A Unified Random Coding Bound

Abstract

In this paper, we prove a unified achievability bound that generalizes and improves random coding bounds for any combination of source coding, channel coding, joint source-channel coding, and coding for computing problems assuming blockwise node operation. As a general network setup, we consider an acyclic discrete memoryless network, where the network demands and constraints are specified by a joint-typicality constraint on the whole channel input and output sequences. For achievability, a basic building block for node operation consists of simultaneous nonunique decoding, simultaneous compression, and symbol-by-symbol mapping. Our hound can be useful for deriving random coding bounds without error analysis, especially for large and complex networks. In particular, our hound can be used for unifying and generalizing many known relaying strategies. For example, a generalized decode-compressamplify-and-forward bound is obtained as a simple corollary of our main theorem, and it is shown to strictly outperform the previously known relaying schemes. Furthermore, by exploiting the symmetry in our bound, we formally define and characterize three types of network duality based on channel input-output reversal and network flow reversal combined with packing-covering duality.