Fun Coursera class - discrete optimization

[gkm]

I thought others might be interested. I have just started this Coursera class on discrete optimization:
https://class.coursera.org/optimization-002
and it is a lot of fun. Very well presented.

Come join to learn some crazy techniques to make programs work out hard problems much faster.
And you can use any programming language.

 

[Waansin]

gkm: how is it going for you?

The biggest hassle I'm having is finding the time to watch the videos and to create solutions to the assignments. I managed, after a lot of time and effort, to get 7/10 for all of the Graph Colouring assignments and I'll never find the extra time to go back to improve my result.

 

[gkm]

I am really enjoying it. I downloaded all the videos, so that I could watch them offline. I have written programs for all the problems and managed to get 7's or better for everything except the last facility location problem where I am unable to get better than 3 so far. I have refrained from using third party solvers, even through that is allowed by the course, since I enjoy the coding of the solvers myself.

For the graph coloring I initially also got 7's but was able to improve to 10's with local search. My local search was done by setting an infeasible solution (but as close to feasible as possible) that have the target number of colors that would get a 10, then LS'ing until the solution is feasible.

Currently I am back to trying to get 10's for the last couple of TravelingSalesPerson problems, but that is proving to be hard going, so not sure how far I shall get with this. However, it has been great fun to "solve" something that has been studied as a hard problem for many years.

(Note, if anybody else is interested in doing this, there is still plenty time left on the course. I think this has significantly expanded my programming skills.)