Background: I ran into numerical problems when computing the posterior pdf for a high dim problem. The trick is to use the log of the posterior which includes computing the log of the determinant of a covariance matrix. This can be done using, for example,
d = sum(log(eig(Q))), where Q is the (pos-def) covariance matrix [the direct call log(det(Q)) will result in an Inf value]. A faster method is: RQ = chol(Q); d = 2*sum(log(diag(RQ))); A logdet function would probably also be useful for other applications so I'm not sure which package to add it to [one should perhaps test for positive definiteness of Q as well (or let chol report the error)]. Regards, Fredrik ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Octave-dev mailing list Octave-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/octave-dev
