Sum elements over indices in cell array

I have a cell array, in which each cell is a vector of integer numbers. These numbers are certain indices of another vector, say A. For each cell, I want to compute the sum of the elements in A, which are being indexed by my cell. For example, I can easily do this with a for loop (see code below). However, I was wondering if it is possible to vectorize this piece of code.

What I have done now, is artificially edit the cells so that they have the same length, by adding a dummy index. This dummy index then refers to a value 0 in vector A, so that I can easily sum over them by indexing A with the matrix induced by the cell array. However, this takes up a lot of memory.

Any ideas?

A = 1:15;
% some vector containging the values I need

myCell = {[1 3 4], [2 5 8 12]};
% the cell array generally contains arrays of different lengths

sumValues = zeros(1,2);
% I want to fill the array SumValues efficiently

for j = 1:2
   sumValues(j) = sum(A(myCell{j})); 
end

[Rik]

Loops can be surprisingly fast if there isn't a straightforward way to vectorize your code. Did you run any profiling to see if this is actually the bottleneck in your process?

[Lennart Sinjorgo]

Yes, its rather time consuming, although definitely not the bottleneck. That doesn't mean I shouldn't look for ways to speed it up though.

Answer 1

[Matt J]

Once you've split things into cells, there is nothing faster than a loop. The better data organization would be to use group labels instead:

A = 1:15;

I = [1 3 4,   2 5 8 12]; %The indices as a vector
G= [1 1 1,    2 2 2 2]; %Group labels


sumValues=accumarray(G(:),A(I(:)))

sumValues = 2×1

8 27

[Lennart Sinjorgo]

Thank you! I did not know of this function! I'll have a look and see if it's faster than what I'm doing right now.

[Matt J]

You're welcome, but if you conclude that it works well for you, please Accept-click the answer.