In this paper, the problem of designing linear precoders for fixed multiple-input-multiple-output (MIMO) receivers is considered. Two different design criteria are first, transmitted power minimized subject to signal-to-interference-plus-noise-ratio (SINR) constraints. second, worst case SINR maximized a constraint. It shown that both problems can be solved using standard conic optimization packages. addition, conditions developed optimal precoder these problems, and two simple fixed-point...

Abstract Does the default mode network (DMN) reconfigure to encode information about changing environment? This question has proven difficult, because patterns of functional connectivity reflect a mixture stimulus-induced neural processes, intrinsic processes and non-neuronal noise. Here we introduce inter-subject correlation (ISFC), which isolates stimulus-dependent inter-regional correlations between brains exposed same stimulus. During fMRI, had subjects listen real-life auditory...

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> We consider the problem of linear zero-forcing precoding design and discuss its relation to theory generalized inverses in algebra. Special attention is given a specific inverse known as pseudo-inverse. begin with standard under assumption total power constraint prove that precoders based on pseudo-inverse are optimal among this setting. Then, we proceed examine individual per-antenna...

We address covariance estimation in the sense of minimum mean-squared error (MMSE) for Gaussian samples. Specifically, we consider shrinkage methods which are suitable high dimensional problems with a small number samples (large p n). First, improve on Ledoit-Wolf (LW) method by conditioning sufficient statistic. By Rao-Blackwell theorem, this yields new estimator called RBLW, whose dominates that LW variables. Second, to further reduce error, propose an iterative approach approximates...

In this paper we consider Multiple-Input-Multiple-Output (MIMO) detection using deep neural networks. We introduce two different architectures: a standard fully connected multi-layer network, and Detection Network (DetNet) which is specifically designed for the task. The structure of DetNet obtained by unfolding iterations projected gradient descent algorithm into network. compare accuracy runtime complexity purposed approaches achieve state-of-the-art performance while maintaining low...

In this paper, we consider the use of deep neural networks in context Multiple-Input-Multiple-Output (MIMO) detection. We give a brief introduction to learning and propose modern network architecture suitable for detection task. First, case which MIMO channel is constant, learn detector specific system. Next, harder parameters are known yet changing single must be learned all multiple varying channels. demonstrate performance our using numerical simulations comparison competing methods...

We address high dimensional covariance estimation for elliptical distributed samples, which are also known as spherically invariant random vectors (SIRV) or compound-Gaussian processes. Specifically we consider shrinkage methods that suitable problems with a small number of samples (large $p$ $n$). start from classical robust estimator [Tyler(1987)], is distribution-free within the family distribution but inapplicable when $n<p$. Using coefficient, regularize Tyler's fixed point iterations....

Wyner's wiretap channel is generalized to the case when sender, receiver and eavesdropper have multiple antennas. We consider two cases: deterministic fading case. In case, matrices of intended are fixed known all nodes. experience block sender has only receiver's state information (CSI) statistical knowledge eavesdropper's channel. For a scheme based on generalized-singular-value-decomposition (GSVD) proposed shown achieve secrecy capacity in high signal-to-noise-ratio (SNR) limit. When one...

Geodesic convexity is a generalization of classical which guarantees that all local minima g-convex functions are globally optimal. We consider with positive definite matrix variables, and prove Kronecker products, logarithms determinants g-convex. apply these results to two modern covariance estimation problems: robust in scaled Gaussian distributions, structured models. Maximum likelihood settings involves non-convex minimizations. show problems fact This leads straight forward analysis,...

Phasor measurement units (PMUs) are time synchronized sensors primarily used for power system state estimation. Despite their increasing incorporation and the ongoing research on estimation using measurements from these sensors, with imperfect phase synchronization has not been sufficiently investigated. Inaccurate is an inevitable problem that large scale deployment of PMUs to face. In this paper, we introduce a model mismatch. We propose alternating minimization parallel Kalman filtering...

We consider covariance estimation in the multivariate generalized Gaussian distribution (MGGD) and elliptically symmetric (ES) distribution. The maximum likelihood optimization associated with this problem is non-convex, yet it has been proved that its global solution can be often computed via simple fixed point iterations. Our first contribution a new analysis of based on geodesic convexity requires weaker assumptions. second framework for structured under sparsity constraints. show...

We develop a computationally efficient approximation of the maximum likelihood (ML) detector for 16 quadrature amplitude modulation (16-QAM) in multiple-input multiple-output (MIMO) systems. The is based on convex relaxation ML problem. resulting optimization semidefinite program that can be solved polynomial time with respect to number inputs system. Simulation results random MIMO system show proposed algorithm outperforms conventional decorrelator by about 2.5 dB at high signal-to-noise ratios.

We consider regularized covariance estimation in scaled Gaussian settings, e.g., elliptical distributions, compound-Gaussian processes and spherically invariant random vectors. Asymptotically the number of samples, classical maximum likelihood (ML) estimate is optimal under different criteria can be efficiently computed even though optimization nonconvex. propose a unified framework for regularizing this order to improve its finite sample performance. Our approach based on discovery hidden...

In this paper, we consider the problem of estimating an unknown deterministic parameter vector in a linear regression model with random Gaussian uncertainty mixing matrix. We prove that maximum-likelihood (ML) estimator is (de)regularized least squares and develop three alternative approaches for finding regularization maximizes likelihood. analyze performance using Cramer-Rao bound (CRB) on mean squared error, show degradation due not as severe may be expected. Next, address again assuming...

We consider large scale covariance estimation using a small number of samples in applications where there is natural ordering between the random variables. The two classical approaches to this problem rely on banded and inverse structures, corresponding time varying moving average (MA) autoregressive (AR) models, respectively. Motivated by analogy spectral well known modeling power (ARMA) processes, we propose novel ARMA structure. Similarly results context AR MA, address completion an...

Non-data-aided (NDA) signal-to-noise-ratio (SNR) estimation is considered for binary phase shift keying systems where the data samples are governed by a normal mixture distribution. Inherent accuracy limitations examined via simple, closed-form approximation to associated Cramer-Rao bound which eliminates need numerical integration. The expectation-maximization algorithm proposed iteratively maximize NDA likelihood function. Simulation results show that resulting estimator offers statistical...

We consider distributed estimation of the inverse covariance matrix in Gaussian graphical models. These models factorize multivariate distribution and allow for efficient signal processing methods such as belief propagation (BP). The classical maximum likelihood approach to this problem, or potential function BP terminology, requires centralized computing is computationally intensive. This motivates suboptimal alternatives that tradeoff accuracy communication cost. A natural solution each...

The scalar shrinkage-thresholding operator is a key ingredient in variable selection algorithms arising wavelet denoising, JPEG2000 image compression and predictive analysis of gene microarray data. In these applications, the decision to select given as solution sparsity penalized quadratic optimization. some other one seeks multidimensional variables. this work, we present natural extension shrinkage thresholding operator. Similarly case, threshold determined by minimization convex form...

We address structured covariance estimation in elliptical distributions by assuming that the is a priori known to belong given convex set, e.g., set of Toeplitz or banded matrices. consider General Method Moments (GMM) optimization applied robust Tyler's scatter M-estimator subject these constraints. Unfortunately, GMM turns out be non-convex due objective. Instead, we propose new COCA estimator-a relaxation which can efficiently solved. prove tight unconstrained case for finite number...

We provide new results on the sphere demodulator (SD) and its variant, list (LSD). A reduced complexity SD algorithm is presented. suggest implementing necessary sorting of constellation symbols through a look up table. In addition, we propose use modified QR which tries to begin search in layers are "easier demodulate". These improvements also apply recently published soft output LSD, implement using heap structure.

We address covariance estimation under mean-squared loss in the Gaussian setting. Specifically, we consider shrinkage methods which are suitable for high dimensional problems with small number of samples (large p n). First, improve on Ledoit-Wolf (LW) method by conditioning a sufficient statistic via Rao-Blackwell theorem, obtaining new estimator RBLW whose error dominates LW model. Second, to further reduce error, propose an iterative approach approximates clairvoyant estimator. Convergence...

We consider principal component analysis (PCA) in decomposable Gaussian graphical models. exploit the prior information these models order to distribute its computation. For this purpose, we reformulate problem sparse inverse covariance (concentration) domain and solve global eigenvalue using a sequence of local problems each cliques graph. demonstrate application our methodology context decentralized anomaly detection Abilene backbone network. Based on topology network, propose an...

Graphical models are a framework for representing and exploiting prior conditional independence structures within distributions using graphs. In the Gaussian case, these directly related to sparsity of inverse covariance (concentration) matrix allow improved estimation with lower computational complexity. We consider concentration mean-squared error (MSE) as objective, in special type model known decomposable. This includes, example, well banded structure other cases encountered practice....

This paper analyzes the performance of Tyler's Mestimator scatter matrix in elliptical populations.We focus on non-asymptotic setting and derive estimation error bounds depending number samples n dimension p.We show that under quite mild conditions squared Frobenius norm inverse estimator decays like p 2 /n with high probability.