On sources and networks: Can computational tools derive information theoretic limits?

Abstract

A source-network coding problem is defined by a set of communication devices (nodes), a collection of error-free, capacitated channels (links) through which they communicate, a collection of random processes (sources) available at some nodes in the network, and a set of requirements (demands) to reconstruct those sources. The limits of such a system's performance are described by a set of achievable distortion vectors; each vector in the relative interior of this set captures the accuracies with which a single code reconstructs the demands at all receivers. The goal of this work is to build computational tools to derive provable inner and outer bounds on the set of achievable distortions for any source-network coding problem.

Additional details