Re: Exim4, Clamav, SA-Exim,
On Tuesday 07 October 2003 23:27, Alan Shutko wrote:
> In my experience, still really, really good. I haven't been getting
> any additional false possitives, and I haven't been missing any more
> spam than normal.
I see. Well, I haven't read up on how the Bayesian filter works (though
I have been studying quite a lot of Bayesian statistics in astronomy, I
have friends who belong to the Bayesian school of thought in statistics
rather than the more common frequentist), I haven't even read Paul
Graham's paper.
However, intuitively, you would have to use a finite number of
characteristics to filter on, and characteristics are computed from ham
and spam. Given the finite number of characteristics, if you used them
to both classify spam and viruses, it will necessarily be less
characteristics to classify either, and so, your statistics is less
favourable, and therefore the probability of error increases.
Obviously, my intuition may be wrong (that's when it starts getting
interesting, usually), or the effect may be neglible. I'd love to see
it studied and quantified, though... Allthough I might not have time to
read it anyway... :-)
Cheers,
Kjetil
--
Kjetil Kjernsmo
Astrophysicist/IT Consultant/Skeptic/Ski-orienteer/Orienteer/Mountaineer
kjetil@kjernsmo.net webmaster@skepsis.no editor@learn-orienteering.org
Homepage: http://www.kjetil.kjernsmo.net/ OpenPGP KeyID: 6A6A0BBC
Reply to: