Multiply matrices of different sizes

Background: I've got 7 years worth of Power data for a wind farm binned into 100 bins of wind speed i.e. 100 x 7 matrix. I've got probability of each of that bin happening i.e. 100 x 1 matrix. I would like to multiply the elements of 100x7 matrix by elements in a 100x1 matrix to get cumulative probability of power output.

In simple term: how do I multiply 100x7 matrix by 100x 1 matrix? Tried bsxfun but didn't work.

[Jan]

"Didn't work" is a weak explanation. Better post your code and the complete error message. Not that bsxfun will be the solution, so letting us guess, what fails in your case, is not useful.

[NMans]

I've got Pow_ave which is 100x7 matrix i.e. 7 years of average power data and prob_bins which is probability of wind speed occuring from hiscounts i.e. 100X1 matrix. My codes are as below:

 for yrs = 1:length(years)
            [N,edges,bin] = histcounts(SPEED(:,yrs),100);
            COUNT = [N]';
            nelements = sum(COUNT);
            prob_bins = COUNT./nelements;
Area = bsxfun(@times, Pow_ave(:,yrs), prob_bins)
end

The results should be 100x7 matrix but I've got 100x1 matrix instead.

I could plot Pow_ave vs prob_bins and find area under graph but thought this is a simpler method.

Answer 1

[Jan]

Your code:

for yrs = 1:length(years)
  [N,edges,bin] = histcounts(SPEED(:,yrs),100);
  COUNT         = N.';  % Here [] is a waste of time
  nelements     = sum(COUNT);
  prob_bins     = COUNT ./ nelements;
  Area          = bsxfun(@times, Pow_ave(:,yrs), prob_bins);
end

Area is overwritten in each iteratotion. Maybe you want:

    Area(:, yrs)  = Pow_ave(:,yrs) .* prob_bins;

As far as I can see, Pow_ave(:, yrs) and prob_bins are both 100x1 vectors, so there is no need for bsxfun.

[NMans]

Thank you! This works now - I just have to add (:,yrs) after Area.

[Jan]

@NMans: Pre-allocate theoutput before the loop:

Area = zeros(100, 7);

This avoids the iterative growing of the array, which is expensive in general - although the runtime will not matter much in this case.

Answer 2

[Geoff Hayes]

NMans - if you want to multiply each column of the 100x7 matrix by the 100x1 matrix, then you could use bsxfun as

 X = randi(255,100,7);   % generate some dummy data
 Y = rand(100,1);
 Z = bsxfun(@times,X,Y);

Or are you trying to do something else?

[NMans]

Geoff - thank you I've tried that but it gives me 100X100 instead of 100x7 matrix

[Geoff Hayes]

NMans - when I run the code that I posted in my answer, Z is a 100x7 matrix. Are you sure that your answer is 100x100?

[NMans]

My codes are as below:

for yrs = 1:length(years)
            [N,edges,bin] = histcounts(SPEED(:,yrs),bins);
            COUNT = [N]';
            nelements = sum(COUNT);
            prob_bins = COUNT./nelements;
Area = bsxfun(@times, Pow_ave(:,yrs), prob_bins);
end

And I ended up with 100x1 matrix.