Formulations of clique problems¶
Part 2 · Formulations and their properties · Chapter 12 · Lecture notes, volume 2 (PDF) · Slides · Exercises: Clique problems
Prerequisites
-
From this module, Chapter 5 «Clique problems»: the problem, its assumptions, cliques, stable sets, the complement graph and \(\omega(G)=\alpha(\overline{G})\).
-
From this module, Chapter 4 «Coloring problems» and Chapter 11 «Formulations of coloring problems»: the chromatic number \(\chi(G)\) and an ILP model with its LP relaxations.
-
From the module Integer linear programming, Chapter 2 «Primal and dual bounds», Chapter 3 «Ideal formulations» and Chapter 4 «Valid inequalities»: LP relaxations, stronger formulations and valid inequalities.
-
From the module Linear Programming, Chapter 9 «Duality theory»: the dual of an LP problem, weak and strong duality.
Objectives
At the end of this chapter you will be able to:
-
write the edge formulation of the Maximum Clique Problem on the non-edges of \(G\), that is, on the edges of \(\overline{G}\);
-
explain why its LP relaxation is weak: the point \(x_v=\tfrac{1}{2}\) is always feasible;
-
write the clique formulation and compare the two LP bounds on a small graph;
-
prove \(\omega(G)\le z(\mathrm{LP}^{\mathrm{CL}}_{\mathrm{MCP}})\le\chi(G)\) through LP duality and a coloring formulation;
-
write the edge and the clique formulations of the Maximum Stable Set Problem;
-
write the ILP models of the vertex-weighted and of the edge-weighted maximum clique problems, linearizing the product \(x_u\,x_v\) with the variables \(y_{uv}\).
12.1 Overview of the family¶
-
The ILP models of the clique problems use one binary selection variable \(x_v\) for each vertex \(v \in V\), equal to \(1\) if the vertex is in the clique.
-
Two vertices that are not adjacent cannot be both selected: the conflict constraints are written on the non-edges of \(G\), that is, on the edges of the complement graph \(\overline{G}\), or on its cliques.
-
The objective function maximizes the number of selected vertices or their total weight; the edge-weighted problem adds one variable per edge.
The clique models select vertices with binary variables and forbid the selection of two non-adjacent vertices. The same models, written on the edges of \(G\) instead of \(\overline{G}\), select stable sets.
- The table lists the members of the family, with the names of their ILP models and the sections that write them.
| Problem | Models | Conflict constraints | Section |
|---|---|---|---|
| MCP | \(\mathrm{ILP}_{\mathrm{MCP}}\) (edge formulation) | on the non-edges of \(G\) | 12.2 |
| MCP | \(\mathrm{ILP}^{\mathrm{CL}}_{\mathrm{MCP}}\) (clique formulation) | on the maximal stable sets of \(G\) | 12.5 |
| MSSP | \(\mathrm{ILP}_{\mathrm{MSSP}}\), \(\mathrm{ILP}^{\mathrm{CL}}_{\mathrm{MSSP}}\) | on the edges and the cliques of \(G\) | 12.7 |
| VWMCP | \(\mathrm{ILP}^{\mathrm{VW}}_{\mathrm{MCP}}\) | on the non-edges of \(G\) | 12.8 |
| EWMCP | \(\mathrm{ILP}^{\mathrm{EW}}_{\mathrm{MCP}}\) | on the non-edges of \(G\) | 12.8 |
-
Overview
Overview of the family
-
12.2 An ILP model for the Maximum Clique Problem: the edge formulation
A second Linear Programming relaxation · Clique inequalities: a stronger formulation · The relation with vertex coloring · Stable sets on the complement graph
-
12.8 ILP models for the weighted variants
The edge-weighted maximum clique problem · Typical mistakes · In summary
Next: 12.2 An ILP model for the Maximum Clique Problem: the edge formulation
Exercises on this chapter: Clique problems