[Paper Review] Multicasting correlated multi-source to multi-sink over a network

Te Sun Han|arXiv (Cornell University)|Jan 6, 2009

Cooperative Communication and Network Coding24 references3 citations

TL;DR

This paper establishes a necessary and sufficient condition for reliable multicasting of correlated multiple sources to multiple sinks over a noisy network with general capacities and stationary ergodic sources. By unifying single-source multicast (Ahlswede et al.) and correlated source-single sink (Han, 1980) models, it introduces a matching condition based on conditional entropy rates and network capacity functions, enabling joint source-channel coding without relying on Menger’s theorem or restrictive assumptions like integer capacities or error-free channels.

ABSTRACT

The problem of network coding with multicast of a single source to multisink has first been studied by Ahlswede, Cai, Li and Yeung in 2000, in which they have established the celebrated max-flow mini-cut theorem on non-physical information flow over a network of independent channels. On the other hand, in 1980, Han has studied the case with correlated multisource and a single sink from the viewpoint of polymatroidal functions in which a necessary and sufficient condition has been demonstrated for reliable transmission over the network. This paper presents an attempt to unify both cases, which leads to establish a necessary and sufficient condition for reliable transmission over a network multicasting correlated multisource to multisink. Here, the problem of separation of source coding and channel coding is also discussed.

Motivation & Objective

• To unify the single-source multicast model (Ahlswede et al.) and the correlated multiple sources to single sink model (Han, 1980) into a general framework for multicasting correlated sources to multiple sinks.
• To establish a necessary and sufficient condition for reliable transmission over noisy networks with general (non-integer, arbitrary real-valued) capacities and stationary ergodic correlated sources.
• To analyze the separation problem between distributed source coding and network coding in the context of correlated multiple sources and multiple sinks.
• To provide a computable, information-theoretic matching condition between source and channel characteristics, avoiding reliance on graph-theoretic tools like Menger’s theorem.
• To clarify the role of polymatroidal and co-polymatroidal structures in enabling or characterizing separability of source and network coding.

Proposed method

• Formulates the network as an acyclic directed graph with edge capacities representing channel capacities, allowing arbitrary non-negative real values.
• Introduces a matching condition between source and channel using conditional entropy rates of correlated sources and the network’s capacity function ρₙ(S).
• Applies polymatroid and co-polymatroid functions to characterize the feasible rate region for reliable transmission.
• Derives a necessary and sufficient condition for transmissibility based on the intersection of Slepian-Wolf rate region and network capacity region: R_SW ∩ (∩ₜ∈Ψ Cₜ) ≠ ∅.
• Uses the strong converse property of channels (Verdú-Han) to ensure reliability under general channel models.
• Demonstrates that the condition remains valid even without assuming integer bit rates, error-free channels, or memoryless sources, avoiding reliance on Menger’s theorem.

Experimental results

Research questions

• RQ1 What is the necessary and sufficient condition for reliable multicasting of correlated multiple sources to multiple sinks over a noisy network with general capacities?
• RQ2 How can joint source-channel coding be characterized in networks with correlated sources and multiple sinks, without assuming independent or memoryless channels?
• RQ3 Under what conditions does separation between distributed source coding (Slepian-Wolf) and network coding hold in multi-source, multi-sink networks?
• RQ4 Can the matching condition between source and channel be expressed in a computable and information-theoretic form without relying on graph-theoretic constructs like Menger’s theorem?
• RQ5 What is the role of polymatroidal and co-polymatroidal structures in enabling or characterizing separability in joint source-channel coding?

Key findings

• A necessary and sufficient condition for reliable transmission is established using conditional entropy rates of correlated sources and the network’s capacity function ρₙ(S), generalizing both single-source multicast and correlated source-single sink models.
• The condition is formulated as the non-emptiness of the intersection between the Slepian-Wolf rate region and the network’s capacity region: R_SW ∩ (∩ₜ∈Ψ Cₜ) ≠ ∅.
• The result holds for general stationary ergodic correlated sources and statistically independent noisy channels with arbitrary non-negative real capacities, including non-integer rates.
• The approach avoids reliance on Menger’s theorem and thus applies beyond the restrictive assumptions of prior works (e.g., integer capacities, error-free channels, memoryless sources).
• Separability of distributed source coding and network coding is both necessary and sufficient when the condition R_SW ∩ (∩ₜ∈Ψ Cₜ) ≠ ∅ holds, even without assuming polymatroidal structure.
• The framework generalizes the non-network compound Slepian-Wolf system and provides a unified information-theoretic foundation for multicasting correlated sources to multiple sinks.