CIF: Small: Collaborative Research: Towards a Paradigm-shift in Distributed Information Processing: Harnessing Group-structure and Interaction

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

"This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)." With a vision to fully realize the potential of next generation communication network infrastructure based on ubiquitous sensor nodes, this research introduces new architectures and strategies, for distributed in-network information processing, which directly harness: the structure of the objective performance metrics for information processing, the structure of the underlying statistical dependencies in the information gathered by sensing nodes, the topology of the network, and the capability for bidirectional interactive information exchange. This research advocates a two-fold paradigm shift in network information processing: 1) a shift from the traditional goal of data transport to the goal of function computation and 2) a shift from unidirectional models of information flow to interactive information flow models. This research supports the education of future scientists and engineers by integrating research advances with curriculum development and supports diversity by encouraging the participation of women and under-represented groups. This research develops two fundamentally new classes of code ensembles for interactive information processing. The first is based on techniques from abstract algebra and random graph theory to capture the structure of functions being computed at destinations. The second is based on techniques from communication complexity and multiterminal information theory to capture interactive structures of information flow in the network. These new code classes have superseded the performance of random code ensembles used in network information theory since its inception. This research develops new analytical frameworks and tools to uncover the fundamental performance limits of interactive information processing in sensor networks. This research facilitates the cross-pollination of research fields by providing components which build bridges between four fundamental areas, namely, information and coding theory, abstract algebra, random graph theory, and communication complexity theory.