using find on each column of a matrix independently

Hello, I would like to use the find function on each row independently in a vectorized way.

for example, I want to find the first point that is above a threshold defined by which column I am in, let this be thresh(j) where j is the column.

let M be an m by n matrix, then for loop way of doing this is as follows:

firstoverthresh = zeros(1,n);
for j = 1:n
    firstoverthresh(j) = find(M(:,j)>thresh(j),1);
end

this should return the n sized vector (since there are n columns) of values that I am trying to find.

Thank you :)

Answer 1

[Image Analyst]

There's nothing wrong with that way. It's fine: it works, it's intuitive, and it's fast. I don't know of any non-looping way that will be faster.

[Image Analyst]

To compare Matt's solution and your "for loop" solution (with a constant threshold of 0.5:

n = 10000
M = rand(n,n);
tic
firstoverthresh = zeros(1,n);
for j = 1:n
  firstoverthresh(j) = find(M(:,j)>.5,1);
end
toc

tic
map=bsxfun(@gt,M,0.5);
[~, firstoverthresh]=max(map,[],1);
toc

For a large matrix, a hundred million elements or 10,000 by 10,000, the for loop is faster.

Elapsed time is 0.125944 seconds.
Elapsed time is 0.151905 seconds.

while for a smaller array, like n = 100, Matt's bsxfun solution is faster.

Elapsed time is 0.000433 seconds.
Elapsed time is 0.000138 seconds.

though either one is so fast (less than a millisecond) you wouldn't notice the difference.

Answer 2

[Matt J]

 map=bsxfun(@gt,M,thresh(:).');

 [~, firstoverthresh]=max(map,[],1);

Answer 3

[Matt J]

There is also a MEX accelerated version on the FEX,

http://www.mathworks.com/matlabcentral/fileexchange/24641-vectorized-find-with--first--option