Sparse matrices and memory
5 views (last 30 days)
Show older comments
Why does this line consume several GB of my RAM
a = sparse(1,1,1,2^30,2^30, 1);
and this line results in out of memory error?
a = sparse(1,1,1,2^34,2^34, 1);
I am specifically telling matlab that I wish to store exactly one nonzero value, how can there be problem with memory? What can I do to be able to use large matrices containing very little data (<1% of RAM)? Isn't this the whole point of sparse matrices and if so, why are they so poorly implemented? Matlab 2017a. Thanks for any suggestions.
1 Comment
Answers (2)
Jan
on 4 Oct 2017
Edited: Jan
on 4 Oct 2017
maxsize, the maximum number of elements allowed in an array with this
version of MATLAB.
In R2017a it is 2^48 - 1. You cannot address more elements in an array. You sparse matrix does have 1 element only, but it could be:
a = sparse(2^30, 2^30, 1, 2^30, 2^30, 1);
also, and this cannot be addressed.
The error messages are different:
a = sparse(1,1,1,2^30,2^30, 1);
Out of memory. Type HELP MEMORY for your options.
b = sparse(1,1,1,1,2^60, 1);
Sparse matrix sizes must be non-negative integers less than
MAXSIZE as defined by COMPUTER. Use HELP COMPUTER for more details.
The later message is more useful. But I assume both have the same cause.
Does this work:
a = sparse(1, 1, 1, 2^24, 2^23, 1);
This is smaller than MAXSIZE.
0 Comments
Steven Lord
on 4 Oct 2017
The amount of memory required to store a sparse matrix is a function not only of the number of nonzero elements but also of the number of columns in the sparse matrix. See the documentation for the MEX API function mxSetJc, the paper "Sparse Matrices in MATLAB: Design and Implementation" by Gilbert, Moler, and Schreiber in the PDF documentation, and the description of the compressed sparse column format in the Wikipedia page on sparse matrices.
When you try to create a sparse matrix of size [2^30, 2^30] that will require a jc array of length 2^30+1 which is fairly large.
0 Comments
See Also
Categories
Find more on Sparse Matrices in Help Center and File Exchange
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!