Re: more graphs

Bill Yerazunis wrote:
> I'm even more concerned with your .css files.
> 
> I see that they are well under 10% full- that's very strange
> for well-trained systems ( slot usage is asymptotic and dependent
> on the language and corpus, but I've never seen email corpora 
> with this small a footprint).

I will run the tests again and give you actual message counts of what I trained 
on for the default case but, this is my five-day message sample.  This is 
typical for me and other people I work with.  The important thing to note is it 
didn't used to do this with even smaller samples.  Within three to five day run, 
it would normally train out just fine to where you would only have five or 10 
messages a day tops sitting in the uncertain zone (as defined by me)

> I'm not surprised.  CRM114 is very cache sensitive; once you bust
> cache, it's gonna slow way down.

well, I bust the cache  at four or five parallel scoring runs.  I'm beginning to 
think this is where a demon or some sort of process queuing mechanism wouldn't 
be such a bad idea.  I really hate the idea of adding another queue because it 
does slow things down but, it might be necessary.  While I'm running the tests 
again, I will try to get you some timings.

> That's strange.  It should speed up when you enforce 
> serialization... unless you're doing something like spinwaiting.

I only enforce serialization for training.  Like I said, scoring is shared, 
training is exclusive.  on the other hand, my test case is serialized completely 
and it takes forever and a day.  That reminds me, I better look into the VMWare 
logs to see if it's doing something really horrible.

> Please tell me that you're not spinwaiting?  Please?

I'm not spin waiting.  Don't know what the hell the file system lock is doing 
however.

-------------------------------------------------------------------------
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/