Matlab Busy since four hours

Question

Muhammad Ali Haider 2016-4-1

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/276576-matlab-busy-since-four-hours

评论： Steven Lord 2019-5-24

Here is my code which when run makes the matlab show busy since hours. What can be wrong? My data is big though!

Y = importdata('Dr_Kucukvar_data.xlsx');
linverse = importdata('L_inverse.txt');
X= Y.data.Diagonal;
mlinverse_1 = X(:,1);
mlinverse_1 = diag(mlinverse_1);
mlinverse_2 = X(:,2);
mlinverse_2 = diag(mlinverse_2);
mlinverse_3 = X(:,3);
mlinverse_3 = diag(mlinverse_3);
mlinverse_4 = X(:,4);
mlinverse_4 = diag(mlinverse_4);
mlinverse = [mlinverse_1 mlinverse_2 mlinverse_3 mlinverse_4];
Result_matrix = zeros(7824,1);
for i= 1:4
    u= 7824*(i-1)+1;
    g= u +7823;
    for j= 0:26
        for k= 96:107
            h= 163*j + k;
            z= zeros(7824,1);
            z(h)= 1;
            S = mlinverse(:,u:g) * linverse * z;
            Result_matrix = [Result_matrix S];
        end
    end
end

3 个评论
显示 1更早的评论隐藏 1更早的评论

Geoff Hayes 2016-4-1

在 MATLAB Online 中打开

Muhammad - look at the innermost loop iterating over k

 h= 163*j + k;
 z= zeros(7824,1);
 z(h)= 1;

So on every iteration, you are re-creating z to be a 7824x1 array just to set one element, z(h), to one. Why? Is this intended?

Muhammad Ali Haider 2016-4-1

Thank you so much Geoff. Stafford's code below really worked!

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Roger Stafford 2016-4-1

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/276576-matlab-busy-since-four-hours#answer_215980

在 MATLAB Online 中打开

Your method of appending columns to 'Result_matrix' is very time consuming. Also as Geoff points out, your creation of a new 'z' matrix each time in the inner loop is very inefficient. I suggest you alter your code as follows:

   ......
   Result_matrix = zeros(7824,1297); % 1297 = 4*27*12+1
   c = 1;
   for i= 1:4
     u= 7824*(i-1)+1;
     g= u +7823;
     for j= 0:26
       for k= 96:107
         h= 163*j + k;
         S = mlinverse(:,u:g) * linverse(:,h);
         c = c+1;
         Result_matrix(:,c) = S;
       end
     end
   end

This should give you the same result and be considerably faster.

Final note: Do you really want the first column of Result_matrix to be all zeros?

4 个评论
显示 2更早的评论隐藏 2更早的评论

Stephen23 2019-5-24

"is there any problem?"

Yes: https://www.mathworks.com/help/matlab/matlab_prog/preallocating-arrays.html

Steven Lord 2019-5-24

Preallocation will help, but this is also a problem that can be easily vectorized (unless this is a homework problem where you're required to use a for loop, in which case do preallocate as Stephen suggested.) See the Array Operations section on that page for an example that looks somewhat like the problem you're trying to solve.

I vectorized the code and the computation piece took under a minute on my machine, most of which was probably spent allocating the three 8 GB arrays needed to store n, en, and sol_err. I didn't let the plotting finish, since that's a lot of data to plot. It's probably more data points than you have pixels on your screen, so you might want to plot only every thousandth point or something. Or you could divide your upper limit Nex by a thousand or something. When I used Nex = 1000000 the whole operation was done in a couple seconds.

请先登录，再进行评论。