Sunday, July 11, 2010
Research
My research this summer is follow-up on the project I did last semester--Optimization of the Double Row Layout Program using an ant colony algorithm. It is based on a published paper where the author used an exact algorithm to solve the problem. He could actually only "solve" the simplest of problems--that's why I'm proposing the ant colony algorithm to approximate a solution when the problem is more complicated. I've had trouble verifying that my calculations are correct since the data used in the other paper is not available to me. The author randomly generated several matrices from a uniform distribution of numbers between 0 and 50. Obviously if the random numbers are all closer to 0 that will create a very different solution than if the random numbers are all closer to 50. So, for the past two weeks I've been working to learn yet another new computer program--AMPL. Successfully running AMPL on the simple 6-machine problem will duplicate the author's research so that I can verify my calculations. The program is a mathematical modeling language that is supposed to be intuitive. I've spent hours trying to learn the language; I'm going back to campus this morning to work on my intuition.
Subscribe to:
Post Comments (Atom)
If they're truly uniform between 0 and 50 then there shouldn't be any bias toward 0 or 50. Its a real challenge to recreate someone else's "random" data because they rarely do exactly what they say. There's always some little thing they forgot to mention in there. Matlab can easily generate uniform matrices from 0 to 50, if they're truly uniform.
ReplyDelete