Robinson-Fisher (aka chi) beats Bayes chain rule
Greg Louis
glouis at dynamicro.on.ca
Wed Dec 4 23:25:04 CET 2002
On 20021204 (Wed) at 1526:37 -0500, Scott Lenser wrote:
> > My problem is, of course, that a double between 1 and 1 - 1.0e-16
> > is indistinguishable from 1; that's why I get at least 71 1.0's at the
> > top of the nonspam list.
> >
>
> A common technique used by AI researchers to get around this is to use the
> log of the probability.
Did that to avoid dynamic-range problems, but I never thought of doing
it for the outcome (you can tell I'm new at this game). Doesn't change
the experimental outcome, of course, but it would be a trick to bear in
mind if I ever encounter an overly dogmatic method that gives fewer
total errors than what I use now.
--
| G r e g L o u i s | gpg public key: |
| http://www.bgl.nu/~glouis | finger greg at bgl.nu |
More information about the Bogofilter
mailing list