Decomposition object saveobj method not supported?

15 views (last 30 days)
Hi there,
I need to perform a lot of computation. I am using some sort of 3D histogram, which is filled with logs about entitites for each cell. Each cell is associated with some particular matrix, which plays the role of RHS of system of linear equations. In order to speed up my computations, solving systems with variable LHS but for specific cell with constant RHS (mentioned earlier), I want to use decomposition object, which is preferable way of handling such issue. The problem is, that creating the grid can be computationally exhaustive and because of that I create it once, save to the mat file and load. This approach is much faster. But if part of the grid is cell array with decomposition object, I receive this warning
Warning: Saving a decomposition is not supported.
> In decomposition/saveobj (line 667).
Minimal not working example could be
decomp = decomposition([1 2; 3 5], 'lu');
save('decomp.mat', 'decomp', '-v7.3');
I use Matlab R2018a. Thank you for advice.

Accepted Answer

Christine Tobler
Christine Tobler on 17 May 2018
This is a limitation on the decomposition object: It cannot be saved into .mat file. The background for this is that some types of decompositions are stored in an internal format defined by a library we use: We can't practically save these to a file, and if we tried, there would likely be problems when loading such a mat-file on a different computer.
As a workaround, you could instead store the factors of an LU decomposition of each matrix, and then use this matrix to solve the linear system. Unfortunately, this is less elegant than using the decomposition object.
  5 Comments
Shumao Zhang
Shumao Zhang on 18 Aug 2022
Edited: Shumao Zhang on 18 Aug 2022
@Michal Unfortunately I don't see any progress... I eventually gave up parallelizing the decomposition objects.
Michal
Michal on 18 Aug 2022
Edited: Michal on 18 Aug 2022
Take a look on this thread. Should be interesting for you, too!

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!