Manohar Shamaiah

Greedy sensor selection: Leveraging submodularity

We consider the problem of sensor selection in resource constrained sensor networks. The fusion center selects a subset of k sensors from an available pool of m sensors according to the maximum a posteriori or the maximum likelihood rule. We cast the sensor selection problem as the maximization of a submodular function over uniform matroids, and demonstrate that a greedy sensor selection algorithm achieves performance within (1 − 1 over e ) of the optimal solution. The greedy algorithm has a complexity of O(n3mk), where n is the dimension of the measurement space. The complexity of the algorithm is further reduced to O(n2mk) by exploiting certain structural features of the problem. An application to the sensor selection in linear dynamical systems where the fusion center employs Kalman filtering for state estimation is considered. Simulation results demonstrate the superior performance of the greedy sensor selection algorithm over competing techniques based on convex relaxation.

Distributed Algorithms for Spectrum Access in Cognitive Radio Relay Networks

We develop distributed algorithms for efficient spectrum access strategies in cognitive radio relay networks. In our setup, primary users permit secondary users access to the resource (spectrum) as long as they consent to aiding the primary users as relays in addition to transmitting their own data. Given a pool of primary and secondary users, we desire to optimize overall network utility by determining the best configuration/pairing of secondary users with primary users. This optimization can be stated in a form similar to the maximum weighted matching problem. Given such formulation, we develop an algorithm based on affinity propagation technique that is completely distributed in its structure. We demonstrate the convergence of the developed algorithm and show that it performs close to the optimal centralized scheme.

Message-Passing Algorithms for Coordinated Spectrum Sensing in Cognitive Radio Networks

This letter presents distributed algorithms that enable efficient spectrum sensing in cognitive radio networks. In the proposed setting, each node in the system detects different channels in a cooperative fashion in order to improve access to the spectral resources available to the cognitive radio network. The developed scheme utilizes a message-passing framework for efficient channel assignment. Simulation results establish that this class of algorithms outperform existing techniques and improves the overall sensing performance of the system.

Greedy Sensor Selection under Channel Uncertainty

Estimation in resource constrained sensor networks where the fusion center selects a fixed-size subset from a pool of available sensors observing the states of a linear dynamical system is considered. With some probability, the communication between a selected sensor and the fusion center may fail. It is shown that when the fusion center employs a Kalman filter and desires to minimize a function of the error covariance matrix, sensor selection under communication uncertainty can be cast as the maximization of a submodular function over uniform matroids. We propose a computationally efficient greedy sensor selection scheme achieving performance within (1 -1/ e ) of the optimal non-adaptive policy. Additionally, we propose an efficient adaptive greedy algorithm which achieves (1-1/e) of the optimal adaptive policy. Structural features of the problem are exploited to reduce the complexity of the greedy selection algorithms. We analyze the complexity and present simulation studies which demonstrate efficacy of the proposed techniques.

Graphical Models and Inference on Graphs in Genomics: Challenges of high-throughput data analysis

Recent technological advances in high-throughput molecular screening and DNA sequencing have enabled acquisition of enormous amounts of biological data that may provide critical information about the functionality of cells and organisms [1], help reveal mechanisms of genetic diseases and disorders [2], improve the efficiency of the drug discovery process [3], and enable development of diagnostic techniques and therapies [4]. Novel sequencing methods allow fast and affordable deciphering of individual genomes and thus enable studies of genetic variations and the effects they have on human health and medical treatments.

Message passing algorithm for inferring consensus sequence from next-generation sequencing data

In order to determine an individuals DNA sequence, sequencing platforms often employ shotgun sequencing where multiple identical copies of the DNA strand of interest are randomly fragmented and then the nucleotide content of the short fragments is determined. Assembly of the long DNA strand from short fragments is a computationally challenging task that has attracted significant amount of attention in recent years. We formulate reference-guided assembly as the inference problem on a bipartite graph and solve it using a message-passing algorithm. The message-passing algorithm does not need to rely on the quality score information which expresses reliability of the short reads. To assess the performance of the proposed methodology, we derive an expression for the probability of error of a genie-aided MAP consensus scheme. Simulation results on a Neisseria meningitidis data set demonstrate that the proposed message-passing algorithm performs close to the idealistic MAP consensus scheme.

Distributed routing in networks using affinity propagation

This paper applies affinity propagation (AP) to develop distributed solutions for routing over networks. AP is a message passing algorithm for unsupervised learning. This paper demonstrates that AP can be generalized and applied to a wide class of problems in networking. In particular, AP can be used to develop distributed routing mechanisms for networks. Simulation results demonstrate that the proposed schemes compare favorably with the existing methods.

Estimating Time-Varying Sparse Signals Under Communication Constraints

In this correspondence, we consider reconstruction of time-varying sparse signals in a sensor network with communication constraints. In each time interval, the fusion center transmits the predicted signal estimate and its corresponding error covariance to a selected subset of sensors. The selected sensors compute quantized innovations and transmit them to the fusion center. We present algorithms for sparse signal estimation in the described scenario, analyze their complexity, and demonstrate their near-optimal performance even in the case where sensors transmit a single bit (i.e., the sign of innovation) to the fusion center.

Inference of gene-regulatory networks using message-passing algorithms

We present an application of message-passing techniques to gene regulatory network inference. The network inference is posed as a constrained linear regression problem, and solved by a distributed computationally efficient message-passing algorithm. Performance of the proposed algorithm is tested on gold standard data sets and evaluated using metrics provided by the DREAM2 challenge [1]. Performance of the proposed algorithm is comparable to that of the techniques which yielded the best results in the DREAM2 challenge competition.

Rao-Blackwellized unscented Kalman filter for nonlinear systems with bandwidth constraints

We consider the state estimation problem in distributed nonlinear systems with bandwidth constraints. In particular, we focus on the sensor networks with limited communication between the sensor nodes and the fusion center. Two practical bandwidth-saving methods are considered: (1) recursive filtering with quantized innovations, and (2) compressive sampling of sparse signals. For both scenarios, Rao-Blackwellized unscented Kalman filter (RBUKF) based methods are developed. The simulation results demonstrate that the proposed algorithms closely track the original signal.