Sparse Recovery in Peer to Peer Networks via Compressive Sensing

Please enable javascript in your browser.

Fattaholmanan Najafabadi, Ali | 2014

484 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 47089 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Rabiei, Hamid Reza
Abstract:
Monitoring large-scale networks is a critical yet challenging task. Enormous number of nodes and links, limited power, and lack of direct acceß to the entire network,are the most important difficulties. In applications such as network routing,where all nodes need to monitor the status of the entire network, the situation is even worse. In this thesis, a collaborative model in which nodes pick up information from measurements generated by other nodes, is proposed. Considering the fact that in most cases the networked data is sufficiently sparse, we used the Compreßive Sensing theory in the recovery phase of the proposed method. Using this model, for the first time, an upper bound is derived for the number of measurements that each node must generate, such that the expected number of measurements observed by each node is sufficient to provide a global view of the entire networked data. Finally, by using this upper bound, an efficient optimization method is introduced to minimize the total number of measurements. The feasibility and accuracy of the proposed method is verified through extensive numerical simulations both on real and synthetic datasets
Keywords:
Peer-to-Peer Network ; Compressive Sensing ; Sparse Recovery ; Sparse Data Recovery ; Distributed Monitoring