MATLAB: Do I receive an error when using a large sparse matrix in the LU function within MATLAB 6.5 (R13)

factorizationlargeluMATLABmemoryofoutsparseumfpack

Why do I receive an error when using a large sparse matrix in the LU function within MATLAB 6.5 (R13)?
If I have a 60,000-by-60,000 matrix, and attempt to call the LU function with four output arguments, I receive the following error:
 Sparse lu with 4 outputs (UMFPACK) failed.
Looking at the diagnostic information provided by using this command before attempting to call LU:
spparms('spumoni',1);
it appears that UMFPACK runs out of memory. Is there a way to compute the LU factorization of this matrix?

Best Answer

This is a bug in MATLAB 6.5 (R13) in the way that the LU function handles the factorization of large sparse matrices. Our development staff is investigating this issue.
To avoid this problem, you can use the 3-output form of LU, which does not use the UMFPACK libraries. Alternately, the documentation from the UMFPACK website indicates that this may be a bug in the UMFPACK 4.0 routines that MATLAB 6.5 (R13) uses for sparse LU factorization.
Related Question