How to quickly accumulate values into sparse matrix from very large row,col and value arrays containing duplicates indices?

Hi,

I'd like to fill a large (sparse) matrix. I have three very long arrays of rowidx, colidx and values. row col combination can appear several times in those arrays and I'd like to add the corresponding values into the matrix at that location

% alloc sparse zero matrix
A = spalloc(vnum,vnum,vnum*maxdeg);
% I and J are row and col idxs
% val is the vector of values (all same length)
for i = 1:length(I)
    A(I(i),J(i)) = A(I(i),J(i)) + val(i);
end

Takes forever. Is there a faster way? I tried also this

Idx = I+vnum*(J-1);
v = accumarray(Idx,val);

to sum up the duplicates first, but that gives me out of memory (maybe it creates a dense large matrix internally?).

Any help appreciated

Answer 1

[Martin]

Just found the answer:

A=sparse(I,J,val);

does the trick (it actually accumulates at duplicated index values).