[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: KDE Naive Bayes (or other ML algorithm)
From:       Yves Glodt <yg () mind ! lu>
Date:       2005-05-26 17:23:09
Message-ID: 200505261923.11560.yg () mind ! lu
[Download RAW message or body]

On Thursday 26 May 2005 18:13, Nathan Toone wrote:
> Seeing as I haven't really gotten a reply that such a thing exists,
> and that it might be something useful, I think that I will probably
> implement one myself - I'd love any help that I can get...

Maybe it could be helpful taking a look at the Mozilla implementation, I 
believe the source can be found here:
http://bonsai.mozilla.org/cvsview2.cgi?subdir=mozilla/mailnews/extensions/bayesian-spam-filter&files=bayesianSpamFilter.js&command=DIRECTORY&branch=HEAD&root=/cvsroot


regards,
Yves

> I'll probably set up a sf account or something to do it in the next
> couple of days or so...
> 
> -Nathan
> 
> On Thursday 26 May 2005 04:43 am, Oliver Brdiczka wrote:
> > Hi Nathan,
> > 
> > I'm currently working with libraries for Bayesian networks and HMM
> > (gmtk, jahmm etc), but these are not really integrated in KDE. I
> > don't know of a library that is really integrated in KDE (via
> > dcop).
> > 
> > I would be very interested in such a library though :-)
> > 
> > In case that you want to implement it yourself, do you need help ?
> > 
> > Oliver
> > 
> > > --- Ursprüngliche Nachricht ---
> > > Von: Nathan Toone <nathan@toonetown.com>
> > > An: kde-devel@mail.kde.org
> > > Kopie: kde-devel@kde.org
> > > Betreff: Re: KDE Naive Bayes (or other ML algorithm)
> > > Datum: Tue, 24 May 2005 17:59:23 -0600
> > > 
> > > On Tuesday 24 May 2005 05:50 pm, Dirk Mueller wrote:
> > > > What is your target ? predict user interaction?
> > > 
> > > I think that there are lots of applications for machine learning
> > > in an operating system.  Automatic learning of mail filtering
> > > rules, better music
> > > playlist randomization, and guessing which news stories a user is
> > > more likely
> > > to want to read in an RSS reader are just a few applications.
> > > 
> > > I work with machine learning algorithms all day long, and there
> > > are
> > 
> > nearly
> > 
> > > limitless possibilities for application.  I was thinking of some
> > > sort of core
> > > library (maybe part of kdelibs, but it doesn't have to be) that
> > > will
> > 
> > take
> > 
> > > calls (either "train" calls or "analyze" calls) via DCOP and
> > > handle all the
> > > algorithm stuff under the covers.  You could even create user
> > > interfaces for
> > > classifier creation for all types of filtering applications - web
> > > page content classification, spam filtering, phishing
> > > recognition, virus detection, etc.
> > > 
> > > It's really just a pet project I'd like to adopt - because ML is
> > > where
> > 
> > my
> > 
> > > interests lie, and it seems that having a "smart" operating
> > > system (or desktop manager, as the case may be) would be a useful
> > > thing...
> > > 
> > > Sorry if that sounds too much like an advertisement - I'm pretty
> > > passionate
> > > about it.  :)
> > > 
> > > -Nathan

-- 
Linux 2.6.11.10 #1 Wed May 18 17:50:55 CEST 2005 i686
 19:10:14 up 1 day, 10:34,  1 user,  load average: 2.38, 1.97, 2.03
 
> > Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<


[prev in list] [next in list] [prev in thread] [next in thread]