Scalable Effective Approaches for Quadratic Assignment Problems Based on Conic Optimiation and Applications

Project: Research project

Description

Hans Mittelmann will do the bulk of the computations for this proposal. This includes implementation of the various lower bounding techniques and fine-tuning in exchange with investigator Jiming Peng. Due to the variety of approaches suggested this will be ongoing over the entire duration of the grant and much of it will be done with existing software. A more substantial part of his work will be the implementation of the exact solution methods. This will require programming from scratch of a complex branch&bound&cut method, ultimately resulting in software that will be made available. This work will also extend over the entire grant period with a B&B approach for the basic lower-bounding techniques being done in the first year and extensions to other models and more powerful branching rules in the second year. In the third year based on earlier experiments with cuts based on lifting techniques the most promising of these cuts will be added. It is expected that improved bounds over current knowledge will result from this work early on, while in the second half exact solutions for some unsolved cases will be obtained. The computations will be done on state-of-the-art computational platforms utilizing both distributed and parallel computing.
StatusFinished
Effective start/end date3/1/0911/30/11

Funding

  • DOD-USAF-AFRL: Air Force Office of Scientific Research (AFOSR): $175,571.00

Fingerprint

Distributed computer systems
Parallel processing systems
Tuning
Experiments