Copy Link
Add to Bookmark
Report
Machine Learning List Vol. 2 No. 07
Machine Learning List: Vol. 2 No. 7
Wednesday, April 25, 1990
Contents:
Analogy and Case Based Reasoning.
Generating random data to test machine learning programs.
EWSL 91 CALL FOR PAPER
The Machine Learning List is moderated. Contributions should be relevant to
the scientific study of machine learning. Mail contributions to ml@ics.uci.edu.
Mail requests to be added or deleted to ml-request@ics.uci.edu. Back issues
may be FTP'd from ics.uci.edu in /usr2/spool/ftp/pub/ml-list/V<X>/<N> or N.Z
where X and N are the volume and number of the issue; ID & password: anonymous
------------------------------
Date: Tue, 10 Apr 90 14:08:04 +0200
From: Fred Tusveld <hcsrnd!tusveld@relay.EU.NET>
Subject: Analogy and Case Based Reasoning.
I'm wondering if there are mailing-lists or newsgroups on the net that
discuss analogical and/or case-based reasoning, or other groups
discussing learning. Also, the amount of interest in this group seems
small compared to all the research being carried out in the ML area.
Thanks in advance,
Fred Tusveld
HCS Industrial Automation R&D
tusveld@hcsrnd.uucp.nl
tel. +31 55 498600
[I'm not aware of any mailing lists discussing those topics. However, they
would be appropriate for discussion here. I mail ML-LIST to over 250 people
and it is redistributed to many others at universities and corporations.
Of course, fewer than this number have submitted messages. My largest
concern in starting this was that the volume would be too high and the
messages would not be interesting to researchers active in Machine Learning.
If you'd like to see the volume increased, submit a message on a topic
of interest to you, or provide a list of recent tech reports from your
organization with ordering information, or a review of a book or article. -MP]
------------------------------
Subject: Generating random data to test machine learning programs.
Date: Wed, 25 Apr 90 23:18:48 -0700
From: Michael Pazzani <pazzani@ICS.UCI.EDU>
Message-ID: <9004252318.aa11909@ICS.UCI.EDU>
What is a good method for generating positive and negative examples to
test learning programs that accept relations as input. For example,
most EBL programs deal with complex training instances such as
color(obj1,red).
part_of(obj1,concavity1).
part_of(obj1,handle1).
part_of(obj1,bottom1).
light(obj1).
owner(obj1,edgar).
direction(concavity1,upward_pointing).
shape(bottom1,flat).
type(handle1,handle).
type(concavity1,concavity).
type(bottom1,bottom).
Given a domain theory for this concept:
cup(X) :- open_vessel(X),stable(X),liftable(X).
open_vessel(X) :- type(Y,concavity),part_of(X,Y),direction(Y,upward_pointing).
stable(X) :- type(Y,bottom),part_of(X,Y),shape(Y,flat).
liftable(X) :- type(Y,handle),part_of(X,Y),light(X).
Positive examples can be generated easily by taking the leaves of a
random proof tree (and perhaps adding some irrelevant features).
However, it's not clear that negative examples can be generated as easily.
Any suggestions?
The goal of generating the positive and negative examples is to test
learning programs that deal with incomplete or incorrect theories.
That is, a complete and correct domain theory will be used to
generate examples, and then errors are purposely introduced into
the domain theory.
Mike Pazzani
------------------------------
Date: Fri, 30 Mar 90 09:43:21 +0200
From: yk%FRLRI61.BITNET@cunyvm.cuny.edu
Subject: EWSL 91 CALL FOR PAPER
EUROPEAN WORKING SESSION ON LEARNING. EWSL-91
EWSL-91 will be held in Porto 6,7,8 March 1991
Important dates
Oct. 15th 1990: deadline for paper submission
Dec. 15th 1990: notification to authors
Jan. 15th 1991: camera-ready copies received by the publisher
Traditionally, EWSL reflects the latest results obtained by the artificial
intelligence approach to machine learning. Submissions relative to this topic
are the most welcome, with an emphasis on "real-world" applications. We would
like to also open discussions with other approaches such as
- knowledge acquisition when it makes use of ML techniques,
- genetic algorithms and neural networks, when their relations to background
knowledge are carefully related,
- data analysis and statistical approaches to induction, when they take care
of the understandability of their results by their field users,
- existing or potential applications of ML to other fields (robotics,
vision, etc.).
Local Chairman: Pavel BRAZDIL Univ. Porto Rua Dr. R. Frias 4200 Porto Portugal
Programme Chairman: Yves KODRATOFF Univ. Paris-Sud Equipe
Inference et Apprentissage Building 490, LRI, F-91405 Orsay France
Submit FIVE copies of a full paper (abstracts will not be considered)
to the Programme Chairman, before the 15th of October 1990. The
length of the submitted paper must be in accordance with the
achievement of the research. Papers with more than 6000 words will not
be considered. Long papers must state clearly their relation with
other approaches, their implementation state, and their validation
(experimental or theoretical). Also very much welcome are short papers
in which the main idea is briefly presented, illustrated by a
significant example. The first page of submitted papers must give
each of the following informations: name, address, affiliation of the
author(s), title of the paper, number of words in the paper, summary
of the paper.
Keynote Speaker
Jaime CARBONELL (Carnegie Mellon University, Pittsburgh, USA)
Panels (tentative)
Knowledge acquisition and ML.
Applications of ML.
Learnability and its relationship to existing programs.
Members of the Programme Committee
BRATKO Ivan (Yugoslavia), BRAZDIL Pavel (Portugal), CLARK Peter (UK), DE JONG
Kenneth (USA), DE RAEDT Luc (Belgium), GANASCIA Jean Gabriel (France),
KODRATOFF Yves (France), LAVRAC Nada (Yugoslavia), LOPEZ DE MANTARAS
Ramon (Spain), MORIK Katharina (FRG), MOZETIC Igor (Austria), MUGGLETON Stephen
(UK), SAITTA Lorenza (Italy), SEGRE Alberto (USA), SHAVLIK Jude (USA), SLEEMAN
Derek (UK), TECUCI Gheorghe (Rumania), VAN SOMEREN Marteen (The Netherlands).
------------------------------
END of ML-LIST 2.7
