Question about problem F in ACM ICPC 2017.
Difference between en1 and en2, changed 0 character(s)
### Question about Problem F: Posterize↵
My question is, can we solve [this problem](https://icpc.kattis.com/problems/posterize) using algorithms like k-means or something? After all, all we're trying to do is minimize the error for each pixel with respect to the nearest cluster/value out of the **K** values, right? this idea hit me first before the straightforward dynamic programming solution which passes nicely here [link](https://icpc.kattis.com/submissions/1883959), but when I tried this solution [link](https://icpc.kattis.com/submissions/1883959) I got WA on test #9 and I have no idea if the error is from my implementation or that the algorithm doesn't even fit here, in which case I'd very much like to know explicitly why?↵

Thanks in advance.

History

 
 
 
 
Revisions
 
 
  Rev. Lang. By When Δ Comment
en9 English Hossam 2017-12-02 23:03:01 0 (published)
en8 English Hossam 2017-12-02 23:02:17 23 Tiny change: 'ut of the **K** values`, ' -> 'ut of the K values`, ' (saved to drafts)
en7 English Hossam 2017-09-15 19:01:38 70
en6 English Hossam 2017-06-24 14:11:03 41 Tiny change: '### Question about Problem F: Posterize\nMy questio' -> 'My questio'
en5 English Hossam 2017-06-04 15:55:45 6 Tiny change: 'ssions/1883959) I got WA' -> 'ssions/1882552) I got WA'
en4 English Hossam 2017-06-04 13:36:56 0 (published)
en3 English Hossam 2017-06-04 13:35:31 0 (saved to drafts)
en2 English Hossam 2017-06-04 13:31:28 0 (published)
en1 English Hossam 2017-06-04 13:29:01 801 Initial revision (saved to drafts)