Lots of zero entries in L factor found by chol and CHOLMOD

2 views (last 30 days)
Kadir
Kadir on 23 Apr 2014
I have used Matlab's chol() and CHOLMOD for decomposing A into L and L'. I know that there will be fill-ins. However, L factor found by chol() has lots of very very small numbers. Number of zero entries in L factor found by CHOLMOD has lots of zero entries. I have run these two methods on a large set of symmetric positive definite sparse matrices from UFL. For instance, L factors of Muu matrix show the above-mentioned anomalies.
For Muu, there are 88.618 nonzeros in the lower triangular part of A. L factor found by CHOLMOD contains 239.912 entries. 186.826 entries are nonzero, remaining 53.086 are zero, which are explicitly stored in sparse data structures. L factor found by chol() contains 1.497.725 entries.
I used SuiteSparse/CHOLMOD/Demo/cholmod_demo.c source code for generating L factor.
I wonder why these zero entries exist in L? Is it correct to remove very very small numbers below a certain threshold or zero entries from L factors? What may be the threshold?

Sign in to answer this question.

Answers (0)

Categories

Find more on Descriptive Statistics in Help Center and File Exchange

Tags

Community Treasure Hunt

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

Start Hunting!