Good afternoon,
I have an application in which I'm using gsl_linalg_cholesky_decomp()
to decompose a symmetric positive-definite matrix, A. However, the
decomp fails in a very small percentage of the matrices that I feed to
this function due to round-off error and all of that fun stuff.
Does anyone happen to know of a good method that will still give me
the L L^T decomposition for A when one of these errors is detected? For
example, is there a method whereby I can add a small-magnitude diagonal
matrix to A, Cholesky decomp that, and modify the result to obtain the
decomp as though I'd done it with A instead of the modified A?
-Daniel
_______________________________________________
Help-gsl mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-gsl
