Abstract
|
This presentation will give a broad overview of some of the recent enhancements of interior-point algorithms for the improved solution of linear/semidefinite relaxations in combinatorial optimization and binary quadratic programming. Our central topics are planned to cover general interior-point cutting-plane schemes, the efficient handling of free variables and large numbers of linear inequalities, and several warm-starting strategies. Considering applications from graph theory and combinatorics, facility layout design, and molecular biology, this talk will conclude with a brief discussion of selected computational results and summarize open questions for further research.
|