Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 104
AIList Digest Monday, 28 Apr 1986 Volume 4 : Issue 104
Today's Topics:
Seminars - Learning Representation by Backpropagation (GTE) &
Distributed Object-Oriented Programming (CMU) &
Prolog and Geometry (CSLI) &
Inferring Domain Plans in Question Answering (UPenn) &
Possible Worlds Planning (UPenn) &
Parallel Algorithms for Term Matching (MIT),
Conference - 19TH Hawaii Int. Conf. on Systems Sciences &
1st Australian AI Congress (Extended Deadline)
----------------------------------------------------------------------
Date: Thu, 24 Apr 86 10:08:52 EST
From: Bernard Silver <SILVER@AI.AI.MIT.EDU>
Subject: Seminar - Learning Representation by Backpropagation (GTE)
GTE Laboratories Inc
Machine Learning Seminar Series
Speaker: David E. Rumelhart
Institute of Cognitive Science
and
University of California San Diego
Title: Learning Representation by Backpropagation
Date: Monday April 28 9am
Place: GTE Laboratories
40 Sylvan Rd
Waltham MA 02254
Recent work will be presented on using the backpropagation learning
procedure for developing internal representations in parallel
distributed processing systems. The talk will include a brief
introduction to the backpropagation learning procedure followed by a
report on a number of new applications of the procedure to storage of
semantic information and the discovery of new features.
Visitors welcome: Contact Oliver Selfridge (617) 466-2855
or Bernard Silver (617) 466-2663
------------------------------
Date: 23 Apr 86 11:49:56 EST
From: David.Anderson@SPICE.CS.CMU.EDU
Subject: Seminar - Distributed Object-Oriented Programming (CMU)
Thesis Proposal:
Object Oriented Programming for Distributed Systems
David B. Anderson
Computer Science Department
Carnegie-Mellon University
dba@k.cs.cmu.edu
28 April 1986
3:30 pm
WeH 5409
ABSTRACT
Object oriented programming has often been advocated for a variety of
programming tasks, particularly interactive, graphical applications and
window managers. Software engineers are attracted to this
programming methodology because of the modularity, data abstraction and
information hiding that it promotes. On the other hand, object oriented
techniques have not generally been used in building distributed
systems and applications.
The difficulty in using object oriented programming techniques for
implementing distributed applications lies in the requirements that
object oriented languages and systems place on their runtime
environment. For example, the remote procedure call mechanisms
typically used in building distributed applications must be replaced with
a mechanism for remote method invocation. This means that a static
remote procedure call stub generator, such as Matchmaker, must be replaced
with a mechanism for dynamically locating the correct method to call
based on the runtime types of objects. Furthermore, mechanisms are needed
to allow objects, classes and methods to be created and destroyed as
the system is running. Other difficulties and issues that must be
addressed include the naming and scope of objects, garbage collection,
error recovery and protection.
The proposed dissertation research will develop a solution to these
problems in the form of an object manager for distributed systems.
This proposal looks at these issues in some detail, and discusses the
design of an object manager to meet these requirements. A prototype
system is planned, and will be used to implement a distributed,
object oriented user interface environment.
------------------------------
Date: Fri 25 Apr 86 19:31:48-PST
From: Fred Lakin <LAKIN@SU-CSLI.ARPA>
Subject: Seminar - Prolog and Geometry (CSLI)
Pixels and Predicates meeting: note ==> TUESDAY <==
PROLOG AND GEOMETRY
Who: Randolph Franklin, UC at Berkeley
wrf@degas.berkeley.edu
Where: CSLI trailers
When: 1:00pm - TUESDAY, April 29, 1986
Abstract:
The Prolog language is a useful tool for geometric and graphics
implementations because its primitives, such as unification,
match the requirements of many geometric algorithms. We have im-
plemented several problems in Prolog including a subset of the
Graphics Kernal Standard, convex hull finding, planar graph
traversal, recognizing groupings of objects, and boolean combina-
tions of polygons using multiple precision rational numbers.
Certain paradigms, or standard forms, of geometric programming in
Prolog are becoming evident. They include applying a function to
every element of a set, executing a procedure so long as a cer-
tain geometric pattern exists, and using unification to propagate
a transitive function. Certain strengths and weaknesses of Pro-
log for these applications are now apparent.
------------------------------
Date: Sun, 27 Apr 86 21:36 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Inferring Domain Plans in Question Answering (UPenn)
Forwarded From: Bonnie Webber <Bonnie@UPenn> on Sun 27 Apr 1986 at 9:40
INFERRING DOMAIN PLANS IN QUESTION-ANSWERING
Martha E. Pollack
Thesis Defense
Monday, April 28, 1986
Noon-2pm, 216 Moore
The importance of plan inference (PI) in models of conversation has
been widely noted in the computational-linguistics literature, and its
incorporation in question-answering systems has enabled a range of
cooperative behaviors. The PI process in each of these systems,
however, has assumed that the questioner (Q) whose plan is being
inferred and the respondent (R) who is drawing the inference have
identical beliefs about the actions in the domain. I demonstrate that
this assumption is too strong, and often results in failure not only
of the PI process, but also of the communicative process that PI is
meant to support. In particular, it precludes the principled
generation of appropriate responses to queries that arise from invalid
plans. I present a model of PI in conversation that distinguishes
between the beliefs of the questioner and the beliefs of the
respondent. This model rests on an account of plans as mental
phenomena: "having a plan" is analyzed as having a particular
configuration of beliefs and intentions. Judgements that a plan is
invalid are associated with particular discrepancies between the
beliefs that R ascribes to Q, when R believes Q has some particular
plan, and the beliefs R herself holds. I define several types of
invalidities from which a plan may suffer, relating each to a
particular type of belief discrepancy, and show that the types of any
invalidities judged to be present in the plan underlying a query can
affect the content of a cooperative response. The PI model has been
implemented in SPIRIT -- a System for Plan Inference that Reasons
about Invalidities Too -- which reasons about plans underlying queries
in the domain of computer mail.
Advisor: Bonnie Webber
Committee: Aravind Joshi, chair
Tim Finin
Barbara Grosz
------------------------------
Date: Sun, 27 Apr 86 23:35 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Possible Worlds Planning (UPenn)
Forwarded From: Bonnie Webber <Bonnie@UPenn> on Thu 24 Apr 1986 at 14:24
In addition to his talk on Tuesday afternoon, 29 April, on Multi-Valued
Logic, Matt Ginsberg will also give a talk on Wednesday morning at
10:30 on Possible Worlds Planning.
Room to be announced.
POSSIBLE WORLDS PLANNING
Matt Ginsberg
Stanford University
The size of the search space is perhaps the most intractable
of all of the problems facing a general-purpose planner. Some
planning methods (means-ends analysis being typical) address this
problem by encouraging the system designer to give the planner domain-specific
information (perhaps in the form of a difference table) to help govern
this search.
This paper presents a domain-independent approach to this problem
based on the examination of possible worlds in which the planning
goal has been achieved. Although a weak method, the ideas presented
lead to considerable savings in many examples; in addition, the natural
implementation of this approach has the attractive property that
incremental efforts in controlling the search provide incremental
improvements in performance. This is in contract to many other
approaches to the control of search or inference, which may require
large expenditures of effort before any benefits are realized.
------------------------------
Date: Thu 24 Apr 86 14:29:00-EST
From: Lisa F. Melcher <LISA@XX.LCS.MIT.EDU>
Subject: Seminar - Parallel Algorithms for Term Matching (MIT)
DATE: Thursday, May 1, 1986
TIME: 3:45 - Refreshments
4:00 - Lecture
PLACE: NE43 - 512A
"PARALLEL ALGORITHMS FOR TERM MATCHING"
CYNTHIA DWORK
IBM Almaden Research Center
San Jose, CA
Unification of terms is a well known problem with applications to a variety
of symbolic computation problems. Two terms s and t, involving function
symbols and variables, are unifiable if there is a substitution for the
variables which makes s and t syntactically identical. For example, f(x,x)
and f(g(y),g(g(c))) are unified by substituting g(c) for y and g(g(c)) for
x. A special case of unification is term matching where one of the terms
contains no variables. Previous work on parallel algorithms for unification
by Dwork, Kanellakis and Mitchell (DKM) showed that unification is P-complete
in general, even if terms are represented as trees so that common
subexpressions must be repeated. However, DKM give an NC2 algorithm for term
2
matching using M(n ) processors where M(m) is the number of operations needed
to multiply m-by-m matrices. This algorithm allows a compact dag
representation of terms. These resuts have been tightened in two ways.
First, the processor bound for term matching of dags has been improved to
2
M(n), while retaining the O(log n) running time, using a randomizing
algorithm. There is also some evidence that improving the processor bound
further will be difficult since there is an efficient parallel reduction from
the graph accessibility problem (GAP) to the term matching problem for dags,
2
so that any improvement in the processor bound for term matching (say, to n )
would imply the same for GAP. The second improvement is a sharper
P-completeness result which shows that unification of tree terms is
P-complete even for linear terms where each variable can appear at most once
in each term.
This is joint work with Paris Kanellakis and Larry Stockmeyer.
Shafi Goldwasser
Host
------------------------------
Date: 25 Apr 1986 10:09:56-EST
From: Vasant.Dhar@ISL1.RI.CMU.EDU
Subject: Conference - 19TH Hawaii Int. Conf. on Systems Sciences
CALL FOR PAPERS: 19TH HAWAII INTERNATIONAL CONFERENCE ON SYSTEMS SCIENCES
(HICSS), Hawaii, January 1987.
Papers are invited for the KNOWLEDGE-BASED/DECISION-SUPPORT SYSTEMS track.
The following are representative areas:
1. Knowledge-Based approaches to large systems development
2. Knowledge-Based support systems in business organizations
3. Knowledge Engineering in Management Science Applications
4. Knowledge Engineering in Database Management/Intelligent Retrieval Systems
5. Decision Support Systems for Group decision making
6. Model Management in Decision Support Systems
7. User Interfaces in Decision Support Systems
Papers falling into area 1 above should be sent to:
Vasant Dhar
Department of Information Systems
New York University
90 Trinity Place
New York, NY 10006.
Papers in all other areas should be sent to Edward Stohr at the same
address. The deadline for submission is July 7 1986. Authors will be notified
of acceptance before September 8, 1986. Camera ready copies are due on
October 20, 1986. The conference is on the island of Oahu, January 6-9, 1987.
------------------------------
Date: 23 Apr 86 13:26:07 +1000 (Wed)
From: "ERIC Y.H. TSUI" <decvax!mulga!aragorn.oz!eric@decwrl.DEC.COM>
Subject: Conference - 1st Australian AI Congress (Extended Deadline)
1
11 st
111 AUSTRALIAN
11 ARTIFICIAL
11 INTELLIGENCE
11 CONGRESS
11
1111 Melbourne, November 18-20, 1986
CALL FOR PAPERS
========================================================================
DEADLINE EXTENDED...DEADLINE EXTENDED...DEADLINE EXTENDED...DEADLINE EXT
========================================================================
Abstract (300 words) of papers to be selected for presentation
to the 1st Australian Artificial Intelligence Congress are now invited.
The three-part program comprises:
i) AI in Education
- Intelligent tutors
- Computer-managed learning
- Course developers environment
- Learning models
- Course authoring software
ii) Expert System Applications
- Deductive databases
- Conceptual schema
- Expert system shells (applications and limitations)
- Interactive knowledge base systems
- Knowledge engineering environments
- Automated knowledge acquisition
iii) Office Knowledge Bases
- Document classification and retrieval
- Publishing systems
- Knowledge source systems
- Decision support systems
- Office information systems
Tutorial presenters are also sought. Specialists are required
in the areas of:
- Common loops
- Natural language processing
- Inference engines
- Building knowledge databases
- Search strategies
- Heuristics for AI solving
Reply to:
ACSnet address: brian!aragorn.oz
CSNET address: brian@aragorn.oz
UUCP address: seismo!munnari!aragorn.oz!brian
decvax!mulga!aragorn.oz!brian
ARPA address: munnari!aragorn.oz!brian@seismo.arpa
decvax!mulga!aragorn.oz!brian@Berkeley
or post to: Dr. B. Garner, Division of Computing and Mathematics,
Deakin University, Victoria 3217, Australia.
NEW DEADLINES: All submissions by May 31, 1986. Notification by July 14.
^^^ ^^^^^^ ^^^^^^^
||| |||||| |||||||
Inquiries: Stephen Moore, Director, 1AAIC86, tel: (02)439-5133.
Eric Tsui eric@aragorn.oz
------------------------------
End of AIList Digest
********************
