Re: Mixed 64-bit system GerH binaries / BillYscripts

   From: "Ger Hobbelt" <[email protected]>

   [...]

   Example: today, you learn spam or ham using two different statements:

   learn (spam.css) [message]
   learn (ham.css) [message]     -- my [] may be the incorrect ones,
   doesn't matter for this sample.

That's correct.

   and you classify using this:

   classify (spam.css | ham.css) [message]



   The 'unified' approach would look like this:

   learn (spam.css | ham.css) [message] <learn_spam>
   learn (spam.css | ham.css) [message] <learn_ham>

   and you classify using this - as usual:

   classify (spam.css | ham.css) [message]

Ooooh, nice syntax.  I was going to say you just would:

   learn (good.stat) [my_good.txt]
   learn (good.stat) <refute> [my_spam.txt]

and then
    
   classify (spam.stat) [my_unknown.txt]

but allowing polyvalent LEARN (learn positively for before the
OR-bar, learn negatively after the OR-bar) also cleans up
the use of <refute>  - it can now go back to being "forget".  Or
probably we should add a <forget>.  

   See how learn looks more like classify now? Some classifiers already
   _require_ this, and some don't use it. The latter ones can, however,
   use the new interface as well: they just pick that one CSS file
   they're interested in and ignore the other one. Meanwhile, algorithm
   adjustments like I was talking about _can_ be built in and tested
   using the new interface, why I _cannot_ do them using the existing
   interface.

Indeed.  One standard way to access any classifier as desired, no
more funny stuff depending on whether you were using OSB, Winnow,
Hyperspace, SVM, or neural net.

	 - Bill Yerazunis

-------------------------------------------------------------------------
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=/