This project was conducted at the Institute of Mathematical Modelling, at the Technical University of Denmark, and at UNIC, Danish Computing Center for Research and Education, during the late spring, summer, and autumn of 1998.
The original idea behind this project, came from Claus Bendtsen, of UNIC. In short, he was interested in investigating whether discrete optimization of QR factorization would be beneficial, if at all possible, in relation to another project he was involved in.
During the time of the project, I have had much good advice, and some interesting and most enlightening conversations with Professor Jens Clausen, of IMM, DTU, who supervised this project.
I will present the methods I used for finding ``good'' solutions, as well as some of the approaches that failed. This is by no means a thorough mathematical treatment of the problem. It is a document describing a number of approaches taken, and their usefulness in actual applications.