Problem with for loop

Hallo,

I have a large matrix let's say A 150 * 220000 , including columns that are set entirely to zero .

I want to find '' column wise'' the first element that is larger than 5 and the first element smaller than 200 and store them as two vectors

I creat a for loop, but it breaks when it reaches the first zero column in the matrix. so instead of idx_start 1 * 220000 , I get idx_start 1 * 15000

how can i modify my code so the loop continue over the zero columns?

and would be better if I replaced the zero columns with NaN?

I tried something with isempty but it does not work like I want.

wenn the condition is not met, it is enough to be replaced with NaN

[nx,ny] = size(A) ;
for j = 1:ny  
idx_start(:,j) = find(A(:,j)> 5 ,1,'first') ; 
 if (isempty(A(:,j)))
                continue  
 idx_end(:,j) = find (A(:,j) < 200 , 1, 'first');
  
   end
end

[madhan ravi]

Dwf?

[YH]

I edited it, it was a mistake

[madhan ravi]

Give a short example of your matrix and your desired output

[Jan]

isempty(A(:,j)) is always false, because it checks the number of elements, not the contents of the elements. See any().

Answer 1

[Andrei Bobrov]

s = size(A,2);
[ii,jj] = find(cumsum(cat(3,A > 5,A < 200)) == 1);
out = accumarray([rem(jj-1,s)+1,ceil(jj/s)],ii,[s,2],[],nan);

[YH]

thanks! it is a great code, short, fast and works as I wanted

Answer 2

[Dennis]

The problem is that find might return an empty vector and in that case the assignment fails.

You could catch this error by checking if there are any values > 5.

if any(A(:,j)>5)
idx_start(:,j) = find(A(:,j)> 5 ,1,'first') ; 
else
idx_start(:,j)=0 %or NaN or whatever you want to happen
end

I would prefer the use of only one array to store start and end values. Especially since you are always only storing 1 value you could do it like this:

if any(A(:,j)>5) && any(A(:,j)<200)
 idx(1,j) = find(A(:,j)> 5 ,1,'first'); 
 idx(2,j) = find (A(:,j) < 200 , 1, 'first'); 
else
idx(1,j)=0; %or NaN or whatever you want to happen
idx(2,j)=0; %or NaN or whatever you want to happen
end

[YH]

Thank you for your help!

Answer 3

[Jan]

In for j=1:nx you run a loop over the rows, not columns. Do you mean ny here?

[nx, ny]  = size(A);
idx_start = NaN(1, nx);  % Pre-allocate!!!
idx_end   = NaN(1, nx);  % Pre-allocate!!!
for j = 1:ny   % Or really nx?
   m = find(A(:, j) > 5, 1, 'first');
   if ~isempty(m)
       idx_start(j) = m;
   end
   m = find(A(:, j) < 200, 1, 'first');
   if ~isempty(m)
     idx_end(j) = m;
   end
end

Now all elements of the idx_... vectors are NaN, if the corresponding column of A does not contain matching elements. You can skip the search also, if the column contains zeros only:

for j = 1:ny    % Or really nx?
   col = A(:, j);
   if any(col)  % Skip if column contains zeros only
     m = find(col > 5, 1, 'first');
     if ~isempty(m)
       idx_start(j) = m;
     end
     m = find(col < 200, 1, 'first');
     if ~isempty(m)
       idx_end(j) = m;
     end
   end
end

[YH]

thanks alot! it is really ny not nx