ICFP 2012Posted by Dan Gerhardsson Tue, September 11, 2012 18:45:52
The results from the final round have been announced. We did pretty well on the last maps, continued to climb, and ended at position 17, which is the team's best score in any ICFP contest so far. We haven't yet analyzed our performance on the maps of the final round, but our guess at this point is that the solution scaled well as the size of the maps increased, and that the ability to handle large maps was crucial in the final round.
In the lightning division we ended up at position 21. This is also our best lightning performance to this date.
ICFP 2012Posted by Dan Gerhardsson Sat, August 18, 2012 11:30:38
The results from the 2nd round were published a while ago. Our code performed pretty well on the new set of maps, and in the main competition we climbed from position 102 (changed from 97 after the score review) up to position 46. In the lightning division we are now placed as number 14, i.e. we dropped a single position from the 1st round. This means that we made it to the final round in both the main competition and the lightning division. Now we are looking forward to the final results which will be presented at ICFP in September.
ICFP 2012Posted by Dan Gerhardsson Wed, August 01, 2012 19:47:27
This time we actually managed to do well in the lightning division, which we usually don't care that much about. We seem to have pretty good scores on all five maps and reached position 13 after the first round. In the main competition the result is a bit less impressive. After the first round we are only at position 97, but at least that is in the top 50% so we made the cut and have advanced to the next round.
ICFP 2012Posted by Dan Gerhardsson Mon, July 16, 2012 14:44:21
Since the contest is over we have now made our repo public:
ICFP 2012Posted by Pär Bohrarper Mon, July 16, 2012 13:39:02
We have now made our submissions.
Improvements from yesterday:
* Can move rocks if needed in some cases
* Avoid blocking the lift with falling rocks
* Handle beards by shaving them if in the way
* Handle lambda-rocks in simple cases
ICFP 2012Posted by Pär Bohrarper Sun, July 15, 2012 22:35:35
Some progress today:
* The solver can now remove a piece of dirt to be able to move a rock in some cases.
* It can make some use of trampolines.
* It has a panic mode that tries to fix things when it doesn't know what to do.
Still to do:
* Remove rocks in general, when blocking other stuff.
* Stop stuff from falling and blocking other stuff.
* Handle beards.
* Handle lambda-rocks.
* Deciding in what order to do stuff.
ICFP 2012Posted by Pär Bohrarper Sun, July 15, 2012 10:41:39
The viewer can now show all the new refinements. Note the tiny text indicators on trampolines. :)
ICFP 2012Posted by Dan Gerhardsson Sat, July 14, 2012 15:00:50
We managed to do a submission for the lightning round with a couple of minutes to spare. We didn't do much about the flooding and we have issues with some of the test maps, but we made several improvements the last hour and we get decent scores on at least a few maps. It feels like a pretty good start.