On Achievability via Random Binning

In [1], the authors present a novel tool to establish achievability results in network information theoretic problems. The main idea is to study a set of discrete memoryless correlated sources, and the ensuing statistics of the bin indices resulting from random binning of the sources. Sufficient conditions are established on rates of random binning such that these indices are mutually independent and uniformly distributed. In conjunction with the Slepian-Wolf result, such a characterization allows for a new framework to prove achievability in both rate-distortion and channel coding problems. In this paper, we study the main techniques employed in [1], which we then apply to rederive several known results such as Marton’s inner bound for broadcast channels, Gelfand-Pinsker theorem, Wyner-Ziv theorem, BergerTung inner bound etc. While this framework is not fundamentally different from the one involving typicality arguments, it presents a unified view of achievability, and is particularly useful in problems of secrecy and coordination.