Toward A Graph-based Discrete Information Interface for Network Communication

S. Sandeep Pradhan

EECS Department University of Michigan

Abstract:

The communication problem that involves transmission of correlated information sources over multiuser channels is considered. For this problem, traditional separation-approach is not optimal. We consider a graph-based framework for this information transmission problem. The system involves a source coding module and a channel coding module. In the source coding module, the sources are efficiently mapped into a bipartite graph, and in the channel coding module, the edges of this graph are reliably transmitted over a multiuser channel. We consider bipartite graphs as discrete information interface between source coding and channel coding in this multiterminal setting. We provide an information-theoretic characterization of (1) the rate of exponential growth (as a function of the number of channel uses) of the size of the bipartite graphs whose edges can be reliably transmitted over a multiuser channel and (2) the rate of exponential growth (as a function of the number of source samples) of the size of the bipartite graphs which can reliably represent a pair of correlated sources to be transmitted over a multiuser channel.

Tuesday, February 19, 2008

4:00-5:00 pm

Room EECS 1005