And another thing with mailtrainer and thick threshold training

Howdy again, long time no see and all that.

See mailtrainer lines 860 and 915 (test_traingood/test_train_spam):

		eval /:@: :*:pr: < :*:thick_threshold: : /

                eval /:@: :*:pr: > (0 - :*:thick_threshold:) : /

Given the fact that today's classify thresholds are assymmetric and -
for 'good' at least - larger than the 'thick threshold' configured for
training when using mailtrainer, shouldn't these lines be changed to
use the basic classify (assymmetric) check thresholds? Or am I
reasoning in the wrong direction here and should I lift the thick
threshold to the MAX of spam_threshold and good_threshold as I want to
ensure that emails which have just been LEARNed, *do* pass the
post-check classify cycle at all times: with the current TRAIN thick
threshold at 5 and the CLASSIFY thresholds at -5/+10, many 'good'
trained items will be STUCK in the 'DON'T KNOW' section for ever.

And THAT I do not like. Or should I bow my head and live with this?


-- 
Met vriendelijke groeten / Best regards,

Ger Hobbelt

--------------------------------------------------
web: http://www.hobbelt.com/
 http://www.hebbut.net/
mail: [email protected]
mobile: +31-6-11 120 978
--------------------------------------------------

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/