Prev: why an exhaustive search is faster than fmincon when solvinga non-linear programming problem?
Next: MEAN AND FUCTION
From: H on 2 Jun 2010 08:23
Hi!

I'm solving quite large shortest path problems with matlabBGL (which is great btw). I have several weight matrices representing different kinds of optimization criteria. My idea is to add weights to the matrices then then sum them together (and then feed the result matrix to the algorithm).

My problem is that when I sum a pair of large (for example 100MB) sparse matrices, the memory requirement seems to be very high. What I mean that for example I have 512MB of free continuous memory according to memory-function, but I still get "out of memory error when I try to sum the to sparse matrices". I managed to increase my page file and did the 3GB switch -thing but as the ram runs out, the hard drive paging operations take very long time.

The matrices have non zero elements in the exact same indexes so the result matrix takes the same amount of memory than the matrices that I'm summing.

Could I for example try to cut the matrices in smaller blocks, sum the blocks and then put the result matrix together or something?

Any help is appreciated!
 | 
Pages: 1
Prev: why an exhaustive search is faster than fmincon when solvinga non-linear programming problem?
Next: MEAN AND FUCTION