Anonymous user
Talk:Closest-pair problem: Difference between revisions
→About this task: mark pseudocode as "in need of review"? :D
(ok for implementors; pseudocode don't need to be optimized) |
(→About this task: mark pseudocode as "in need of review"? :D) |
||
Line 7:
I am working on it; actually I have less time than before. The pseudocode needs to be cleaned up and corrected; the Smalltalk impl works fine in all my tests, but I've done a C and Perl impl too... and they work 1/6 times... meaning that once a while the result between the "brute force" method and "fast" method disagree; at the very beginning I thought the fault was in the C and Perl impls, since Smalltalk worked always... but now I think it is indeed a deeper problem and the error does not exhibit in Smalltalk because of the distribution of the random number (this is just an idea... I am trying to figure out a set of tests to understand it better what's going wrong and why, and why not in Smalltalk... using the same set of points could be enough, it's in my plan). Once C and Perl code will work, I will fix the pseudocode and the Smalltalk impl. Of course, if anyone can fix the pseudocode (it is deduced by the Smalltalk impl indeed!)... or doing a new impl finds it work properly (this would be my C and Perl impl are wrong someway... but I can't see where!)... I would like to read any idea on it. (First I am going to use the same set of points, then I will take paper and pencil and try to reinvent the wheel in the hope I understand where I get the algorithm explanation wrong!) --[[User:ShinTakezou|ShinTakezou]] 09:17, 11 May 2009 (UTC)
'''A note''': should be the pseudocode marked as "problematic" (or whatever)? (References are there even to allow any good guy to make it better; sorry for this, but sadly as said before the Smalltalk impl does not exhibit a wrong behaviour, so I was deceived by it :D) --[[User:ShinTakezou|ShinTakezou]] 09:23, 11 May 2009 (UTC)
| |||