Vectorizing for loop with with strcmp function inside

Question

Avijit Chowdhury 2019-1-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/439283-vectorizing-for-loop-with-with-strcmp-function-inside

评论： Avijit Chowdhury 2019-1-11

Hi,

I have a very large cell array (A), which has 3 columns with 1066426 rows of data.

What I am trying to do is for each row (row ii), find the one and only row in A (say row j) for which A(j,1) is the same as A(ii,2), & A(j,2) is the same as A(ii,1). Then, get A(j,3) and put it into a structure list (dayne(ii).bsg). Below is my code:

Parfor ii =1:length(A)
dayne(ii).bsg=cell2mat(A((strcmp(A(:,2),A(ii,1)) & strcmp(A(:,1),A(ii,2))),3));
end

The code is working, but its taking a really long time (about 30 mins for 1% to be completed). I have already used parfor to utliize parallel workers.

Can anyone suggest if (and how) this code can be vectorized?

Is there any other way to make it work faster?

Thank you very much for the help.

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

Stephen23 2019-1-11

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/439283-vectorizing-for-loop-with-with-strcmp-function-inside#answer_355957

编辑：Stephen23 2019-1-11

在 MATLAB Online 中打开

Reverse the loop order and get rid of the cell2mat:

for ii = numel(A):-1:1
    idx = strcmp(A(:,2),A(ii,1)) & strcmp(A(:,1),A(ii,2));
    dayne(ii).bsg = A{idx,3};
end

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Avijit Chowdhury 2019-1-11

在 MATLAB Online 中打开

Thanks for the help, it seems to be somewhat faster. However, I am not able to use parfor, and since there maybe iterations where the outpull is null (does not meet the criteria), I have had to add in some more commands:

I have timed it and seems to be taking 14sec/100 iterations.

for ii= length(A):-1:1
    idx = strcmp(A(:,2),A(ii,1)) & strcmp(A(:,1),A(ii,2));
    if any(idx)
    dayne(ii).bsg = A{idx,3};
    else
    end     
end