Copy Link
Add to Bookmark
Report
Machine Learning List Vol. 6 No. 12
Machine Learning List: Vol. 6 No. 12
Tuesday, April 19, 1994
Contents:
MACHINE LEARNING: A Multistrategy Approach, Volume IV
Extension for the AAAI Workshop on Indexing and Reuse in Multimedia
Updated version of FOCL available by anonymous ftp
ASIS SIG/CR's 5th Classification Research Workshop - Call for Papers.
World Congress on Medical Physics and Biomedical Engineering - RIO'94
CFP IBEROAMERICAN CONGRESS ON AI
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 pub/ml-list/V<X>/<N> or N.Z where X and N are
the volume and number of the issue; ID: anonymous PASSWORD: <your mail address>
----------------------------------------------------------------------
Date: Fri, 15 Apr 94 18:06:19 EDT
From: "Ryszard S. Michalski" <michalsk@aic.gmu.edu>
Subject: MACHINE LEARNING: A Multistrategy Approach, Volume IV
A N A N N O U N C E M E N T OF N E W B O O K
MACHINE LEARNING:
A Multistrategy Approach, Volume IV
Edited by Ryszard Michalski (George Mason University) and
Gheorghe Tecuci (George Mason University and Romanian Academy)
Morgan Kaufmann Publishers, 1994, ISBN 1-55860-251-8
Multistrategy learning is one of the newest and most challenging
research directions in the development of machine learning
systems. The objectives of research in this area are to study
trade-offs between different learning strategies and to develop
learning systems capable of employing multiple inference types or
computational paradigms in a learning process. Multistrategy
systems offer significant advantages over monostrategy systems,
which currently dominate the machine learning landscape, as they
are more flexible in the type of input they can learn from and
the type of knowledge they are able to acquire. Therefore,
they have the potential to be applicable to a wide range of
practical problems.
This volume is the first book in this fast-growing field. It
contains tutorial-style contributions by leading researchers
specializing in this area, and an extensive 600-entry bibliography,
indexed by different categories and subcategories. It is the fourth
volume in a series of highly acclaimed books in the field of machine
learning, which includes Volumes I, II, III of Machine Learning:
An Artificial Intelligence Approach, all published by Morgan Kaufmann.
As the first available book on this subject, it is intended to serve
several needs and may be of interest to a wide spectrum of readers.
For students in Machine Learning and related fields, is can serve
as a primary or supplementary texbook covering this important subfield
of machine learning. For researchers in Machine Learning, AI,
Cognitive Science, Intelligent Systems and related disciplines, it provides
a comprehensive source of information about multistrategy learning.
CONTENTS
Preface
PART ONE: GENERAL ISSUES
Chapter 1
Inferential Theory of Learning: Developing Foundations for
Multistrategy Learning
Ryszard S. Michalski
Chapter 2
The Fiction and Nonfiction of Features
Edward J. Wisniewski and Douglas L. Medin
Chapter 3
Induction and Organization of Knowledge
Yves Kodratoff
Chapter 4
An Inference-Based Framework for Multistrategy Learning
Gheorghe Tecuci
PART TWO: THEORY REVISION
Chapter 5
A Multistrategy Approach to Theory Refinement
Raymond J. Mooney and Dirk Ourston
Chapter 6
Theory Completion using Knowledge-Based Learning
Bradley L. Whitehall and Stephen C-Y. Lu
Chapter 7
GEMINI: An Integration of Analytical and Empirical Learning
Andrea P. Danyluk
Chapter 8
Theory Revision by Analyzing Explanations and Prototypes
Stan Matwin and Boris Plante
Chapter 9
Interactive Theory Revision
Luc De Raedt and Maurice Bruynooghe
PART THREE: COOPERATIVE INTEGRATION
Chapter 10
Learning Causal Patterns: Making a Transition from
Data-driven to Theory-driven Learning
Michael Pazzani
Chapter 11
Balanced Cooperative Modeling
Katharina Morik
Chapter 12
WHY: A System that Learns Using Causal Models and Examples
Cristina Baroglio, Marco Botta and Lorenza Saitta
Chapter 13
Introspective Reasoning using Meta-explanations for
Multistrategy Learning
Ashwin Ram and Michael Cox
Chapter 14
Macro and Micro Perspectives of Multistrategy Learning
Yoram Reich
PART FOUR: SYMBOLIC AND SUBSYMBOLIC LEARNING
Chapter 15
Refining Symbolic Knowledge Using Neural Networks
Geoffrey G. Towell and Jude W. Shavlik
Chapter 16
Learning Graded Concept Descriptions by Integrating
Symbolic and Subsymbolic Strategies
Jianping Zhang
Chapter 17
Improving a Rule Induction System Using Genetic Algorithms
Haleh Vafaie and Kenneth DeJong
Chapter 18
Multistrategy Learning from Engineering Data by Integrating
Inductive Generalization and Genetic Algorithms
Jerzy Bala, Kenneth DeJong, and Peter Pachowicz
Chapter 19
Comparing Symbolic and Subsymbolic Learning: Three Studies
Janusz Wnek and Ryszard S. Michalski
PART FIVE: SPECIAL TOPICS AND APPLICATIONS
Chapter 20
Case-Based Reasoning in PRODIGY
Manuela Veloso and Jaime Carbonell
Chapter 21
Genetic Programming: Evolutionary Approches to
Multistrategy Learning
Hugo de Garis
Chapter 22
Experience-Based Adaptive Search
Jeffrey Gould and Robert Levinson
Chapter 23
Classifying for Prediction: A Multistrategy Approach to
Predicting Protein Structure
Lawrence Hunter
Chapter 24
GEST: A Learning Computer Vision System That
Recognizes Hand Gestures
Jakub Segen
Chapter 25
Learning with a Qualitative Domain Theory by Means of
Plausible Explanations
Gerhard Widmer
BIBLIOGRAPHY OF MULTISTRATEGY LEARNING
Janusz Wnek and Mike Hieb
------------------------------
Date: Mon, 11 Apr 1994 16:04:03 -0600
From: kedar@ils.nwu.edu
Subject: Extension for the AAAI Workshop on Indexing and Reuse in Multimedia
***DEADLINE EXTENDED to April 22, 1994 for submissions to the
AAAI 94 Workshop on Indexing and Reuse in Multimedia Systems***
Due to overlapping deadlines with related conferences (e.g. ACM Multimedia)
we have extended our deadline to April 22, 1994. Please send submissions
to:
Catherine Baudin (chair)
AI Research Branch
NASA Ames Research Center. MS 269/2
Moffett Field, CA 94035. USA.
Tel (415) 604 4745
FAX (415) 604 3594
baudin@ptolemy.arc.nasa.gov
We are especially encouraging contributions from video archivists,
videographers, editors, or sound designers who may not have an AI
background, but have experience in representing and indexing multiple
media.
AAAI 94 Workshop on Indexing and Reuse in Multimedia Systems
Multimedia systems for documentation, training, entertainment, or news
use a variety of indexing schemes to annotate and access media for
further use. As these systems grow, issues of how to represent,
acquire, and refine indexing knowledge for effective information
access and reuse become crucial. Yet, analysis of the indexing task
and research on indexing methods and tools for multimedia are just
beginning to emerge.
Because automated parsing techniques for time-based and visual media
are currently in the early stages of development, systems for
multimedia material must often operate with partial representations of
their information content. However, these systems can still produce
information that is useful for the end-user. Such hybrid human-machine
representation and communication provides opportunities to investigate
interactive methods for index acquisition and refinement, which may
complement and improve automatic parsing techniques.
This workshop focuses on methods and tools for indexing, searching,
and reusing information in multimedia systems. We would like to bring
together researchers in knowledge acquisition, case-based reasoning,
multimedia system design, human-computer interaction, speech/visual
perception, and other relevant disciplines. We also encourage
contributions from video archivists, videographers, editors, or sound
designers who may not have an AI background, but have experience in
representing and indexing multiple media.
We welcome contributions in the areas of: (1) indexing tools; (2)
index representations and refinement for information reuse; and (3)
studies of indexing practice. Specifically:
INDEXING TOOLS:
o Knowledge acquisition or machine learning techniques to acquire or
refine indexing knowledge. Incremental refinement of semiformal
representations. Incremental generation of hypermedia links, etc.
o Integration of automated parsing techniques that process signals or
text with conceptual (semantic) indexing techniques.
o Analysis of the types of interactions with humans that a system can
use to reorganize its memory.
INDEX REPRESENTATIONS AND REFINEMENT FOR REUSE:
o Index representations and reuse for different purposes/classes of
users.
o Completing, generalizing, adding context to indexing schemes.
o Index representations and acquisition which address the media-specific
properties of time-based and visual media.
o Comparison between different indexing schemes: domain-dependent vs.
domain-independent, semantic vs. episodic, absolute (information
content) vs. relative (links between information fragments as in
hypermedia).
o Empirical evaluation of information organization and reuse in
multimedia systems.
THE PRACTICE OF INDEXING:
o Case studies and analysis of indexing practice which deal with
pragmatic questions: Who are the indexers? How deep must their
understanding be of the content of the indexed information? When is
indexing performed, in real-time during the data generation or after
the fact?
The duration of the workshop will be one day and will consist of at
most fifty (50) participants. To leave time for discussion only a
subset of the selected papers will be presented. Each session will
comprise: (1) presentations by authors of selected papers; (2) a short
analysis/critique of the papers presented by the session chair; (3) a
discussion with the authors and the workshop participants. In
addition we will have a panel and one or two invited speakers.
Workshop committee:
Catherine Baudin (chair)
AI Research Branch
NASA Ames Research Center. MS 269/2
Moffett Field, CA 94035. USA.
Tel (415) 604 4745
FAX (415) 604 3594
baudin@ptolemy.arc.nasa.gov
Smadar Kedar, ILS, Northwestern University, kedar@ils.nwu.edu
(co-chair) Daniel M. Russell, Apple computer inc.,
dmrussel@taurus.apple.com (co-chair) Marc Davis, MIT Media Laboratory
and Interval Research Corp., davis@interval.com (co-chair)
Ray Bareiss, ILS, Northwestern University,
Guy Boy, EURISCO,
Tom Gruber, Stanford University,
Ken Haase, MIT Media Lab,
Doug Lenat, MCC,
Nathalie Mathe, NASA Ames Research Center
Scott Minneman, Xerox PARC
Dick Osgood, ILS, Northwestern University
Jim Spohrer, Apple Computer Inc.
Meg Withgott, Interval Research Corp.
Please send four copies of a five to twelve page paper to Catherine
Baudin. The papers should be received by April 22. Those who would like
to attend without submitting a paper should send a one to two page
description of their relevant research interests. Presenters are
encouraged to bring live demos and/or videos of their work. Organizers
intend to publish a selection of the accepted papers as a book or as a
special issue of a journal.
------------------------------
Subject: Updated version of FOCL available by anonymous ftp
Date: Tue, 19 Apr 1994 13:48:34 -0700
From: Michael Pazzani <pazzani@pan.ICS.UCI.EDU>
FOCL is a machine learning system that extends Quinlan's FOIL program
by containing a compatible explanation-based learning component. FOCL
learns Horn Clause programs from examples and (optionally) background
knowledge.
An updated version of FOCL (v 2.1) is now available by anonymous ftp
from ics.uci.edu
For details on FOCL, see:
Pazzani, M. & Kibler, D. (1992). The role of prior knowledge in
inductive learning. Machine Learning, 9, 54-97.
It is available in one of two forms:
1. A (binhexed, Compacted) Macintosh application. This is stored in
pub/machine-learning-programs/FOCL-1-2-3.cpt.hqx
In addition to the machine learning program, this contains a graphical
interface that displays the search space explored by FOCL, so it is a
useful pedagogical tool.
This application also contains a graphical interface for building rule
bases, so you can ignore the machine learning aspects, and use it as
an expert system shell with the following capabilities:
* A backward-chaining rule interpreter.
* A graphical rule and fact editor.
* Graphical display of the rule base.
* (Simple) Natural language explanation of inferences
* Menu-based facilities for editing rules and adding natural language
translations to rules.
* Optional typing of variables and checking the rule base for type conflicts
* Tracing of rules
* Analysis of the accuracy of rules in a rule base.
The expert system has been used successfully in an undergraduate
laboratory course. Sample rule bases are included. A minimum of 6MB
of memory is recommended for the application. In addition, a
manual (that should print on any Macintosh printer) is available in
pub/machine-learning-programs/FOCL-1-2-3-manual.hqx
2. Common lisp source code.
This is portable source code for the machine learning program
only, since the interface currently depends on the MAC.
Available by Anonymous ftp from ics.uci.edu in
pub/machine-learning-programs/FOCL-1-2-3.tar.Z
1. FTP FOCL-1-2-3.tar.Z to your machine.
2. Uncompress FOCL-1-2-3.tar.Z creating FOCL-1-2-3.tar
3. Extract the FOCL-1-2-3 files using tar -xf FOCL-1-2-3.tar
4. There will be a new directory called FOCL-1-2-3 containing
the subdirectories source, compiled, and sample-domains, and
the files load.lisp.
5. Start up common lisp in the FOCL-1-2-3 directory.
6. Type the following command to load and compile FOCL
(load "load.lisp")
(load-source)
(compile-all)
(load-comp)
7. After step 6 has been done once, to load the compiled code
(load "load")
(load-comp)
8. Load one of the sample domain files from the directory sample-domains
e.g. (load "sample-domains/cup.lisp")
since all of these sample files contain a def-focl-problem learning
parameter setting it is possible to invoke FOCL using the command:
(learn)
most of these files also contains a function, (e.g. (test-cup))
that calls focl with an explicit set of parameters.
9. Warning. FOCL takes a large number of keyword parameters. Most
of them are there to either
a. permit us to turn off some capability of FOCL to see how it
performs without that capability.
b. To make FOCL run faster by ignoring some capability that we
know will not affect the answer
c. experiment with some capability (e.g., the evaluation function)
It is possible to give inconsistent parameter settings and have
strange things happen.
Unfortunately, we haven't yet written a manual to explain all the
settings of the learning program. It is best to start with the
parameter settings of an example from the sample-domains directory.
If you have problems, or run into bugs, send mail to pazzani@ics.uci.edu
or brunk@ics.uci.edu and we will try to address them as time permits.
If you provide us with data and a description of your problem, we can
help getting you started with FOCL.
Mike Pazzani
Cliff Brunk
ICS Dept
UC Irvine,
Irvine, CA 92717
USA
If you use a copy of FOCL, please send mail to pazzani@ics.uci.edu so
we can inform you of upgrades
------------------------------
Date: Mon, 18 Apr 1994 22:22:09 -0400 (EDT)
From: Ray Schwartz <schwart@gandalf.rutgers.edu>
Subject: ASIS SIG/CR's 5th Classification Research Workshop - Call for Papers.
CALL FOR PARTICIPATION
5th ASIS SIG/CR Classification Research Workshop
QUESTIONS, CONTROVERSIES AND CONCLUSIONS
IN CLASSIFICATION RESEARCH
The American Society for Information Science Special Interest
Group on Classification Research (ASIS SIG/CR) invites
submissions for the 5th ASIS Classification Research Workshop, to
be held at the 57th Annual Meeting of ASIS in Alexandria, VA. The
workshop will take place Sunday, October 16th, 1994, 8:30 a.m. -
5:00 p.m. ASIS '94 continues through Thursday, October 20th.
The CR Workshop is designed to be an exchange of ideas among
active researchers with interests in the creation, development,
management,representation, display, comparison, compatibility,
theory, and application of classification schemes. Emphasis will
be on semantic classification, in contrast to statistically based
schemes. Topics include, but are not limited to:
- Warrant for concepts in classification schemes.
- Concept acquisition.
- Basis for semantic classes.
- Automated techniques to assist in creating classification
schemes.
- Statistical techniques used for developing explicit semantic
classes.
- Relations and their properties.
- Inheritance and subsumption.
- Knowledge representation schemes.
- Classification algorithms.
- Procedural knowledge in classification schemes.
- Reasoning with classification schemes.
- Software for management of classification schemes.
- Interfaces for displaying classification schemes.
- Data structures and programming languages for classification
schemes.
- Image classification.
- Comparison and compatibility between classification schemes.
- Applications such as subject analysis, natural language
understanding, information retrieval, expert systems.
The CR Workshop welcomes submissions from various disciplines.
Those interested in participating are invited to submit a short
(1-2 page single-spaced) position paper summarizing substantive
work that has been conducted in the above areas or other areas
related to semantic classification schemes, and a statement
briefly outlining the reason for wanting to participate in the
workshop. Submissions may include background papers as
attachments. Participation will be of two kinds: presenter and
regular participant. Those selected as presenters will be invited
to submit expanded versions of their position papers and to speak
to those papers in brief presentations during the workshop. All
position papers (both expanded and short papers) will be
published in proceedings to be distributed prior to the workshop.
The workshop registration fee is $35.00.
Traditionally, a revised version of the proceedings is published the
following year as a volume of Advances in Classification Research
(ASIS Monograph Series, published by Learned Information, Medford,
New Jersey, USA)
Submissions should be made by email, or diskette accompanied by
paper copy, or paper copy only (fax or postal), to arrive by May
15, 1994, to:
*Raya Fidel, Graduate School of Library and Information Science,
University of Washington, FM-30, Seattle, WA 98195; Internet:
fidelr@u.washington.edu; Phone: 206-543-1888; Fax: 206-685-8049*
------------------------------
Date: Tue, 12 Apr 94 18:05:03 EST
From: Marcelo de Carvalho Bossan <bossan@bio1.peb.ufrj.br>
Subject: World Congress on Medical Physics and Biomedical Engineering - RIO'94
Cc: cybsys-l@BINGVMB
World Congress on Medical Physics and Biomedical Engineering - RIO'94
The next World Congress on Medical Physics and Biomedical
Engineering will be held in Rio de Janeiro, Brazil, 21-26 August 1994.
The scientific program includes approximately 1900 papers. Round
table, state-of-the-art and tutorial sessions are being finalized, and
nearly 250 of the world's leading experts have so far sent abstracts
for their invited presentations.
Issues in Neural Networks (NN) are inserted mainly in the
topic Expert and Decision Support Systems (chairmen: N. Saranummi,
Finland, and R. J. Machado, Brazil). A tutorial lecture on Hybrid
Expert Systems will be presented by A. Rocha (Brazil) and oral
sessions are programmed on both Neural Networks and Hybrid Expert
Systems. A Round-table on NN in Electrocardiology (speakers: R. G.
Mark, USA, N. Maglaveras, Greece, L. Glass, Austria and R. M.
Sabbatini, Brazil) is also included in the program, in adiction to
scientific sessions focusing NN in various applications.
Attracting the largest possible audience is our present
priority, as we can now present an extensive scientific program of
high standard. We particularly hope that a large number of students
will be able to participate as we offer courses and over 30 tutorial
sessions. We are currently also making arrangements for low-priced
board and lodging for students. Cut-price air-fairs to Rio (or
packages including hotels) are available. Supporting from funding
agencies has allowed reduced registration fees for Latin American
participants. Together with the Final Announcement (and the final
letter to the authors) we will give details of the reception of
conference participants at the airport and transports to hotels, which
are currently being organized. A free bus service will operate on
Saturday (20th August) and Sunday (21st) during peak arrival hours.
One of the taxi-companies operating within the airport will be
available to participants arriving at other times (fixed prices fairs,
around US$ 15), with a counter on information regarding the congress
and hotels.
We look forward to welcoming you here, where you can enjoy
stimulating scientific presentations and discussions, renew old
friendships and make new ones, all in exotic and beautiful
surroundings.
For more information:
World Congress on Medical Physics and Biomedical Engineering
Rio de Janeiro - 21-26 August 1994
Congrex do Brasil
Rua do Ouvidor 60, Gr.414
Tel: +55 (021) 224-6080 - Fax: +55 (021) 231-1492
or
bossan@bio1.peb.ufrj.br
------------------------------
From: "Jose Ramirez G. (AVINTA" <jramire@conicit.ve>
Subject: CFP IBEROAMERICAN CONGRESS ON AI
Date: Fri, 15 Apr 1994 10:20:13 -0400 (AST)
CALL FOR PAPERS
IBEROAMERICAN CONGRESS ON ARTIFICIAL INTELLIGENCE
IBERAMIA 94
NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE
CNIASE 94
IBERAMIA 94/CNIASE 94 will be sponsored by the Venezuelan Association
for AI -AVINTA-, the Mexican AI Society -SMIA- and the Spanish
Association for IA -AEPIA-. The goal of the conference is to promote
research and development in Artificial Intelligence and scientific
interchange among AI researchers and practitioners.
IBERAMIA 94/CNIASE 94 will be hosted by the Centro de Investigaciones
Oficina Metropolis and the School of Systems Engineering of the
Universidad Metropolitana of Caracas -UNIMET-, between Tuesday 25th
October and Friday 28th October 1994.
We invite authors to submit papers describing original work in all areas
of AI, including but not limited to:
Machine Learning
Knowledge Acquisition
Natural Language Processing
Genetic Algorithms
Evolutionary Programming
Knowledge Based Systems
Knowledge Representation and Reasoning
Automated Reasoning
Knowledge-based Simulation
Cognitive Modelling
Robotics
Case-based Reasoning
Distributed Artificial Intelligence
Neural Networks
Virtual Reality
All submissions will be refereed for quality and originality. Authors
must submit three (3) copies of their papers (not electronic or fax
transmisions) by June 30, 1.994 to the following address:
AVINTA
Apartado 67079
Caracas 1061
Venezuela
+58-2-2836942,fax:+58-2-2832689
jramire@conicit.ve
or
Universidad Metropolitana
Centro de Investigaciones Oficina Metropolis
Autopista Petare-Guarenas
Distribuidor Universidad
Terrazas del Avila
Caracas 1070-A
Venezuela
+58-2-2423089, fax:+58-2-2425668
abianc@conicit.ve
All copies must be clearly legible. Notification of receipt will be
mailed to the first author.
Papers can be written in English, Spanish or Portuguese and must be
printed on 8 1/2
x 11 inches paper using 12 point type (14 point type for headings). The
body of submitted papers must be at most 12 pages. Each copy must have a
title page (separate from the body of the paper) containing title of the
paper, names and addresses of all authors, telephone number, fax number,
electronic mail address and a short (less than 200 word) abstract.
All accepted papers will be published in full length by McGraw-Hill.
Important dates:
Deadline for paper submission: June 30, 1994
Notification of acceptance: July 30, 1994
Camera Ready Copy: September 9,1994
Location:
Caracas is located on the north of South America, facing the Caribbean
Sea; is a modern city with an enjoyable wheather all year long (20 C to
30 C), with many interesting sites including cultural complexes,
historical downtown, shopping malls and excelent hotels and restaurants
offering the best food from all over the world.
The Simon Bolivar International Airport is 45 minutes from downtown, and
have regular flights to all major cities in the world.
------------------------------
End of ML-LIST (Digest format)
****************************************
