--- Toad <[EMAIL PROTECTED]> wrote: 
> A theory:
> NGRouting is not working because it can't predict classic routing unless
> the nodes using classic routing specialize quite obviously, which they
> don't.

Seems plausible.  The old system used a really "grainy" estimator, with serveral 
orders of
magnitude more points than the new 10 point estimators.

> How to falsify: 
> 1. Prove that classic routing nodes DO indeed specialize in a fairly
> obvious manner.

Maybe you could mix the estimators to have that happen better.  It seems an old system 
could learn
what an NG node is specialized in, but the other way around probably won't work.

Could you run both models, grade them on thier accuracy and then use the "best 
estimator"?

The old nodes would eventually specialize "obviously", if thier neighbors are mostly 
NG routing.

> Or
> 2. A very large NGRouting-only test network might do it...
> 
> any other ideas?

Question: I thought you guys did spot tests watching the performance improvements at a 
single node
doing NG.  You saw an improvement in peformance right?  How come?

__________________________________________________________________

Gesendet von Yahoo! Mail - http://mail.yahoo.de
Logos und Klingelt�ne f�rs Handy bei http://sms.yahoo.de
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to