🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.ai
Article #9260

Re: METAXA and multiple theories

#9260
From: Stefan Wrobel
Date: Thu, 08 Aug 1991 08:29
50 lines
2929 bytes
Dear Y.O. Busia,

In comp.ai you write:

>I am looking for up-to-date information on the METAXA program.
>The 1989 reference I have, Lecture Notes in Computer Science no.347,
>describes it as "An Inference Engine for Representing Multiple Theories".
>(The program is able to simultaneously hold contradictory points of view
>and thus come to better decisions, eventually).

>I am interested in (1) how successful the project has been, and
>		   (2) whether or not other programs can (or should) do
>		   this as well.

I don't know if Werner Emde is reading this newsgroup, so I'll send you
a quick reply and forward it to him.  METAXA is Werner's learning program,
and it is more than IM2, the inference engine described in the above-mentioned
paper.  I think Metaxa was very successful modeling theory revision 
processes and learning, and I'm sure Werner will tell you more about it.

>I shall be grateful especially for news and also for views on this which
>are posted either directly to the net or via e-mail for a summary.

You may be interested to know that Werner's inference engine (in the
basically the same form as described in the paper) is still in heavy use
here at GMD.  We are developing MOBAL, an integrated knowledge acquisition
and learning system, and are using IM2 as our inference engine.  It's
especially important for us to have the kind of revision support offered
by IM2.  Our system also incorporates a horn-clause learning algorithm
named RDT that is based on the same basic idea as Metaxa learning, namely
the use of rule templates.  We have extended this to exploit the generality
ordering of the hypothesis space as well.  If you are interested, I can
send you papers on both MOBAL and its RDT component.

Anyway, I'll forward this to Werner, and he can tell you more about
Metaxa and IM2.  [Hallo Werner! Viele Gruesse!]

Bye from Bonn,

Stefan.

--------------------------------------------------------
Stefan Wrobel
GMD (The German Natl. Research Center for Comp. Science)
Project Machine Learning Toolbox (MLT)
F3.KI, Pf. 1240
W-5205 St. Augustin 1
Fed. Rep. of Germany
Tel.: 02241/14-0           E-mail: wrobel @ gmdzi.gmd.de
--------------------------------------------------------
Technical Details
Message-ID: <5346@gmdzi.gmd.de>
Path: arc1-ca.pugleaf.net!gmdzi!wrobel
References: <1407@keele.keele.ac.uk>

Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: fdcf7,8358b475b4986377
X-Google-Attributes: gidfdcf7,public
X-Google-Thread: 10461e,8358b475b4986377
X-Google-Attributes: gid10461e,public
X-Google-ArrivalTime: 1991-08-08 01:29:20 PST
Path: arc1-ca.pugleaf.net!gmdzi!wrobel
From: wrobel@gmdzi.gmd.de (Stefan Wrobel)
Newsgroups: comp.ai,eunet.esprit
Subject: Re: METAXA and multiple theories
Summary: Metaxa, IM2, etc.
Keywords: Hypothesis, Contradiction, Knowledge Representation
Message-ID: <5346@gmdzi.gmd.de>
Date: 8 Aug 91 08:29:19 GMT
References: <1407@keele.keele.ac.uk>
Followup-To: poster
Organization: GMD, St. Augustin, F.R. Germany
Lines: 52