Cooperative diversity (CD) networks have been receiving a lot of attention recently as distributed means improving error performance and capacity. For sufficiently large signal-to-noise ratio (SNR), this paper derives the average symbol probability (SEP) for analog forwarding CD links. The resulting expressions are general they hold an arbitrary number cooperating branches, hops per branch, various channel fading models. Their simplicity provides valuable insights to suggests optimizing...

We deal with distributed estimation of deterministic vector parameters using ad hoc wireless sensor networks (WSNs). cast the decentralized problem as solution multiple constrained convex optimization subproblems. Using method multipliers in conjunction a block coordinate descent approach we demonstrate how resultant algorithm can be decomposed into set simpler tasks suitable for implementation. Different from existing alternatives, our does not require centralized estimator to expressible...

We study deterministic mean-location parameter estimation when only quantized versions of the original observations are available, due to bandwidth constraints. When dynamic range is small or comparable with noise variance, we introduce a class maximum-likelihood estimators that require transmitting just one bit per sensor achieve an variance close (clairvoyant) sample mean estimator. larger than standard deviation, show optimum quantization step exists best possible for given constraint....

This paper provides an overview of distributed estimation-compression problems encountered with wireless sensor networks (WSN). A general formulation compression-estimation under rate constraints was introduced, pertinent signal processing algorithms were developed, and emerging tradeoffs delineated from information theoretic perspective. Specifically, we designed rate-constrained estimators for various models variable knowledge the underlying data distributions. We proved theoretically,...

When dealing with decentralized estimation, it is important to reduce the cost of communicating distributed observations-a problem receiving revived interest in context wireless sensor networks. In this paper, we derive and analyze state estimators dynamical stochastic processes, whereby low communication effected by requiring transmission a single bit per observation. Following Kalman filtering (KF) approach, develop recursive algorithms for estimation based on sign innovations (SOI). Even...

Network topology inference is a prominent problem in Science. Most graph signal processing (GSP) efforts to date assume that the underlying network known, and then analyze how graph's algebraic spectral characteristics impact properties of signals interest. Such an assumption often untenable beyond applications dealing with e.g., directly observable social infrastructure networks; typically adopted construction schemes are largely informal, distinctly lacking element validation. This...

Two architectures that generalize convolutional neural networks (CNNs) for the processing of signals supported on graphs are introduced. We start with selection graph network (GNN), which replaces linear time invariant filters shift to generate features and reinterprets pooling as a possibly nonlinear subsampling stage where nearby nodes pool their information in set preselected sample nodes. A key component architecture is remember position sampled permit computation at deeper layers. The...

We address the problem of identifying structure an undirected graph from observation signals defined on its nodes. Fundamentally, unknown encodes direct relationships between signal elements, which we aim to recover observable indirect generated by a diffusion process graph. The fresh look advocated here leverages concepts convex optimization and stationarity signals, in order identify shift operator (a matrix representation graph) given only eigenvectors. These spectral templates can be...

A new scheme to sample signals defined on the nodes of a graph is proposed. The underlying assumption that such admit sparse representation in frequency domain related structure graph, which captured by so-called graph-shift operator. Instead using value signal observed at subset recover entire sampling proposed here uses as input observations taken single node. correspond sequential applications operator, are linear combinations information gathered neighbors When corresponds directed cycle...

We study the optimal design of graph filters (GFs) to implement arbitrary linear transformations between signals. GFs can be represented by matrix polynomials graph-shift operator (GSO). Since this captures local structure graph, naturally give rise distributed network operators. In most setups, GSO is given so that GF consists fundamentally in choosing (filter) coefficients polynomial resemble desired transformations. determine spectral conditions under which a specific transformation...

We consider the problem of optimally allocating resources across a set transmitters and receivers in wireless network. The resulting optimization takes form constrained statistical learning, which solutions can be found model-free manner by parameterizing resource allocation policy. Convolutional neural networks architectures are an attractive option for parameterization, as their dimensionality is small does not scale with network size. introduce random edge graph (REGNN), performs...

This paper considers the design of optimal resource allocation policies in wireless communication systems, which are generically modeled as a functional optimization problem with stochastic constraints. These problems have structure learning statistical loss appears constraint, motivating development methodologies to attempt their solution. To handle constraints, training is undertaken dual domain. It shown that this can be done small optimality when using near-universal parameterizations....

Stationarity is a cornerstone property that facilitates the analysis and processing of random signals in time domain. Although time-varying are abundant nature, many practical scenarios information interest resides more irregular graph domains. This lack regularity hampers generalization classical notion stationarity to signals. The contribution this paper twofold. Firstly, we propose definition weak for takes into account structure where process place, while inheriting meaningful properties...

Modern neuroimaging techniques provide us with unique views on brain structure and function; i.e., how the is wired, where when activity takes place. Data acquired using these can be analyzed in terms of its network to reveal organizing principles at systems level. Graph representations are versatile models nodes associated regions edges structural or functional connections. Structural graphs model neural pathways white matter, which anatomical backbone between regions. Functional built...

This paper develops the Decentralized Linearized Alternating Direction Method of Multipliers (DLM) that minimizes a sum local cost functions in multiagent network. The algorithm mimics operation decentralized alternating direction method multipliers (DADMM) except it linearizes optimization objective at each iteration. results iterations that, instead successive minimizations, implement steps whose is akin to much lower gradient descent step used distributed (DGM). proven converge optimal...

Effective communication is key to successful, decentralized, multi-robot path planning. Yet, it far from obvious what information crucial the task at hand, and how when must be shared among robots. To side-step these issues move beyond hand-crafted heuristics, we propose a combined model that automatically synthesizes local decision-making policies for robots navigating in constrained workspaces. Our architecture composed of convolutional neural network (CNN) extracts adequate features...

In this paper, we address tracking of a time-varying parameter with unknown dynamics. We formalize the problem as an instance online optimization in dynamic setting. Using gradient descent, propose method that sequentially predicts value and turn suffers loss. The objective is to minimize accumulation losses over time horizon, notion termed regret. While existing methods focus on convex loss functions, consider strongly functions so provide better guarantees performance. derive regret bound...

This paper presents methods to analyze functional brain networks and signals from graph spectral perspectives. The notion of frequency filters traditionally defined for supported on regular domains such as discrete time image grids has been recently generalized irregular domains, defines frequencies associated with different levels spatial smoothness across the regions. Brain network also enables decomposition into pieces corresponding smooth or rapid variations. We relate principal...

Graph neural networks (GNNs) have emerged as a powerful tool for nonlinear processing of graph signals, exhibiting success in recommender systems, power outage prediction, and motion planning, among others. GNNs consist cascade layers, each which applies convolution, followed by pointwise nonlinearity. In this work, we study the impact that changes underlying topology on output GNN. First, show are permutation equivariant, implies they effectively exploit internal symmetries topology. Then,...

We study the problem of minimizing a sum convex objective functions, where components are available at different nodes network and allowed to only communicate with their neighbors. The use distributed gradient methods is common approach solve this problem. Their popularity notwithstanding, these exhibit slow convergence consequent large number communications between optimal argument because they rely on first-order information only. This paper proposes Newton (NN) method as algorithm that...

Wireless sensor networks (WSNs) deployed to perform surveillance and monitoring tasks have operate under stringent energy bandwidth limitations. These motivate well distributed estimation scenarios where sensors quantize transmit only one, or a few bits per observation, for use in forming parameter estimators of interest. In companion paper, we developed algorithms studied interesting tradeoffs that emerge even the simplest setup estimating scalar location presence zero-mean additive white...

Distributed algorithms are developed for optimal estimation of stationary random signals and smoothing (even nonstationary) dynamical processes based on generally correlated observations collected by ad hoc wireless sensor networks (WSNs). Maximum a posteriori (MAP) linear minimum mean-square error (LMMSE) schemes, well appreciated centralized estimation, shown possible to reformulate distributed operation through the iterative (alternating-direction) method multipliers. Sensors communicate...

This paper considers the control of a linear plant when state information is being transmitted from sensor to controller over wireless fading channel. The power allocated these transmissions determines probability successful packet reception and allowed adapt online both channel conditions state. goal design input transmit policies that minimize an infinite horizon cost combining expenses conventional quadratic regulator cost. Since inputs powers are in general coupled, restricted structure...