Re: Mixed 64-bit system GerH binaries / BillYscripts --> two-sided training? YES!

On Sun, Aug 31, 2008 at 08:57:38PM -0300, Fidelis Assis wrote:
> 
> I thought he was mainly interested in noise reduction. EDDC is not
> specific to email and might be of some help. What is tailored to email
> is TONE-HR.

being 'intuitively and empirically derived', unless you studied other data
streams, that's tailor-made (on email, and perhaps on somewhat specific
mail stream - suffice to say plain OSB is better here, which doesn't mean
EDDC per se is worse, simply it's default params don't shape well here and
I hadn't time to fine tune). 

> Storing only diffs?  Wouldn't that make class occurrence rate to be lost?

yep, sounds like early erroneous thinking in crm114 by someone here (me),
but Ger's case is very specific and so such a highly non-linear tailored
(again) filter on the training chain might yield better results, actually
leaving the stats filter to work on diffs only. Kinda special DSP vs 
GP CPU. We'll see.

-- 
paolo

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/