Layered Noisy Network Coding

Noisy network coding naturally combines two parallel lines of work on network information flow --- network coding over noiseless networks by Ahlswede, Cai, Li, and Yeung, and compress--forward coding for the relay channel by Cover and El Gamal and extends both results to general discrete memoryless and Gaussian networks. In particular, it achieves the best known capacity inner bounds for several multi-source multicast networks including deterministic networks by Avestimehr, Diggavi, and Tse and erasure networks by Dana, Gowaikar, Palanki, Hassibi, and Effros. In this paper we further improve noisy network coding for the two-way relay channel. In the new scheme, instead of the relay node just sending a common compression index to both destinations, the relay compresses its observation into a description of two layers. The common layer is used at both decoders, while the refinement layer is used only at one of the decoders. The advantage of the new scheme is demonstrated through an example of the Gaussian two-way relay channel.