Comments on: Big Sunday Push Continues http://blogs.mathworks.com/contest/2007/05/13/big-sunday-push-continues/ The MATLAB Programming Contest is a semi-annual competition where contestants submit MATLAB code to try to solve a challenge. For more information, see http://www.mathworks.com/contest/overview.html Sat, 21 Jan 2012 14:45:27 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: Yi Cao http://blogs.mathworks.com/contest/2007/05/13/big-sunday-push-continues/#comment-2104 Yi Cao Tue, 15 May 2007 12:43:59 +0000 http://blogs.mathworks.com/contest/?p=67#comment-2104 The minimum result has been pushed to 38681. I think it still has some space to go if we can make the code faster. The minimum result has been pushed to 38681. I think it still has some space to go if we can make the code faster.

]]> By: Alan Chalker http://blogs.mathworks.com/contest/2007/05/13/big-sunday-push-continues/#comment-2099 Alan Chalker Mon, 14 May 2007 15:45:35 +0000 http://blogs.mathworks.com/contest/?p=67#comment-2099 I forgot to include my analysis of the current tradeoff between time and results. Right now the leading entry has a results of 39582 and time of 48.7. We are way past the 'knee' of the exponential curve, which is somewhere in the 25-30 seconds range depending on how you look at it. Each 1/2 in time that is shaved off the code right now is equal to about 6 points of results. While we don't know what the makeup of the test suite is, the best result obtained so far is 39028 by David Jones (who is as usual making significant and numerous contributions to the contest). I think we can assume that the minimum possible result is somewhere around 39000. I forgot to include my analysis of the current tradeoff between time and results. Right now the leading entry has a results of 39582 and time of 48.7. We are way past the ‘knee’ of the exponential curve, which is somewhere in the 25-30 seconds range depending on how you look at it.

Each 1/2 in time that is shaved off the code right now is equal to about 6 points of results.

While we don’t know what the makeup of the test suite is, the best result obtained so far is 39028 by David Jones (who is as usual making significant and numerous contributions to the contest). I think we can assume that the minimum possible result is somewhere around 39000.

]]> By: Alan Chalker http://blogs.mathworks.com/contest/2007/05/13/big-sunday-push-continues/#comment-2098 Alan Chalker Mon, 14 May 2007 15:32:21 +0000 http://blogs.mathworks.com/contest/?p=67#comment-2098 I don't think the complexity factor has significantly helped the competition due to the fact you set the bar too high. I've calculated out the scoring equation coefficients which are: k1 = 0.1 k2 = 2 k3 = 0.05 k4 = 1 (Kudos for changing them significantly from past competitions). Based upon this, each point of complexity above 10 only affects the score by 1 point. While most of the leading entries are hovering around complexity 10, I've seen competitive entries as high as 20 something and as low as 6 or 7. People don't seem to be worrying too much about complexity and are just tweaking the parameters and timing now. I'd suggest in the future you lower the complexity 'freebie' bar to 4 or 5 and make the penalty factor significantly greater (i.e. k4=10 perhaps?? or make it exponential). Perhaps a new midterm contest would be the best scoring code with a complexity of 3 or less??? That would seem to really push us based upon the entries so far and would be in line with what the 1000 character challenge calls for (which is typically a ~3x reduction in code length). I don’t think the complexity factor has significantly helped the competition due to the fact you set the bar too high. I’ve calculated out the scoring equation coefficients which are:
k1 = 0.1
k2 = 2
k3 = 0.05
k4 = 1
(Kudos for changing them significantly from past competitions). Based upon this, each point of complexity above 10 only affects the score by 1 point. While most of the leading entries are hovering around complexity 10, I’ve seen competitive entries as high as 20 something and as low as 6 or 7. People don’t seem to be worrying too much about complexity and are just tweaking the parameters and timing now.

I’d suggest in the future you lower the complexity ‘freebie’ bar to 4 or 5 and make the penalty factor significantly greater (i.e. k4=10 perhaps?? or make it exponential).

Perhaps a new midterm contest would be the best scoring code with a complexity of 3 or less??? That would seem to really push us based upon the entries so far and would be in line with what the 1000 character challenge calls for (which is typically a ~3x reduction in code length).

]]>