me rindo: training to exhaustion

[Tom Anderson]

On Sun, 2005-05-08 at 07:05, Boris 'pi' Piwinger wrote:
> David Relson <relson at osagesoftware.com> wrote:
> >I don't recommend train to exhaustion.  True, it'll give you short term
> >accuracy.  However, it will also lessen your long term accuracy.  
> 
> There is really no hint for that. I use it for many months
> and have no need to mess around with my database. I just do
> the corrections runs needed every few days.

I agree.  It works great with no long term degradation that I can see. 
I receive very few unsures, even less false negatives, and virtually
zero false positives.  And when I train to exhaustion, it ensures that
any which do slip by are the last ones like it.  

If there were any support for long term accuracy affects, they are
nullified by the fact that incoming spam is likely to have changed
significantly over time anyway.  It's far more important to address
today's and tomorrow's email than trends from a year ago.  Train to
exhaustion and you'll almost certainly reduce tomorrow's spam.  You
can't predict what will occur in a year, but today's training will not
likely be very accurate then anyway.  Bogofilter is not a one-time
fix... it is a lifestyle of tweaking and correcting.  Ensuring proper
behavior today is most important.

Tom