Graph-theoretical Constructions for Graph Entropy and Network Coding Based Communications

The guessing number of a directed graph (digraph), equivalent to the entropy of that digraph, was introduced as a direct criterion on the solvability of a network coding instance. This paper makes two contributions on the guessing number. First, we introduce an undirected graph on all possible configurations of the digraph, referred to as the guessing graph, which encapsulates the essence of dependence amongst configurations. We prove that the guessing number of a digraph is equal to the logarithm of the independence number of its guessing graph. Therefore, network coding solvability is no more a problem on the operations made by each node, but is simplified into a problem on the messages that can transit through the network. By studying the guessing graph of a given digraph, and how to combine digraphs or alphabets, we are thus able to derive bounds on the guessing number of digraphs. Second, we construct specific digraphs with high guessing numbers, yielding network coding instances where a large amount of information can transit. We first propose a construction of digraphs with finite parameters based on cyclic codes, with guessing number equal to the degree of the generator polynomial. We then construct an infinite class of digraphs with arbitrary girth for which the ratio between the linear guessing number and the number of vertices tends to one, despite these digraphs being arbitrarily sparse. These constructions yield solvable network coding instances with a relatively small number of intermediate nodes for which the node operations are known and linear, although these instances are sparse and the sources are arbitrarily far from their corresponding sinks.

💡 Research Summary

The paper tackles the long‑standing problem of determining when a network coding instance is solvable by recasting it in purely graph‑theoretic terms. The authors begin by recalling the notion of the guessing number g(D) of a directed graph D, which is known to coincide with the graph entropy H(D). Their first major contribution is the introduction of the “guessing graph” G(D), an undirected graph whose vertices correspond to all possible assignments of symbols (from a fixed alphabet) to the vertices of D. Two assignments are joined by an edge precisely when at least one node of D cannot correctly infer its own symbol from the symbols on its in‑neighbors under those assignments. In this construction, an independent set in G(D) is a collection of mutually compatible configurations; no two configurations in the set cause a conflict at any node. The authors prove the central theorem:

 g(D) = log₂ α(G(D)),

where α(G(D)) denotes the independence number of the guessing graph. The proof proceeds in two directions. First, any independent set yields a set of configurations that can be simultaneously realized, guaranteeing a feasible coding scheme whose rate equals the logarithm of the set size. Second, if two configurations are adjacent, they cannot coexist in a feasible scheme, so any feasible scheme must be an independent set. Consequently, the guessing number reduces to a classic combinatorial optimization problem: finding the largest independent set in G(D).

This reduction has immediate algorithmic and conceptual benefits. Instead of reasoning about the local functions each node may apply, one can work directly with the global set of admissible messages. Moreover, the authors show how standard graph operations translate into simple algebraic rules for guessing numbers. For example, expanding the alphabet from size q to kq multiplies the number of vertices of G(D) by k^{|V(D)|} but leaves the independence structure unchanged, so the guessing number increases by log₂ k. Similarly, the guessing number of the disjoint union of two digraphs is the sum of their guessing numbers, while the tensor product yields a multiplicative upper bound. These observations allow the derivation of a suite of bounds for composite networks.

The second major contribution is the explicit construction of families of digraphs with provably large guessing numbers. The first family is built from cyclic codes. Given a binary cyclic code of length n with generator polynomial g(x), the authors construct a digraph D_g whose adjacency matrix mirrors the parity‑check matrix of the code. They prove that the guessing number of D_g equals the degree of g(x). Because cyclic codes are well‑understood, this construction provides a systematic way to generate linear network coding instances with a prescribed guessing number, and the associated node operations are purely linear.

The second family addresses the long‑standing belief that sparse graphs cannot achieve high coding rates. By adapting techniques from the theory of high‑girth, low‑density graphs (e.g., Ramanujan‑type constructions), the authors produce an infinite sequence of digraphs with arbitrarily large girth and bounded average degree, yet whose linear guessing number g_lin(D) satisfies

 lim_{|V(D)|→∞} g_lin(D) / |V(D)| = 1.

In other words, despite being extremely sparse and having sources far from their sinks, these graphs support coding schemes that transmit essentially one bit per vertex. The construction relies on defining each vertex’s local function as a linear equation whose coefficient matrix has a sparsity pattern that guarantees large girth, thereby preventing short cycles that would otherwise create linear dependencies and reduce the independence number of the guessing graph.

These two constructions illustrate that (i) high‑rate network coding can be achieved with a small, explicitly known set of linear node operations, and (ii) sparsity and large girth do not preclude near‑optimal coding efficiency. The paper also discusses the computational implications of the guessing‑graph viewpoint. Computing α(G(D)) is a classic NP‑hard problem, but a wealth of approximation algorithms, integer‑programming formulations, and heuristic methods exist for maximum independent set and graph coloring. Consequently, the guessing‑graph framework provides a practical pathway for evaluating solvability of real‑world network coding problems, bypassing the more abstract entropy‑based methods that often lack constructive algorithms.

In summary, the authors (1) establish a precise equivalence between the guessing number of a digraph and the logarithm of the independence number of its guessing graph, thereby converting a network‑coding solvability question into a combinatorial independence problem; (2) leverage this equivalence to derive bounds for composite networks; (3) present two concrete families of digraphs—one based on cyclic codes, the other on high‑girth sparse constructions—both achieving guessing numbers that are linear in the number of vertices; and (4) argue that the guessing‑graph approach is both theoretically elegant and computationally advantageous, opening new avenues for designing efficient, linear network coding schemes in practical communication networks.