Hofman Graph

Family of graphs used as benchmarks for Hamiltonian path and cycle algorithms From Wikipedia, the free encyclopedia

In the mathematical field of graph theory, the Hofman graph family is an infinite family of finite graphs introduced by the Polish researcher Radosław Hofman (2025) as scalable benchmark instances for the study of Hamiltonian path and Hamiltonian cycle algorithms and related approaches to the Travelling Salesman Problem (TSP).[1]

Verticesvariable (scalable family)
Edgesvariable
Quick facts Vertices, Edges ...
Hofman graph family
Example of non-Hamiltonian Hofman Graph H(12,4)
Verticesvariable (scalable family)
Edgesvariable
Table of graphs and parameters
Close

Hofman graphs are designed to be structurally balanced and uniformly connected while allowing explicit control over the existence of Hamiltonian paths and cycles. In many cases, the construction also determines the minimum number of edges that must be added to obtain a Hamiltonian cycle.

The family was motivated by the need for large graph instances with known Hamiltonicity properties that avoid structural bottlenecks or easily detectable decompositions, which may bias the evaluation of algorithms.

Definition

Hofman graphs are parameterized by two integers and are typically denoted

where:

  • denotes the number of vertices in the outer structure of an associated outline graph,
  • denotes the step size (distance) controlling how outer vertices connect to inner components.

The construction is inspired by generalized Petersen graphs, but differs in that the inner structure may consist of multiple disconnected components and that outer–inner connections are mediated by special connector subgraphs called H-connections.

Outline graphs

The first stage of the construction produces smaller graphs called outline Hofman graphs, denoted

Outline graphs consist of:

  • an outer cycle-like structure,
  • one or more inner components (often odd cycles such as triangles),
  • connector vertices between outer and inner parts.

Instead of connecting the outer and inner parts directly, Hofman introduced a connector vertex that forces any Hamiltonian traversal to alternate between outer and inner regions. This alternation constraint can prevent the existence of Hamiltonian paths or cycles depending on the parity and number of inner components.

Examples of outline Hofman graphs:

More information Graph image, Description ...
Graph imageDescription
Outline of Hofman Graph H(6,2) as an example of a graph containing Hamiltonian path
Hamiltonian path follows vertices: 17, 11, 5, 0, 6, 12, 14, 8, 2, 1, 7, 13, 15, 9, 3, 4, 10, 16
Outline of Hofman Graph H(9,3) as an example of a graph not containing Hamiltonian path
After third triangle is added to the inner structure the Hamiltonian path requires at leas one edge to be added to the graph
Outline of Hofman Graph H(12,3) as an example of a graph containing Hamiltonian cycle
The inner structure consists of three squares, therefore a Hamiltonian cycle may be constructed
Outline of Hofman Graph H(12,5) as an example of a graph containing Hamiltonian path
In this case the inner structure has a form of a star, it is possible to construct a Hamiltonian path connecting all vertices but not Hamiltonian cycle
Outline of Hofman Graph H(12,4) as an example of a graph not containing Hamiltonian path
The inner structure consists of four triangles, therefore the graph requires at least two edges to be added to construct a Hamiltonian path
Close

Balanced H-connections

To obtain large benchmark graphs with uniform degree and connectivity, Hofman introduced a balanced H-connection replacing the simple connector.

In the balanced construction, each H-connection is replaced by a fixed 15-vertex subgraph that preserves the Hamiltonian alternation constraint while ensuring that the resulting graph is regular. In the resulting graphs, every vertex has degree 4.

The comparison between H-connectors of the outline graph and full graph:

More information H-connection for outline graph, H-connection for full graph ...
H-connection for outline graphH-connection for full graph
oH-connection
H-connection
Close

The connector has the same properties from the perspective of a Hamilton path construction - if the flow does not alternate between the inner and outer structure, the middle vertice will be impossible to connect.

H-connection example flow

When the balanced connector is used throughout the outline, the resulting Hofman graphs are typically:

  • 4-regular (all vertices have degree 4),
  • 4-connected (in the intended construction),
  • scalable to arbitrarily large size.

For example, the graph contains vertices.

Hamiltonian properties

A central purpose of the Hofman graph family is that Hamiltonian properties are controlled by construction. Depending on the parameters and the chosen inner components, a Hofman graph may:

  • contain a Hamiltonian cycle,
  • contain a Hamiltonian path but no Hamiltonian cycle,
  • contain neither a Hamiltonian path spanning all vertices nor a Hamiltonian cycle.

In addition, for many instances the construction yields the minimum number of edges required to be added to obtain Hamiltonicity.

For example:

  • contains a Hamiltonian cycle,
  • contains a Hamiltonian path but no Hamiltonian cycle,
  • contains neither a Hamiltonian path spanning all vertices nor a Hamiltonian cycle.

Scaling and use in TSP benchmarks

Hofman introduced a scaling method for constructing large TSP benchmark instances from Hofman graphs by embedding them into a weighted complete graph.

The method adds two external vertices:

  • a designated start vertex, and
  • a designated end vertex,

and assigns low cost to edges inside the Hofman subgraph and very high cost to all other edges. This forces optimal tours to traverse the Hofman subgraph in a prescribed manner. Because the number of required "expensive" edges can be derived from the Hamiltonicity properties of the underlying Hofman graph, the optimal tour value can be computed analytically for large instances.[1]

This technique can be extended by concatenating multiple Hofman graphs into layered structures, producing instances with hundreds or thousands of vertices while preserving known optimal costs.

Counterexample to LP-based TSP formulations

The Hofman graph family was introduced as part of a counterexample to a proposed polynomial-size linear programming formulation for the TSP by Diaby et al.

Diaby's approach attempted to model valid TSP tours using flow-like variables and constraints. Hofman constructed a Hofman-based benchmark instance for which the true optimal TSP cost is known, while the LP relaxation yields a strictly smaller objective value. This demonstrates the existence of feasible LP solutions that do not correspond to valid tours ("phantom solutions").[1]

This use case supports broader concerns that polynomial-size linear programs cannot fully characterize the TSP polytope without additional proof, consistent with earlier arguments in the literature.[2]

Examples

The following table lists example instances described by Hofman.

More information , ...
Selected Hofman graphs and their Hamiltonian properties
Graph Outline vertices Total vertices Hamiltonian cycle Hamiltonian path
18 90 No (1 edge needed) Yes
27 135 No (2 edges needed) No (1 edge needed)
36 180 No (3 edges needed) No (2 edges needed)
24 150 Yes Yes
36 180 Yes Yes
36 180 Yes Yes
24 120 No (1 edge needed) Yes
36 180 No (1 edge needed) Yes
45 225 No (4 edges needed) No (3 edges needed)
54 270 No (5 edges needed) No (4 edges needed)
Close

Relationship to Petersen-type constructions

The Hofman graph family is inspired by generalized Petersen graphs such as the classical Petersen graph, but differs in several ways:

  • The inner structure may consist of multiple disconnected components.
  • Outer–inner links are mediated by H-connections rather than direct edges.
  • Hamiltonian properties are controlled explicitly rather than being incidental.
  • Large balanced instances are obtained by systematic substitution of connectors.

Unlike the Petersen graph, which is a well-known hypohamiltonian graph, Hofman graphs can be constructed to be Hamiltonian, non-Hamiltonian, or to lack Hamiltonian paths.

See also

Notes

Related Articles

Wikiwand AI