Sum subsets of vector by indices stored as vectors in cell array

Question

L is a cell.

L=
 2,4,6   % j=1
 1,6,8   % j=2
 4,6     % j=3

r is a vector 1x8:

23 1 24 5 4 3 7 8

I want to vectorize this code:

UC=zeros(1,J);
for j=1:J
    if ~isempty(L{j})                      
        UC(j)=sum(r(L{j}));
    end
end

I tried this:

UC = arrayfun(@(x)r(x), L, 1, 'UniformOutput', false);

but it looks like cells are not suitable for this function.

Error using subsindex
Function 'subsindex' is not defined for values of class 'cell'.

Answer 1

Listed in this post is an almost vectorized approach based upon accumarray. I am calling it as almost vectorized because it uses cellfun which isn't really a vectorized way, but since it uses it to find the lengths of each cell only, so its effect would be minimal. Here's the implementation -

lens = cellfun('length',L)
id = zeros(1,sum(lens))
id([1 cumsum(lens(1:end-1))+1]) = 1;
out = accumarray(cumsum(id(:)),r([L{:}]))

Answer 2

You just want to sum up elements of r according to every of your cell array elements? Then you need indeed cellfun:

%// given
L = { ...
 [2,4,6]   % j=1
 [1,6,8]   % j=2
 [4,6] }

r = [23 1 24 5 4 3 7 8]

%// output
out = cellfun(@(x) sum(r(x)),L)
%// or in case r is not a vector, but a matrix
out = cellfun(@(x) sum(r(x(:))),L)

which is the same as:

out = arrayfun(@(x) sum(r(x{:})),L)

---

out =

     9
    34
     8