Vectorize nested loop to increase efficiency

Hi, I have a 3 for loops and I would like if possible to vectorize the two inner loops.

for t=1:size(datesdaily1)
for i=1:size(secids,1)
    sum=0;
    if inc(t,i)==1 
    for j=1:size(secids,1)
        if inc(t,j)==1 
            sum=sum+weig1(t,j)*sqrt(Rates(t,j))*rhoneutral(i,j);         
        end
    end
    b(t,i)=sqrt(Rates(t,i))*sum/MRates(t,1);
    end
end    
end

Any idea on how to accomplish that? Here 'weig', 'inc' and 'Rates' are (size(datesdaily1) by size(secids,1)) matrixes, rhoneutral is a (size(secids,1) by size(secids,1)) matrix.

Thanks

[Chad Greene]

A few notes:

1. Make a habit of not using i or j as variables. They are the built-in imaginary unit. Overwriting them works, but can occasionally errors that are tough to debug.

2. Don't use sum as the name of a variable for a similar reason. It's a built-in function.

At the top of your example can you create some dummy variables that we can use to run your code? For example, this could be accomplished by

Rates = rand(10,30);

if Rates is a 10x30 matrix.

What's the goal of the code above? Can you describe the calculation you're trying to perform?

Answer 1

[James Tursa]

I am guessing a bit on some of the dimensions, but try this as a vectorized alternative. Speed improvement is going to depend on the actual sizes involved and how many of those inc values are 1, but I am getting 1000x or so improvement on my machine with large sizes and a large percentage of 1's in inc. For a small percentage of 1's in inc the speed improvement will be somewhat less than that.

sr = (inc == 1) .* sqrt(Rates);
wr = weig1 .* sr;
mr = bsxfun(@rdivide,sr,MRates(:,1));
wrt = wr * rhoneutral.';
b = mr .* wrt;