sum of product of 2 matrices

hi

i have 2 matrices... u is a 2 x 1250 matrix and r is a 1250 x 1250 matrix...

here c = 1

for each j = 1 to 1250 i want to find

how to writ the matlab code for this ?

Answer 1

[Zoltán Csáti]

It is just a standard matrix-matrix product: u*r. So enter

value = u(1,:)*r;

The j-th column of the matrix "value" will contain the sum for the specific j.

[Sreejith]

This is only the product..

i want to find the sum of the products

u(c,1)*r(1,j) + u(c,2)*r(2,j) +.....+u(c,i)*r(i,j) where i = 1,2,3...,1250

what will be the code ?

[Matt J]

But u(1,:)*r is a matrix product. It is equivalent to the summation you've shown.

[Zoltán Csáti]

This is what my code does. Write a few indices and you will see.

[John D'Errico]

+1. What you don't seem to understand is that the * operator IS a dot product, i.e., the sum of products of elements. After all, MATLAB is a matrix language, so it is written to do these computations trivially, with the basic operators.

You sum of products is simply a dot product, what * delivers.

Answer 2

[Guillaume]

Use bsxfun:

sum(bsxfun(@times, u(c, :)', r))

[Sreejith]

In your answer, there is no mentioning of 'j'...

sum(bsxfun(@times, u(c,:)', r(:,j))

Is this correct ?

[Zoltán Csáti]

Guillaume provided a vectorized version, that's why there is no matrix index j.

[Guillaume]

bsxfun replicates u(c,:)' for each column j of r and sum calculates the sum for each of these columns. Exactly what you asked for.

Just try my code without any modification.

Note, this is equivalent to:

tempu = repmat(u(c, :)', 1, size(r, 2));
p = tempu .* r;
result = sum(p)

Or, if you really want to introduce j and make the calculation much slower:

tempu = repmat(u(c, :)', 1, size(r, 2));
p = tempu .* r;
result = zeros(1, size(r, 2));
for j = 1:size(r, 2)
   result(j) = sum(p(:, j));
end

The one line bsxfun does the same.