Graph Colouring: A Visual Tour

Abstract: Graph colouring is a combinatorial optimisation problem with applications in several important domains, including sports scheduling, cartography, street map navigation, and timetabling. It is also of significant theoretical interest and a standard subject in university-level courses on graph theory, algorithms, and combinatorics. In this paper, we consider the topics of node, edge, and face colouring along with their associated algorithms. Theoretical results are reviewed and brought to life through a collection of detailed, visually engaging figures designed to enhance understanding and appeal.

• The paper's main contribution is a comprehensive visual exposition of graph colouring theory, integrating node, edge, and face colourings with formal definitions and reduction techniques.
• The methodology employs rigorous algorithmic analysis using approaches from greedy heuristics to hybrid evolutionary algorithms, validated through empirical testing on various graph classes.
• The study demonstrates practical applications in areas like network routing and scheduling, while elucidating classical challenges such as the NP-hardness of determining chromatic numbers and the four colour theorem.

Authoritative Summary of "Graph Colouring: A Visual Tour" (2602.18246)

Introduction and Formalism

The paper provides a comprehensive and visually guided exposition of graph colouring theory, addressing node, edge, and face colourings in a unified framework. The formal definitions of chromatic number ($\chi(G)$), chromatic index ($\chi'(G)$), and face chromatic number ($\chi_f(G)$) are introduced, serving as the foundational invariants in classical and computational graph theory. Reduction relationships between the colouring problems are highlighted: edge colouring via line graphs, and face colouring via dual graphs, rigorously connecting the subdisciplines.

Figure 1: Node, edge, and face colourings of the dodecahedral graph with $\chi(G) = \chi'(G) = 3$ and $\chi_f(G) = 4$.

The discussion emphasizes the $\mathcal{NP}$-hardness of chromatic number and chromatic index determination for general graphs, with explicit mention of tractable subclasses and the four colour theorem as a unifying result for planar face colouring.

Algorithms and Empirical Analysis

The paper systematically surveys algorithmic approaches for node colouring, ranging from greedy heuristics to exponential-time exact methods and hybrid metaheuristic strategies. The greedy algorithm's sensitivity to vertex orderings, the efficacy of the DSatur heuristic (maximum saturation degree rule), and recursive backtracking with clique-based pruning are all given formal prominence. DSatur's performance guarantees on bipartite, tree, cycle, and wheel topologies are demonstrated, with empirical considerations on runtime scaling and the frequency of optimal orderings.

Metaheuristic frameworks, such as the hybrid evolutionary algorithm (HEA), incorporating tabu search for clash reduction, recombination operators, and population-based evolutionary cycles, are rigorously detailed. This paradigm enables practical scalability to larger dense random instances.

Figure 2: Comparative performance of HEA and backtracking on $G(n, p)$ random graphs across different densities.

The empirical evaluation substantiates that, on Erdős–Rényi random graphs, backtracking yields optimal chromatic numbers for up to $n \sim 100$ in predictable time, whereas HEA provides upper bounds with robust performance. The linear trend and dependence on graph density are statistically exhibited.

Visualization and Structural Observations

The visualisation methodology encompasses diverse layout strategies (force-directed, circular, multipartite), with commentary on the tradeoffs for dense versus sparse topologies. It is demonstrated that colour class clustering and structural layouts can elucidate independent set partitions and highlight symmetry and cycles.

Figure 3: Multiple visual arrangements of a $G(30, 0.5)$ colouring reflecting $\chi(G) = 7$.

Characteristic classes—bipartite, tripartite, and higher—are illustrated with canonical examples (binary trees, lattices, HoG exemplars), with the relationship between odd cycle presence and chromatic number visually encoded.

Figure 4: Bipartite graph node colourings; all depicted graphs verify $\chi(G) = 2$.

Edge Colouring: Theoretical Properties and Applications

The exposition on edge colouring articulates Vizing's theorem, separating Class 1 ($\chi'(G) = \Delta(G)$) and Class 2 ($\chi'(G) = \Delta(G) + 1$) graphs, and addresses the polynomial-time computability of optimal solutions for bipartite and complete topologies. The mapping to matchings and the significance for sports scheduling and network routing are linked to practical algorithmic design.

Figure 5: Optimal edge colourings of complete and tree graphs, demonstrating matching decompositions.

The encoding of routes via colour sequences in navigation contexts is exemplified through real-world planar graphs, with substantial compression ratios illustrated when replacing high-entropy edge tags with minimal colour encodings.

Figure 6: Edge colouring of a real street map and hexagonal lattice, evidencing $\chi'(G) = \Delta(G)$ in both.

Face Colouring and Topological Constraints

A substantial portion of the paper is dedicated to planar face colouring and the interplay with duality. The four colour theorem is treated in a rigorous topological context, and the special case of Eulerian graphs leading to bipartite duals and resultant $\chi_f(G) = 2$ is both proven and illustrated.

Figure 7: Face colourings of diverse planar graphs and maps, with chromatic number 3 or 4 as dictated by adjacency arrangements.

The connection between structural properties (Eulerian, odd-degree nodes, faces forming cycles) and chromatic requirements is detailed, including classical and aperiodic tiling patterns. The peculiarities of Delaunay triangulations and Voronoi diagrams, with respect to typical face chromatic numbers, are analyzed.

Figure 8: Face colouring contrasting a Voronoi diagram (4 colours) and its Delaunay dual (3 colours), both derived from random point sets.

The paper dispels historical misconceptions such as the spurious five-colouring counterexample, emphasizing that all planar embeddings are four-colourable.

Implications and Future Directions

The technical results surveyed have implications for combinatorial optimization, algorithm design, and topological understanding of networks. Novel visual formalisms aid in pedagogical and algorithmic clarity, potentially inspiring further research into colour-critical graphs, parameterized heuristics, and applications in register allocation, scheduling, and map visualization. The extension to surfaces of higher genus (Heawood's bound) and artistic manifestations in geometric and cubist art are suggested as promising lines of exploratory research.

Conclusion

This work synthesizes foundational and advanced aspects of graph colouring with a strong emphasis on visual representation, empirical algorithmics, and applications. The paper combines theoretical rigor with comprehensive illustration, supporting both further algorithmic research and cross-disciplinary applications in combinatorial optimization, art, and data visualization (2602.18246).