Sparse matrix multiplication in MATLAB with spfun

Question

I have a dense column matrix y of size (m,1) and a sparse matrix x of size (m,n).
I want to do element-wise multiplication using y and every column of x.
The resultant sparse matrix is still of size (m,n).
Sparse matrix x, when loaded into memory, is about 10GB.
Can spfun help me accomplish my goal in a memory efficient manner?

I am having difficulties understanding the logic behind it.

Thank you.

Shai · Accepted Answer · 2016-01-06T08:26:19.947

5

Have you tried bsxfun?

out = bsxfun( @times, x, y );

spfun is more suitable for element-wise operations where you manipulate each non-zero element of x. It is not exactly fit for matrix-vector element wise operations.
However, if you want to do something along this line, you might try:

[ii jj xij] = find(x); %// extract non-zeros of x and their locations
out = sparse( ii, jj, xij.*y(ii), size(x,1), size(x,2) );

See doc find for more information.

edited Jan 06 '16 at 08:26

answered Jan 06 '16 at 07:39

Shai

111,146
38
238
371

1

Cool, I never knew multiplying a normal matrix with a sparse matrix was possible and that it results in a sparse as well – Adriaan Jan 06 '16 at 07:43
1

Actually, I wasn't aware that bsxfun is applicable to sparse matrices... . – Jan 06 '16 at 08:02
2

@yi416 `bsxfun` rocks!! – Shai Jan 06 '16 at 08:08

Sparse matrix multiplication in MATLAB with spfun

1 Answers1

Linked