|
From: | Daniel Neilson |
Subject: | [Help-gsl] Recovering from failed Cholesky decomp. |
Date: | Wed, 24 Mar 2010 12:38:53 -0600 |
User-agent: | Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1.8) Gecko/20100227 Thunderbird/3.0.3 |
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
[Prev in Thread] | Current Thread | [Next in Thread] |