Y$ given an auxiliary random variable $Z$, where the maximization is taken over all $P_{Z|Y}$ ensuring that the probability of correctly guessing $X$ given $Z$ does not exceed $ε$. We show that the map $ε\mapsto h(P_{XY}, ε)$ is strictly increasing, concave, and piecewise linear, which allows us to derive a closed form expression for $h(P_{XY}, ε)$ when $X$ and $Y$ are connected via a binary-input binary-output channel. For $(X^n, Y^n)$ being pairs of independent and identically distributed binary random vectors, we similarly define $\underline{h}_n(P_{X^nY^n}, ε)$ under the assumption that $Z^n$ is also a binary vector. Then we obtain a closed form expression for $\underline{h}_n(P_{X^nY^n}, ε)$ for sufficiently large, but nontrivial values of $ε$.

cs.IT↗

arXiv2016

On the Symmetries and the Capacity Achieving Input Covariance Matrices of Multiantenna Channels

In this paper we study the capacity achieving input covariance matrices of a single user multiantenna channel based solely on the group of symmetries of its matrix of propagation coefficients. Our main result, which unifies and improves the techniques used in a variety of classical capacity theorems, uses the Haar (uniform) measure on the group of symmetries to establish the existence of a capacity achieving input covariance matrix in a very particular subset of the covariance matrices. This result allows us to provide simple proofs for old and new capacity theorems. Among other results, we show that for channels with two or more standard symmetries, the isotropic input is optimal. Overall, this paper provides a precise explanation of why the capacity achieving input covariance matrices of a channel depend more on the symmetries of the matrix of propagation coefficients than any other distributional assumption.

cs.IT↗

arXiv2016

Information Extraction Under Privacy Constraints

A privacy-constrained information extraction problem is considered where for a pair of correlated discrete random variables $(X,Y)$ governed by a given joint distribution, an agent observes $Y$ and wants to convey to a potentially public user as much information about $Y$ as possible without compromising the amount of information revealed about $X$. To this end, the so-called {\em rate-privacy function} is introduced to quantify the maximal amount of information (measured in terms of mutual information) that can be extracted from $Y$ under a privacy constraint between $X$ and the extracted information, where privacy is measured using either mutual information or maximal correlation. Properties of the rate-privacy function are analyzed and information-theoretic and estimation-theoretic interpretations of it are presented for both the mutual information and maximal correlation privacy measures. It is also shown that the rate-privacy function admits a closed-form expression for a large family of joint distributions of $(X,Y)$. Finally, the rate-privacy function under the mutual information privacy measure is considered for the case where $(X,Y)$ has a joint probability density function by studying the problem where the extracted information is a uniform quantization of $Y$ corrupted by additive Gaussian noise. The asymptotic behavior of the rate-privacy function is studied as the quantization resolution grows without bound and it is observed that not all of the properties of the rate-privacy function carry over from the discrete to the continuous case.

cs.IT↗

arXiv2014

On Random Operator-Valued Matrices: Operator-Valued Semicircular Mixtures and Central Limit Theorem

Motivated by a random matrix theory model from wireless communications, we define random operator-valued matrices as the elements of $L^{\infty-}(Ω,{\mathcal F},{\mathbb P}) \otimes M_d({\mathcal A})$ where $(Ω,{\mathcal F},{\mathbb P})$ is a classical probability space and $({\mathcal A},φ)$ is a non-commutative probability space. A central limit theorem for the mean $M_d(\mathbb{C})$-valued moments of these random operator-valued matrices is derived. Also a numerical algorithm to compute the mean $M_d({\mathbb C})$-valued Cauchy transform of operator-valued semicircular mixtures is analyzed.

math.PR↗

arXiv2014

Random Matrix Systems with Block-Based Behavior and Operator-Valued Models

A model to estimate the asymptotic isotropic mutual information of a multiantenna channel is considered. Using a block-based dynamics and the angle diversity of the system, we derived what may be thought of as the operator-valued version of the Kronecker correlation model. This model turns out to be more flexible than the classical version, as it incorporates both an arbitrary channel correlation and the correlation produced by the asymptotic antenna patterns. A method to calculate the asymptotic isotropic mutual information of the system is established using operator-valued free probability tools. A particular case is considered in which we start with explicit Cauchy transforms and all the computations are done with diagonal matrices, which make the implementation simpler and more efficient.

math.PR↗