why setdiff have time longer than for ?

Question

huda nawaf 2013-5-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/75476-why-setdiff-have-time-longer-than-for

采纳的回答： Walter Roberson

在 MATLAB Online 中打开

*hi,

Itried to make my code be faster , so i used setdiff instead of a part in my code

I used:*

                  v=(1:18);
                   buf=setdiff(v,unique_x);
                  relevant_item(ttt(i),1:length(buf))=buf;

*instead of *

                  for v=1:18
                    b2=find(unique_x==v);
                   if isempty(b2)
                       relevant_item(ttt(i),h)=v;
                       h=h+1;
                   end
               end

*why the running time of this part is less than of setdiff?

thanks*

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2013-5-12

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/75476-why-setdiff-have-time-longer-than-for#answer_85145

在 MATLAB Online 中打开

setdiff() essentially has to sort each time -- or at least to check if the set is ordered.

Are the unique_x restricted to being in the allowed range for v, 1:18 ? If so, then how about

buf = 1:10;
buf(unique_x) = [];
relevant_item(ttt(i),1:length(buf))=buf;

3 个评论
显示 1更早的评论隐藏 1更早的评论

huda nawaf 2013-5-15

Walter, if buf =[ 3 49 10 4 7 20];

I mean not 1:18, can find way to run the above code neither use setdiff , nor for?

thanks

Jan 2013-5-15

@huda: Look into the code of setdiff. You find the relevant part in the fast MEX function ismembc2, which expects pre-sorted data. So sort your data explicitly, call this function and use the replied indices. Note that ismembc2 is not documented, but you find enough information about it in the net.

请先登录，再进行评论。