Semidefinite Programming for Combinatorial Optimization
Christoph Helmberg
Abstract
Christoph Helmberg
Abstract
This book offers a self-contained introduction to the field of semidefinite programming, its applications in combinatorial optimization, and its computational methods. We equip the reader with the basic results from linear algebra on positive semidefinite matrices and the cone spanned by them. Starting from linear programming, we introduce semidefinite programs and discuss the associated duality theory. We then turn to semidefinite relaxations of combinatorial optimization and illustrate their interrelation. In the second half we deal with computational methods for solving semidefinite programs. First, the interior point approach, its iteration complexity, and implementational issues are discussed. Next, we explain in great detail the spectral bundle method, which is particularly suited for large scale semidefinite programming. One of the most successful techniques in integer linear programming is the cutting plane approach which improves an initial relaxation by adding violated inequalities. We explore possibilities to combine the two solution methods with the cutting plane approach in order to strengthen semidefinite relaxations of combinatorial optimization problems.
OpenAlex reports 147 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This book offers a self-contained introduction to the field of semidefinite programming, its applications in combinatorial optimization, and its computational methods. We equip the reader with the basic results from linear algebra on positive semidefinite matrices and the cone spanned by them. Starting from linear programming, we introduce semidefinite programs and discuss the associated duality theory. We then turn to semidefinite relaxations of combinatorial optimization and illustrate their interrelation. In the second half we deal with computational methods for solving semidefinite programs. First, the interior point approach, its iteration complexity, and implementational issues are discussed. Next, we explain in great detail the spectral bundle method, which is particularly suited for large scale semidefinite programming. One of the most successful techniques in integer linear programming is the cutting plane approach which improves an initial relaxation by adding violated inequalities. We explore possibilities to combine the two solution methods with the cutting plane approach in order to strengthen semidefinite relaxations of combinatorial optimization problems.
Key concepts: Semidefinite programming, Semidefinite embedding, Cutting-plane method, Linear programming, Second-order cone programming, Relaxation (psychology), Mathematics, Interior point method