Copy Link
Add to Bookmark
Report
Machine Learning List Vol. 4 No. 17
Machine Learning List: Vol. 4 No. 17
Tuesday, Sept 1, 1992
Contents:
A neat idea from L. Breiman
Call for paper for the AAAI Spring symposium on Incremental Learning
Genetic Algorithms Conf. Call for Papers
AAAI-93 Call for Papers
INTERNATIONAL CONFERENCE ON CONCEPTUAL STRUCTURES
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>
----------------------------------------------------------------------
From: Tom Dietterich <tgd@icsi.berkeley.EDU>
Date: Wed, 19 Aug 92 10:43:30 PDT
To: ml@ics.uci.edu
Subject: A neat idea from L. Breiman
I recently read the following paper by Leo Breiman:
Breiman, L. (1991) The $\Pi$ method for estimating multivariate
functions from noisy data. {\it Technometrics, 33} (2), 125--160.
With discussion.
In this paper, Breiman presents a very neat technique called "back
fitting" that is a very general algorithm idea for improving greedy
algorithms. Suppose we are executing a greedy algorithm for some
task, and at any given point in the process, we have already made
decisions d_1, d_2, ..., d_{k-1} and we are about to make decision
d_k. In the standard greedy algorithm, we choose d_k to be
the locally best decision and then go on to consider d_{k+1}.
However, with backfitting, we first perform the following double loop:
repeat until quiesence:
for i from 1 to k-1 do
"undo" decision d_i (holding all other decisions d_j, j<>i fixed)
and re-make d_i to be the best decision (locally).
In other words, we first hold {d_2, ..., d_{k-1}} constant and see if
we can improve things by re-making decision d_1. Then we hold {d_1,
d_3, ..., d_{k-1}} constant and consider re-making decision d_2, and
so on. We cycle through the previous k-1 decisions making local
improvements until no further improvements can be found. THEN, we
make decision d_k (and repeat the process, of course).
In general, this backfitting process will cost another factor of n in
the algorithm (assuming there are n decisions to be made). In
experiments in three different learning algorithms, Breiman has found
that this method finds much better solutions than a simple greedy
algorithm. Breiman (in various collaborations) is currently applying
this idea to improving CART trees and neural networks.
This idea would probably also find good application in COBWEB-style
algorithms and greedy MDL algorithms.
Tom
------------------------------
Date: Tue, 1 Sep 92 13:29:01 +0200
From: antoine@lri.fr
To: pazzani@ics.uci.edu
Subject: Call for paper for the AAAI Spring symposium on Incremental Learning
Symposium in the AAAI-93 Spring Symposium Series
Title : "Training Issues in Incremental Learning"
Unlike the batch learning model in which the learner receives
all the data before processing them, in the incremental
(or on-line) learning model the learner must process the data
as it arrives. Thus, an incremental learning algorithm must
make predictions in an on-line manner, typically updating the
hypothesis throughout the learning session. While many
researchers have studied incremental learning systems, until
recently, little effort was spent identifying and studying issues
specific to incremental learning. For example:
- How do the complexity of learning (cost of updating the hypothesis)
and the improvement in the accuracy of the hypothesis change during
the course of the learning session?
- How many observations are required to obtain a "stable" hypothesis?
- How does data ordering affect learning?
- How can concept drift be handled?
In addition to their practical implications, exploring these issues may
uncover important connections with other fundamental issues in Machine
Learning. For example, recent works suggest that deep relationships
exist between data ordering, bias, and theoretical concepts such as
teachability and information content.
Accordingly, the focus of this symposium will not be on implementations,
but on the general properties and common issues addressed by the many
separate research efforts in the area of Incremental Learning. The
symposium will provide an opportunity for researchers interested in this
area to share their results and hopefully in the process gain a better
understanding of the issues and techniques important to this growing
domain.
Incremental Learning and teaching strategies in particular may be
studied within various theoretical fields: machine learning theory,
information theory, non-monotonic reasoning, control theory, ... They
are also amenable to empirical investigations using symbolic machine
learning algorithms as well as neural networks and genetic algorithms.
In accordance, the symposium is open to contributions from these areas.
Topics of interest include but are not restricted to:
- The relationship between data ordering and bias.
- The necessary and sufficient conditions (on the learner and data)
to get order independence.
- The tractability of identifying a best order or teaching strategy.
- Valiant-style theoretical treatments of the quality of results as
a function of data or teaching strategy.
- The study of techniques (that may vary over time) to determine
which examples to ignore during the learning session.
The symposium will be organized in a format that combines research
presentations with open-ended discussions. It will begin with brief
introductory remarks posing a list of issues (prepared in advance by
the organizing committee) to be addressed during the symposium. Papers
will be organized into sessions according to the primary issues
addressed. We expect each session to have roughly three paper presentations.
For each session the committee will have selected a commentator familar
with the area who will lead a open-discussion session about issues raised
by the presentations and interesting research directions to build upon
the results presented. The commentator will be given copies of the paper
in advance, and may briefly summarize other work that is related to the
Submissions information :
Those interested in presenting their work should submit seven copies of
either a short paper (5-10 pages) describing completed work, or an
extended abstract (2-5 pages) describing preliminary work. Those
interested only in participating should submit a one page summary of
their research interest, together with a list of related publications.
Submissions should be sent to arrive by October 16, 1992 to :
Antoine Cornuejols
L.R.I.
Universite de Paris-Sud
Bat. 490
91405 ORSAY Cedex
FRANCE
antoine@lri.lri.fr
Program committee : Antoine Cornuejols (chair), Douglas Fisher,
Sally Goldman, Lorenza Saitta, Jeffrey Schlimmer.
------------------------------
Subject: Genetic Algorithms Conf. Call for Papers
Date: Sat, 08 Aug 92 10:54:17 -0600
From: Robert Elliott Smith <rob@comec4.mh.ua.EDU>
Call for Papers
ICGA-93
The Fifth International Conference on
Genetic Algorithms
17-22 July, 1993
University of Illinois at
Urbana-Champaign
The Fifth International Conference on Genetic Algorithms (ICGA-93),
will be held on July 17-22, 1993 at the Univ. of Illinois at
Urbana-Champaign. This meeting brings together an international
community from academia, government, and industry interested in
algorithms suggested by the evolutionary process of natural selection.
Topics of particular interest include: genetic algorithms and
classifier systems, evolution strategies, and other other forms of
evolutionary computation; machine learning and optimization using
these methods, their relations to other learning paradigms (e.g.,
neural networks and simulated annealing), and mathematical
descriptions of their behavior. Papers discussing how genetic
algorithms and classifier systems are related to biological and
cognitive systems are also encouraged.
Papers describing significant, unpublished research in this area are
solicited. Authors must submit four (4) complete copies of their
paper (hardcopy only), received by February 1, 1993, to the Program
Chair:
Stephanie Forrest
Dept. of Computer Science
University of New Mexico
Albuquerque, N.M. 87131-1386
Papers should be no longer than 10 pages, single-spaced, and printed
using 12 pt. type. Please include a separate title page with authors
names and addresses, and do not include these names in the paper's
body, to allow for anonymous peer review. The title page should also
contain a short abstract. Electronic submissions will not be
accepted.
Evaluation criteria include the significance
of results, originality, and the clarity and quality of the
presentation. Questions on the conference program and submission
should be directed to icga93@unmvax.cs.unm.edu. Other questions
should be directed to rob@comec4.mh.ua.edu.
Important Dates:
February 1, 1993: Submissions must be received
April 7, 1993: Notification to authors mailed
May 7, 1993: Revised, final camera-ready paper due
July 17-22, 1993: Conference dates
ICGA-93 Conference Committee:
Conference Co-Chairs: David E. Goldberg, Univ. of Illinois
at Urbana-Champaign
J. David Schaffer, Philips Labs
Publicity: Robert E. Smith, Univ. of Alabama
Program Chair: Stephanie Forrest, Univ. of New Mexico
Financial Chair: Larry J. Eshelman, Philips Labs
Local Arrangements: David E. Goldberg, Univ. of Illinois
at Urbana-Champaign
------------------------------
Subject: AAAI-93 Call for Papers
Date: Mon, 17 Aug 92 13:40:23 -0700
Call for Papers
AAAI-93
AAAI-93 is the eleventh national conference. The purpose
of the conference is to promote research in artificial
intelligence (AI) and scientific interchange among AI
researchers and practitioners.
Papers may represent significant contributions to all
aspects of AI:
a) the principles underlying cognition, perception, and
action in humans and machines;
b) the design, application, and evaluation of AI
algorithms and intelligent systems; and
c) the analysis of tasks and domains in which
intelligent systems perform.
In recognition of the wide range of methodologies and
research activities legitimately associated with AI, we
invite authors to submit papers describing both
experimental and theoretical results from all stages of
AI research. In particular, we encourage submission of
papers that present promising research directions by
describing innovative concepts, techniques, perspectives,
or observations that are not yet supported by mature
results. To be accepted, such submissions must include
substantial analysis of the ideas, the technology needed
to realize them, and their potential impact. In addition,
because of the essential interdisciplinary nature of AI
and the need to maintain effective communication across
sub-specialties, we encourage authors to position and
motivate their work in the larger context of the general
AI community. While papers concerned with applications
of AI are invited, those that describe working
commercial systems should be submitted to the IAAI
conference.
Requirements for Submission
Authors must submit six (6) complete printed copies of
their papers to the AAAI office by January 13, 1993.
Papers received after that date will be returned
unopened. Notification of receipt will be mailed to the
first author (or designated author) soon after receipt.
All inquiries regarding lost papers must be made by
January 27, 1993. Authors are also requested to send
their paperUs title page in an electronic mail message to
abstract@aaai.org by January 13, 1993. Notification of
acceptance or rejection of submitted papers will be
mailed to the first author (or designated author) by
March 3, 1993. Camera-ready copy of accepted papers
will be due about one month later.
Paper Format for Review
All six (6) copies of a submitted paper must be clearly
legible. Neither computer files nor fax submissions are
acceptable. Submissions must be printed on 8 1/2" x 11"
or A4 paper using 12 point type (10 characters per inch
for typewriters). Each page must have a maximum of 38
lines and an average of 75 characters per line
(corresponding to the LaTeX article-style, 12 point).
Double-sided printing is strongly encouraged.
Length
The body of submitted papers must be at most 11 pages,
including figures, tables, and diagrams, but excluding the
title page and bibliography. Papers exceeding the
specified length and formatting requirements are subject
to rejection without review.
Title page
Each copy of the paper must have a title page (separate
from the body of the paper) containing the title of the
paper, the names and addresses of all authors, a short
(less than 200 word) abstract, and a descriptive content
area or areas. The title page sent via electronic mail to
the AAAI office must be in plain ASCII text with each
section of the title page preceded by the name of that
section as follows:
title: <title>
author: <name of first author>
address: <address of first author> author: <name of last
author>
address: <address of last author>
abstract: <abstract>
content areas: <first area>, I,
<last area>
To facilitate the reviewing process, authors are
requested to select appropriate content areas from the
list below. Authors are invited to add additional content
area descriptors to their title page as needed.
Artificial Life, Automated Reasoning, Behavior-Based
Control, Belief Revision, Case-Based Reasoning,
Cognitive Modeling, Common Sense Reasoning,
Communication and Cooperation, Constraint-Based
Reasoning, Computer-Aided Education, Connectionist
Models, Corpus-Based Language Analysis, Deduction,
Diagnosis, Discourse Analysis, Distributed Problem
Solving, Expert Systems, Geometrical Reasoning,
Information Extraction, Knowledge Acquisition,
Knowledge Representation, Knowledge Sharing
Technology, Large Scale Knowledge Engineering,
Learning/Adaptation, Machine Learning, Machine
Translation, Mathematical Foundations, Multi-Agent
Planning, Natural Language Processing, Neural Networks,
Nonmonotonic Reasoning, Perception, Planning,
Probabilistic Reasoning, Qualitative Reasoning,
Reasoning about Action, Reasoning about Physical
Systems, Reactivity, Robot Navigation, Robotics, Rule-
Based Reasoning, Scheduling, Search, Sensor
Interpretation, Sensory Fusion/Fission, Simulation,
Situated Cognition, Spatial Reasoning, Speech
Recognition, System Architectures, Temporal Reasoning,
Terminological Reasoning, Theorem Proving, Truth
Maintenance, User Interfaces, Virtual Reality, Vision, 3-
D Model Acquisition.
Submissions to Multiple Conferences
Papers that are being submitted to other conferences,
whether verbatim or in essence, must state this fact on
the title page. If a paper appears at another conference
(with the exception of specialized workshops), it must
be withdrawn from AAAI-93. Papers that violate these
requirements are subject to rejection without review.
Review Criteria
Each paper will be carefully reviewed by experts
specializing in the content areas on the paperUs title
page. Questions that will appear on the review form have
been reproduced below. Authors are advised to bear
these questions in mind while writing their papers:
Significance
How important is the work reported? Does it attack an
important/difficult problem or a peripheral/simple one?
Does the approach offered advance the state of the art?
Originality
Has this or similar work been previously reported? Are
the problems and approaches completely new? Is this a
novel combination of familiar techniques? Does the
paper point out differences from related research? Is it
re-inventing the wheel using new terminology?
Quality
Is the paper technically sound? Does it carefully
evaluate the strengths and limitations of its
contribution? How are its claims backed up?
Clarity
Is the paper clearly written? Does it motivate the
research? Does it describe the inputs, outputs and basic
algorithms employed? Does the paper describe previous
work? Are the results described and evaluated? Is the
paper organized in a logical fashion?
Publication
Accepted papers will be allocated six (6) pages in the
conference proceedings. Up to two (2) additional pages
may be used at a cost to the authors of $250 per page.
Papers exceeding eight (8) pages and those violating the
instructions to authors will not be included in the
proceedings.
Copyright
Authors will be required to transfer copyright of their
paper to AAAI.
Please send papers and conference registration inquiries
to:
AAAI-93
American Association
for Artificial Intelligence
445 Burgess Drive
Menlo Park, CA 94025-3496
Registration and call clarification inquiries (ONLY) may
be sent to the CSNET address: NCAI@aaai.org. Please
send program suggestions and inquiries to:
Richard Fikes
Knowledge Systems Laboratory
Stanford University
701 Welch Road, Building C
Palo Alto, CA 94304
fikes@ksl.stanford.edu
Wendy Lehnert
Department of Computer Science
University of Massachusetts
Amherst, MA 01003
lehnert@cs.umass.edu
------------------------------
Date: Tue, 1 Sep 1992 08:51:41 -0700
From: Bob Levinson <levinson@cse.ucsc.EDU>
Subject: INTERNATIONAL CONFERENCE ON CONCEPTUAL STRUCTURES
CALL FOR PAPERS
INTERNATIONAL CONFERENCE ON CONCEPTUAL STRUCTURES
THEORY AND APPLICATIONS
August 4-7 1993 Quebec City, Canada
*************************************************************************
Sponsored by
IBM Canada (requested), Microsoft (requested)
UNISYS Corporation (requested) Sun Systems (requested)
l'Universite Laval, Quebec, Canada Butterworth Heinemann Ltd.
In cooperation with
AAAI American Association for Artificial Intelligence
ACM Association for Computing Machinery (requested)
CEFRIO Centre francophone de recherche en informatisation des organisations
CRIM Le Centre de recherche informatique de Montreal
IEEE Computer Society (requested)
GIRICO Le Groupe de recherche en informatisation des organisations
CSSCI The Canadian Society for Computational Studies of Intelligence
*************************************************************************
Conference Themes
Two fields are devoted to the study and development of knowledge-based
systems (KBS): artificial intelligence and cognitive science. Over the past 25
years, researchers have proposed several approaches for modelling knowledge
in KBS, including several kinds of formalisms: semantic networks, frames, logics
etc.
In the early eighties, John F. Sowa introduced the Conceptual Graph (CG) theory
which provides a knowledge representation framework consisting of a form of
logic with a graph notation, and which integrates several features from semantic
net and frame representations. Since that time, several research teams over the
world have been working on the application and on the extension of CG theory in
various domains ranging from natural language processing to database
modelling and machine learning. This international conference follows a series
of seven annual workshops and aims at providing an active forum for researchers
and practitioners to exchange ideas about the theory and application of
conceptual graphs. It is also opened to researchers proposing alternative
knowledge representation approaches, provided that they compare them to CG
theory.
Subjects may include, but are not limited to the following topics.
Theory Foundations
. Knowledge representation using CG . Operations on CG
. Logic systems using CG . Enhancements to CG theory
. Modality and truth maintenance . Reasoning using CG
Natural Language Processing (NLP)
. Semantic representations based on CG . Pragmatics
. NLP systems using CG
Methods and Tools
. Methodologies for knowledge acquisition . Learning approaches using CG
. Database conceptual modelling using CG . Expert systems and CG
. gRAph manipulation systems for CG . Abstract machines for CG
Alternative Approaches
. Comparing CG and approaches such as KL-ONE and other
. CG compared with various logics
. Alternative cognitive approaches for knowledge representation and
manipulation
Standardization
. Knowledge representation standardization efforts based on CG
Applications of all Kinds
*************************************************************************
Proceedings
Two types of papers will be considered: long papers (up to 5000 words) to be
published by Morgan Kaufmann in a book which will be available at the
conference; short papers (up to 2000 words) to be included in a supplementary
notebook also available at the conference.
*************************************************************************
Important dates
author's submission December 1 1992
notification of acceptance February 15 1993
camera-ready final papers April 1 1993
*************************************************************************
Instructions for authors
Authors are invited to submit five copies of their papers, not exceeding 5000
words (long papers), or 2000 words (short papers), double spaced, written in
English, including a page with: title of paper, author's name and address, phone
and fax numbers, a 20 line abstract, a list of keywords.
Submitted papers should reach the program committee chairmen before
December 1, 1992 at the following address :
Guy Mineau / Bernard Moulin
ICCS'93 Conference
Laval University, Computer Science Department
Pavillon Pouliot
Ste-Foy, Quebec, G1K 7P4 Canada
fax : 1 - 418 - 656 2324
*************************************************************************
General Chairman: John F. Sowa, IBM Systems Research Institute (USA)
Program Committee and Organizing co-Chairmen:
Guy Mineau, Bernard Moulin, Universite Laval (Canada)
Program Committee
Jerrold Aronson SUNY at Binghamton (USA)
Nick Cercone Simon Fraser University (Canada)
Peter Creasy University of Queensland (Australia)
Veronica Dahl Simon Fraser University (Canada)
Peter Eklund Adelaide University (Australia)
Gerard Ellis University of Queensland (Australia)
John Esch Paramax (USA)
Jean Fargues IBM Paris (France)
Norman Foo University of Sydney (Australia)
Carl Frederiksen McGill University (Canada)
Brian Gaines University of Calgary (Canada)
Roger Hartley New Mexico State University (USA)
Martin Janta CMR, St Jean (Canada)
Pavel Kocura Loughborough Univ. (England)
Debbie Leishman Hughes (Canada)
George Lendaris Portland State University (USA)
Robert Levinson Univ. of Calif. at Santa Cruz (USA)
Robert Meersman Tilburg University (The Netherlands)
Sung Myaeng Syracuse University (USA)
Tim Nagle UNISYS Corp. (USA)
Heather Pfeiffer New Mexico State University (USA)
Gerard Sabah LIMSI-CNRS (France)
Doug Skuce University of Ottawa (Canada)
James Slagle University of Minnesota (USA)
Eric Tsui University of Sydney (Australia)
Paola Velardi University of Ancona (Italy)
Eileen Way SUNY at Binghamton (USA)
Yelena Yesha Univ. of Maryland Baltimore (USA)
Michael Zock LIMSI-CNRS (France)
****************************************************************************
About the conference location
Longtime national capital under the French and English regimes, Quebec City
has preserved this role at the provincial level. The oldest city in North America, it
offers an interesting blend of early and modern architecture. It is the only fortified
city north of Mexico. Its impressive walls, originally designed to block access to
the city, today invite visitors to enjoy the subtle harmony of various architectural
styles. Because Le Vieux Quebec is a unique site that has attracted world
attention, it was the first North American city to be included on UNESCO's
prestigious world heritage list. Perched atop cap Diamant, the historic district
overlooks the St. Lawrence River and offers numerous breathtaking views of the
South Shore, lile d'Orleans, the Lower city and the Laurentians. It is a charming
piece of Europe in North America.
****************************************************************************
Information form
Attendance to the conference will be limited. If you wish to receive the early
announcement of the conference program, please fill in this form and send it at
the following address
Guy Mineau / Bernard Moulin
Conference ICCS'93
Laval University, Computer Science Department
Pavillon Pouliot
Ste-Foy, Quebec, G1K 7P4 , Canada
Email: MINEAU@VM1.ULAVAL.CA MOULIN@VM1.ULAVAL.CA
Name:
Affiliation:
Address:
City:
Zip code:
Country:
Telephone number:
Fax number:
------------------------------
End of ML-LIST 4.17 (Digest format)
****************************************