[ https://issues.apache.org/jira/browse/MATH-664?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13100318#comment-13100318 ]
Phil Steitz commented on MATH-664: ---------------------------------- At the expense of returning nonsense if the problem is singular. The only positive rationale for this change is that QR is more stable numerically for near-singular positive definite matrices. It would be great to add some references supporting that statement. While the covariance matrices are generally going to be small, performance should also be considered. Cholesky should also be considered as a possible replacement for LU. I will open a separate issue on QR's lack of a meaningful singularity test. > Replace "LUDecompostionImpl" with "QRDecompositionImpl" in > "AbstractLeastSquaresOptimizer" > ------------------------------------------------------------------------------------------ > > Key: MATH-664 > URL: https://issues.apache.org/jira/browse/MATH-664 > Project: Commons Math > Issue Type: Improvement > Reporter: Gilles > Assignee: Gilles > Priority: Minor > Fix For: 3.0 > > > In some cases, the "getCovariances()" method throws a > "SingularMatrixException". This can be avoided by using "QR" instead of "LU" > decomposition. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira