Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 031
AIList Digest Thursday, 20 Feb 1986 Volume 4 : Issue 31
Today's Topics:
Seminars - Classical Conditioning and Contingency (SU) &
Learnability and the Vapnik-Chervonenkis Dimension (IBM-SJ) &
Hierarchical Reasoning, Simulation (UPenn) &
The Architecture of a Rational Agent (Edinburgh) &
Planning for Robotic Assembly Lines (USC) &
Distributed Problem Solving (USC) &
Adaptive Planning (MIT) &
Deductive and Relational Knowledge Bases (CCA),
Conference - Symbolics National Users Group Meeting
----------------------------------------------------------------------
Date: Wed, 19 Feb 86 18:34:28 pst
From: gluck@SU-PSYCH (Mark Gluck)
Subject: Seminar - Classical Conditioning and Contingency (SU)
The topic of this week's learning seminar will be on associative learning
in animals. We will examine classical conditioning, one of the simplest
and best studied forms of induction. The readings are:
Rescorla & Wagner (1972): Reviews the animal learning data and proposes
a simple linear model of associative learning which predicts
than animals will induce relative contingencies between
stimuli. The algorithm is formally equivalent to the
Widrow-Hoff predictor in adaptive systems and is a
special case of the delta rule used by the Rumelhart et
al. back-propogation algorithm.
The other two papers are two "Cognitive Science" models for classical
conditioning. The first, presented in the Holland et al book, is
a rule-based production system model of classical conditioning. The
second, by Sutton and Barto, is a connectionist/network model for
classical conditioning.
The seminar is in Building 360; Room 364 (near the geology corner).
On Thursday from 1:15-3pm.
------------------------------
Date: 19 Feb 86 14:53:44 PST
From: CALENDAR@IBM-SJ.ARPA
Subject: Seminar - Learnability and the Vapnik-Chervonenkis Dimension (IBM-SJ)
[Excerpted from the IBM Calendar by Laws@SRI-AI.]
IBM Almaden Research Center
650 Harry Road
San Jose, CA 95120-6099
Computer LEARNABILITY AND THE VAPNIK-CHERVONENKIS DIMENSION
Science D. Haussler, Department of Mathematics and
Seminar Computer Science, University of Denver
Fri., Feb. 28 The current emphasis on knowledge-based software has
10: 30 A.M. created a broader interest in algorithms that learn
B1-413 knowledge structures or concepts from positive and
negative examples. Using the learning model recently
proposed by Valiant, we have attempted to determine
which classes of concepts have efficient (i.e.,
polynomial time) learning algorithms. As noticed
earlier by Pearl and by Devroye and Wagner, a simple
combinatorial property of concept classes, the
Vapnik-Chervonenkis dimension, plays an important
role in learning and pattern recognition. We clarify
the relationship between this property and Valiant's
theory of learnability. Our results lead to the design
of efficient learning algorithms that employ a
variant of Occam's Razor. Illustrations are given
for certain classes of conjunctive concepts and for
concepts that are defined by various types of regions
in feature space. The work reported was done jointly
with Anselm Blumer, Andrzej Ehrenfeucht and
Manfred Warmuth of the Universities of Denver,
Colorado and California at Santa Cruz, respectively.
Host: B. Simons
[A BATS announcement said that the seminar would be at 11:00. - KIL]
------------------------------
Date: Mon, 17 Feb 86 00:56 EST
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Hierarchical Reasoning, Simulation (UPenn)
Forwarded From: Paul Fishwick <Fishwick@UPenn> on Sun 16 Feb 1986 at 12:54
HIERARCHICAL REASONING:
SIMULATING COMPLEX PROCESSES
OVER MULTIPLE LEVELS OF ABSTRACTION
Paul A. Fishwick
University of Pennsylvania
Ph.D. Defense
This talk describes a method for simulating processes over multiple levels of
abstraction. There has been recent work with respect to data, object, and
problem-solving abstraction, however, abstraction in simulation has not been
adequately explored. We define a process as a hierarchy of distinct production
rule sets that interface to each other so that abstraction levels may be
bridged where desired. In this way, the process may be studied at abstraction
levels that are appropriate for the specific task: notions of qualitative and
quantitative reasoning are integrated to form a complete process description.
The advantages to such a description are increased control, computational
efficiency and selective reporting of simulation results. Within the framework
of hierarchical reasoning, we will concentrate on presenting the primary
concept of process abstraction.
A Common Lisp implementation of the hierarchical reasoning theory called HIRES
is presented. HIRES allows the user to reason in a hierarchical fashion by
relating certain facets of the simulation to levels of abstraction specified in
terms of actions, objects, reports, and time. The user is free to reason about
a process over multiple levels by weaving through the levels either manually or
via automatically controlled specifications. Capabilities exist in HIRES to
facilitate the creation of graph-based abstraction levels. For instance, the
analyst can create continuous system models (CSMP), petri net models, scripts,
or generic graph models that define the process model at a given level. We
present a four-level elevator system and a two-level "dining philosophers"
simulation. The dining philosophers simulation includes a 3-D animation of
human body models.
Time: Wednesday, February 26, 3pm
Place: Moore School, Room 554
Committee:
Dr. Norman Badler (Adviser)
Dr. Timothy Finin (Chairman)
Dr. Insup Lee
Dr. Richard Paul
------------------------------
Date: Tue, 18 Feb 86 17:55:35 GMT
From: Gideon Sahar <gideon%edai.edinburgh.ac.uk@cs.ucl.ac.uk>
Subject: Seminar - The Architecture of a Rational Agent (Edinburgh)
EDINBURGH AI SEMINARS
Date: Wednesday, 19th February l9986
Time: 2.00 p.m.
Place: Department of Artificial Intelligence
Seminar Room
Forrest Hill
EDINBURGH.
Dr. Robert C. Moore, Computer Laboratory, University of Cambridge
(visiting from SRI International) will give a seminar entitled -
"The Architecture of a Rational Agent".
The ultimate goal of artificial intelligence is to build complete,
autonomous, artificial rational agents. Most research in AI focuses
on one or another component of such an agent: the vision sybsystem,
reasoning subsystem, language subsystem, etc. Recently, however, some
attention has begun to be paid to the over-all architecture in which
these subsystems are combined. The first half of this talk will
discuss how concern for the architecture of rational agents is motivated
by the need to treat language as a form of rational action, and how
this view of language provides a formal framework for treating phenomena
that have been argued to be beyond the scope of formal analysis. In
the second half of the talk, we will compare the three component
belief/desire/intention model of rational agency typically used in AI
to the two component model cannot satisfactorily account for
cooperation among rational agents, proving a theorem to the effect that
there are situations in which there is no strategy for a group of
two-component agents that is rational by the normal standards of
decision theory.
------------------------------
Date: 18 Feb 1986 13:50-EST
From: gasser@usc-cse.arpa
Subject: Seminar - Planning for Robotic Assembly Lines (USC)
USC DPS GROUP MEETING
Wednesday, 2/26/86
3:00 - 5:00 PM
Seaver Science Bldg. 319
Dong Xia (Ph.D. Student, USC) will speak on "An Approach To Planning and
Scheduling for Robotic Assembly Lines"
While extensive studies have been devoted to general robot problem solving
and planning techniques in artifical world in recent years, the progress
towards their practical applications in robotic manufacturing floor has
severely prohibiited by the lack of sound understanding of the assembly
process and an adequate method to deal with real time uncertainties. In
this talk, we are going to address two of the most fundamental and
interrelated problems, namely task planning and temporal action scheduling.
We study these problems in the context of multiple cooperative robots with
assumed perceptual capabilities which work in a highly shared and dynamic
mechanical environment in a coordinated fashion for a common or different
goal(s). In this presentation, a general system architecture and a hybrid
knowledge representation scheme for a class of assembly lines is proposed
and some important design concepts and problems of robot-based intelligent
assembly lines are identified and discussed. Finally a particular prototype
system, called Miniassembler, is given, which exhibits our concepts and
methods to cope with temporal uncertainty.
Questions: Dr. Les Gasser, USC (213) 743-7794
or Dong Xia (XIA@USC-CSE.ARPA).
------------------------------
Date: 18 Feb 1986 15:34-PST
From: gasser@usc-cse.arpa
Subject: Seminar - Distributed Problem Solving (USC)
USC DPS GROUP MEETING
Wednesday, 2/19/86
3:00 - 4:00 PM
Seaver 319
Tom Hinke will speak on "Distributed Problem Solving and Architectural
Design".
The talk will cover some preliminary ideas about the application of
distributed problem solving techniques to the domain of computer aided
architectural design. The talk will include a brief overview of caad
work to date, a concept of how DPS could be applied to this area, and a
brief discussion of some of the anticipated problems in applying DPS to
design. The talk is based on very preliminary work in the area and
should be viewed as a forum to generate some initial comments and
direction for the bulk of the research which lies ahead.
Questions: Dr. Les Gasser, (213) 743-7794, or
Tom Hinke: HINKE@USC-CSE.ARPA
------------------------------
Date: Tue, 18 Feb 1986 17:08 EST
From: David Chapman <ZVONA%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU>
Subject: Seminar - Adaptive Planning (MIT)
[Forwarded from the MIT bboard by SASW@MIT-MC.]
Wednesday, February 26 3:00pm Room: NE43- 8th floor Playroom
The Artificial Intelligence Lab
Seminar
"Adaptive Planning"
Richard Alterman
UC Berkeley
Consider the case where a planner intends to transfer airplanes. A
common-sense approach to the problem of transferring airplanes would
be to try to re-use an old existing plan: exit first airplane via
arrival gate, determine departure gate, walk to the departure gate,
and board second airplane via departure gate. In a small airport this
would work just fine. But in a larger airport, say Kennedy Airport
where there is more than one terminal, if the arrival and departure
gates were in different terminals, the plan would have to be modified
(i.e. the planner would have to take a shuttle between terminals).
The problem of adaptive planning is to refit old plans to novel
circumstances. In the case of the example above, an adaptive planner
would refit the old plan for transferring airplanes to the novel
circumstances at the Kennedy Airport. The importance of adaptive
planning is that it adds a dimension of flexibility to the
common-sense planner.
Key elements in the theory of adaptive planning are its treatment of
background knowledge and the introduction of a notion of planning by
situation matching. The talk will motivate and discuss four kinds of
background knowledge. It will describe a number of kinds of situation
difference that can occur between an old plan and the new planning
situation. It will discuss situation matching techniques that are
based on the interaction of the planner's current circumstances and
its background knowledge. An important theme throughout this
discussion will be the control of access to knowledge.
------------------------------
Date: Tue 18 Feb 86 15:40:17-EST
From: Sunil Sarin <SKS@XX.LCS.MIT.EDU>
Subject: Seminars - Deductive and Relational Knowledge Bases (CCA)
[Forwarded from the MIT bboard by SASW@MIT-MC.]
CCA Colloquium Series
DATE: February 20, 1986-- Thursday
TIME: 10:00-11:00 a.m.
PLACE: 4th floor large conference room, Four Cambridge Center
TITLE: Deductive Databases and a Relational Knowledge Base
A Survey of Work at ICOT, Japan
SPEAKERS: Haruo Yokota and Masaki Murakami (Institute for New
Generation Computer Technology (ICOT--Japan) )
CCA (Computer Corporation of America) is located at Four Cambridge
Center, which is on Broadway, behind Legal Seafood. Tell the
security desk you are visiting CCA and they will send you up to
CCA on the 5th floor. Tell CCA's receptionist to call Barbara
Wong who will show you where the seminar is. (If you can't
remember that, simply say you're here for the colloquium.)
Abstracts of works to be covered:
1. Deductive Database System based on Unit Resolution
by Haruo Yokota, Ko Sakai, Hidenori Itoh
This paper presents a methodology for constructing a deductive
database system consisting of an intensional processor and a
relational database management system. A setting evaluation
is introduced. The intensional processor derives a setting
from the intensional database and a given goal and sends the
setting and the relationship between setting elements to the
management system. The management system performs a unit
resolution with setting using relational operations for the
extensional databases. An extended least fixed point operation
is introduced to terminate all types of recursive queries.
2. A Model and an Architecture for a Relational Knowledge Base
by Hauro Yokota, Hidenori Itoh
A relational knowledge base model and an architecture which
manipulates the model are presented. An item stored in the
relational knowledge base is a term, and it is retrieved by
unification operation between the terms. The relational
knowledge base architecture we propose consists of a number
of unification engines, several disk systems, a control processor,
and a multiport page-memory. The system has a knowledge compiler
to support a variety of knowledge representations.
3. Formal Semantics of a Relational Knowledge Base
by Masaki Murakami, Hauro Yokota, Hidenori Itoh
A mathematical foundations for formal semantics of term relations
[Yokota et al. 85] is presented. A term relation is a basic data
structure of a relational knowledge base. It is an enhanced version
of relational model in a database theory. It may include syntactically
complex structures such as terms or literals containing variables as
items of relations. The items are retrieved with operations called
retrieval-by-unification. We introduce as a semantic domain of
n-ary-term relations n_T_RELATIONS and define a partial order on them.
We characterize retrieval-by-unification as operations on n_T_RELATIONS
with monotone functions and greatest lower bounds.
------------------------------
Date: Wed, 19 Feb 86 13:50:19 pst
From: grover@aids-unix (Mark Grover)
Subject: Conference - Symbolics National Users Group Meeting
[Submitted to AILIST because: 1) it is an initial announcement. Followup will
take place via the address provided. 2) Symbolics computers are a major tool
of AI researchers. 3) The majority of work on Symbolics computers is related
to AI. 4) Users are widespread: well over 100 sites and 1500 machines.]
Are you getting bored with TV:MENU-CHOOSE?
Do you know your FOSS from your CSE?
Are you ready for Release 7 and Common Lisp?
No matter what your answers, you are invited to the
Second Annual
SYMBOLICS NATIONAL USERS GROUP MEETING
(SNUG86)
Georgetown University Campus
Washington, DC
June 2-6, 1986
(organized by the Capital Area SLUG)
with...
Speakers Poster Sessions War Stories
Panels Discussions Tutorials
Debates Wizards BOFs
The SLUG National Board has approved plans from the Capital Area SLUG to hold
a five-day National Symposium (SNUG86). This year's Symposium will consist
of three days of meetings, preceeded by two days of special Symbolics
Educational Services Tutorials at a small additional cost per session.
Planning is well underway to build on the experience of last year's National
SLUG Symposium in San Francisco. This year we hope for an even more exciting
gathering at the beautiful Georgetown University campus on the Potomac River
to discuss, debate and learn the best in Lisp Machine techniques.
This year's theme:
"Programming in Style on the Symbolics"
The goal of this year's Symposium is to make explicit the experience of
long-time users in terms of programming style. There are so many ways of
achieving a particular function, but which are the most efficient, elegant
and consistent? This Symposium is a means to share such important
information, where common needs and individual problems can be addressed.
Registration costs (separate from tutorials) will be considerably less than
comparable meetings. Inexpensive campus housing will be available. A
detailed announcement will be forthcoming.
RESPONSE DEADLINE IS MARCH 28, 1986
It is essential that the Symposium planning committee hear from you in order
to gauge interest. To receive future announcements, you must fill in a
response form BY MARCH 28 to the mailing address below.
We also also invite program suggestions. Please address program-related
correspondence to ATTN: Programs, or via ARPAnet mail to the Program Chair,
Mark Grover (Advanced Decision Systems), at GROVER@AIDS-UNIX.ARPA (or
Grover@AIDS-DC.Dialnet.Symbolics.COM). This address is for technical program
session proposals only! Questions regarding registration, facilities and
exhibits should be directed to address and phone below.
Planned Activities
Monday and Tuesday: Tutorials taught by Symbolics personnel to include
Introduction to Lisp Machine Programming, Site Maintenance, Common Lisp and
advanced topics. Tuesday evening: Third Party Vendor Hospitality suites.
Wednesday: Keynote presentations, concentrating on Release 7 and SLUG
activities such as the national library.
Thursday and Friday: program sessions to include Windows and Processes;
Flavors; Of Mice and Menus; Large Scale Data Management; Networking; File
Storage for Lisp Objects; Group Programming Etiquette; Security; and
Personalizing Your Environment. Many other topics are under consideration.
Please make additional suggestions of session proposals on the form below.
Poster sessions will be held in parallel with the program sessions. A poster
session allows a user to display charts and code on a fixed display in shared
quarters while interested attendees are free to move about, listen to and
discuss these informal talks. Further, there will be free time for Birds of
a Feather (BOF) gatherings. We hope to provide some Lisp Machine time for
these sessions.
This is a USER-oriented meeting. The informal availability of Symbolics
"wizards" was a significant attraction of last year's Symposium which will be
repeated.
Conference Location
Located near the Potomac River and Rock Creek Park, the Georgetown area of
Washington DC is well-known for its many shops and restaurants. Georgetown
University provides excellent meeting facilities and inexpensive
accommodations. The many monuments and museums of Washington are within short
rides via bus or metro.
SNUG86 MAILING LIST
(Mail this form to the address below. No ARPA mail please).
Ms. Annmarie Pittman
SNUG86
655 15th Street NW #300
Washington, DC 20005
(202) 639-4228
First Name: Last Name:
Title:
Organization:
Address:
City: State: Zip Code:
Telephone:
_____ Please add me to the mailing list.
_____ I plan to attend SNUG86.
_____ I would be interested in attending Symbolics Education Services
one-day tutorials on the subjects(s):
_____ I would like to propose sessions on the subject(s):
_____ I would be interested in giving a poster session on the topic:
_____ My company is interested in exhibiting at the conference. Please
send exhibit materials.
------------------------------
End of AIList Digest
********************
