I am a sucker for competitions. The people at PATAT (a conference series Practice and Theory of Automated Timetabling; I was co-chair for their 2004 conference, and a member of their steering committee for a few years) do a really good job at timetabling competitions. I really liked their competition for the 2008 conference on course and exam timetabling. I had some thoughts of using logical Benders’ to solve one of the problems but didn’t quite get my act together for it. But I liked the problems and I liked the organization.
The next competition has just begin, though it is on an accelerated schedule. The topic is nurse rostering and ends June 1, 2010. This is an important practical problem:
Building good rosters for nurses in hospitals has received ample attention in recent years and is recognised as a difficult combinatorial optimisation problem with practical relevance. In hospitals much effort is spent producing solutions which are workable and of a high quality.
It is a nicely designed competition, but be aware that you can’t use commercial codes. So those of us who use CPLEX/XPRESS/Gurobi as a linear or integer programming base are probably out of luck:
Rule 5
Participants have to implement an algorithm to tackle the problem on a single processor machine; they can use any programming language. The use of third-party software is allowed under the following restrictions:
- it is free software
- it’s behaviour is (reasonably) documented
- it runs under a commonly-used operating system (Linux or Windows)
Working with COIN-OR is probably legal, under a broad definition of “reasonably documented”. But it is interesting that Mac operating systems are not deemed “commonly-used”.
Despite these caveats, I do recommend the competition, and perhaps will see you in Belfast!
It’s a pity that the CPLEX/XPRESS/Gurobi guys don’t do an implementation themselves. I would be nice to see if they match up to cpsolver-unitime (1th in ITC2007 track 1&3, Thomas Muller) and my Drools Planner (4th in ITC2007 track 1).
I don’t recall that ITC2007 had that commercial software restriction and none of the finalists seem to use those implementation.
In any case, I’d love to throw the gauntlet at them and hope they do an implementation anyway 🙂 Of course, if they do, they should speak up now. I might have a head start of 2 weeks, but I only work during the weekend, so I am sure they can catch up (see http://fisheye.jboss.org/browse/JBossRules/trunk/drools-planner ).
Hello Mike,
thanks for the compliments about the competitions and for the advertisement for this new one!
You are right: MAC is a “commonly-used” operating system. This will be corrected soon, allowing participants to use MAC.
Regarding CPLEX and other commercial software, if the solver runs also under a free version (that we organizers can install on our machines) it is ok.
Andrea