robinson-fisher - two states vs three states

[David Relson]

At 08:36 AM 1/21/03, Boris 'pi' Piwinger wrote:

>David Relson wrote:
>
> >> > Lastly, for those who really, really don't ever want to see a message
> >> > classified as "unsure", you can set the value of ham_cutoff to 0 (and
> >> > bogofilter will only say "Yes" or "No")
> >>
> >>I don't find a command line option for this. Is there one?
> >>It would be nice to have an option which says "use
> >>spam_cutoff as the value of ham_cutoff". It is great that
> >>setting it to 0 has that effect.
> >
> > Hi pi,
> >
> > You've been quiet for many weeks now.  I hope the new year is going well
> > for you.
>
>Lots of things were going on, now I am back on track:-)) And
>for bogofilter I actually wanted to wait for MIME, so I did
>not really look that much at other improvements.
>
> > We don't have a command line option for that - yet.  I've been thinking
> > that "-2" and "-3" would work.  If you don't like those choices, let me 
> know.
>
>Sounds great.

O.K.  I'll add that to my TODO list for 0.10.1

>Til then I will use the following in ~/.bogofilter.cf:
>algorithm=fisher
>ham_cutoff = 0.00
>
>I hope my understanding is correct that this will do the job.

Looks right.  Let me know if you run into trouble.

>So far there is "min_dev=0.0" in /etc/bogofilter.cf. What
>would be a good value to use? Is it correct that you should
>not need to rebuild the database when changing the value?

0.1 works well.  With min_dev at 0.1, using the "-vv" options to generate 
the histogram should show token counts of zero for the 0.40 and 0.50 
rows.  It's a good, visual check that the settings are working.