PDC 2008 Conference Scheduling Using a Genetic Algorithm

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Florian Krüsch — Sun, 04 May 2008 14:40:00 GMT

If the genetic algorithm doesn't work out, try simulated annealing. It is very good for this kind of problem :)

Man, I really wish I can make it to PDC. With Live Mesh announced and Silverlight, this is going to be exciting.

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

mswanson — Sun, 04 May 2008 19:10:17 GMT

Thanks for the tip, Florian. Although I didn't mention it in the article, I did experiment with simulated annealing, and it's still part of the project. I'm about to do some testing with a much larger dataset, and my guess is that I'll start to see some differences in the two algorithms soon.

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Florian Krüsch — Mon, 05 May 2008 10:16:31 GMT

Interesting! I'm curious how the differences will turn out in terms of compute time and result.

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

briankel — Mon, 05 May 2008 21:14:35 GMT

Well, that just about wraps up my "Is there anything Mike Swanson can't do?" list. (scratches off "#32 - Applied Genetics"). Only two to go... "#57 - Paragliding" and "#138 - Surfing." I expect these will be on your FY09 commitments.

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Paul Osburn — Tue, 06 May 2008 03:07:33 GMT

Hey Mike:

That sounds like a fun challenge, although I like your use of the word 'most' at the end of the post. Years ago, as my senior project in college, I tried writing a scheduling application (in LISP of all things) that would create the class schedule for the Computer Science department for an upcoming quarter. After working on it for a while, I learned about NP-Complete problems. Man. So, at the end, I found solice in the fact that I was able to automatically schedule *most* of the classes. I'm eager to see how your scheduling turns out!

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

BradA — Tue, 06 May 2008 03:18:36 GMT

I love it! You can always introduce a random number generator in there if it helps... :-)

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Dan Ciruli — Thu, 08 May 2008 23:49:29 GMT

How compute intensive do you expect this to be? We've had several customers use our .NET-based grid computing framework to run genetic algorithms across small and large grids alike. Let me know (at dan (at) digipede (dot) net if you're interested). MSR has a grid they may be willing to share if you need it...

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Krip — Thu, 22 May 2008 09:30:44 GMT

Cool stuff. What does your fitness function look like? Seems to me that if you're factoring into your score each attendee's ability to attend all their favourite sessions, then you need for each possible solution some clever algorithm to see how well that can be done. For example, if one session does occur more than once during the day, then there are multiple options for the attendee (for that solution). Does that mean there are then multiple options for each attendee+solution you must score?

PDC2008: A Day in the Life #5

Mike Swanson's Blog — Thu, 02 Oct 2008 08:43:26 GMT

I have an internal PDC2008 gadget running in my Windows Sidebar, and it reports that there are only 25

PDC2008: A Day in the Life #6

Mike Swanson's Blog — Tue, 07 Oct 2008 07:38:02 GMT

Today was our first day of PDC2008 dry-runs on campus. To present a session at the conference, all speakers

Microsoft PDC 2008 : Cloud Computing, SOA et Windows 7 sont à l’honneur

Cloud Computing @ Microsoft France — Wed, 08 Oct 2008 18:44:35 GMT

La liste des sessions de la conférence Microsoft 2008 pour les développeurs et architectes est au 2/3

re: PDC 2008 Conference Scheduling Using a Genetic Algorithm

Dan H — Tue, 28 Oct 2008 23:26:48 GMT

After 1.5 days at PDC, so far I'd have to say that the chosen algorithm was not a very good one. There have been countless room overflows, many times to a second overflow room, and quite often the overflow room is on the opposite side of the facility. Saw in Hanselman's talk yesterday that he had to remove some slides / content since his talk was scheduled before the Windows 7 keynote.

Was there a snafu in the number of attendees, or was all this expected?