Can I vectorise or make this 'find' double-loop run faster?

Hi, Is there a possible way in which the formula that uses 'find' to update a matrix could do away with the double loops and make this calculation faster. Inputs:

 1) Cmat - a 13x13 matrix,
 2) Hist - a 1000x21 matrix of random numbers,
 3) Chain - a 1000x22 matrix, such that Chain(:,1) contains numbers 1:13 and Chain(:,2:22) are zeros.

I run two loops to update the Chain matrix as follows:.

    for i=2:size(Chain,2);

        for k=1:size(Hist,1)

            Chain(k,i) = find(Hist(k,i)< Cmat(Chain(k,i-1),:),1);

        end

    end

any suggestions are welcome.

Thanks for your inputs.

jayant

Answer 1

[Matt J]

Assuming R2016b or later,

    for i=2:size(Chain,2)

      [~,Chain(:,i)] = max(   Hist(:,i) < Cmat(Chain(:,i-1),:)  ,[],2);

    end

[J Yadav]

That's a marvellous reply. I don't understand the ~ notation, but the solution works and reduced my calculation time by > 90%. Thank you very much.

[Stephen23]

https://www.mathworks.com/help/matlab/matlab_prog/ignore-function-outputs.html