Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 159
AIList Digest Tuesday, 1 Jul 1986 Volume 4 : Issue 159
Today's Topics:
Seminars - Chunking and XAPS3 (Rutgers) &
Advanced Planning Systems (Rutgers) &
Real-Time Inferencing with Adaptive Logic Networks (NASA) &
Overview of the MENTOR System (CMU),
Conference - ACM Conference on Office Information Systems
----------------------------------------------------------------------
Date: 23 Jun 86 10:51:24 EDT
From: Tom Fawcett <FAWCETT@RED.RUTGERS.EDU>
Subject: Seminar - Chunking and XAPS3 (Rutgers)
The summer machine learning discussion group meets Tuesdays at 11 in
room 423. This week John Bresina will give a talk on "Chunking and
XAPS3". The abstract follows. [...]
In this talk I discuss the chunking theory of learning, and in
particular how this theory is realized in the XAPS3 production system
architecture. The talk is based on Paul S. Rosenbloom's Ph.D. thesis,
"The Chunking of Goal Hierarchies: A Model of Practice and
Stimulus-Response Compatibility" [Carnegie-Mellon, 1983], for which
Allen Newell was the advisor.
First the chunking theory of learning is described and the desired
behavioral aspects of a chunking mechanism are summarized. I then
present the architectural constraints that an implementation must
satisfy in order to exhibit this desired behavior. Next the XAPS3
production system architecture is described, followed by a detailed
look at the implementation of the chunking theory within XAPS3. In
conclusion I present a brief critique of this implementation as well
as some suggestions for extending and improving it.
------------------------------
Date: 23 Jun 86 14:43:33 EDT
From: Smadar <KEDAR-CABELLI@RED.RUTGERS.EDU>
Subject: Seminar - Advanced Planning Systems (Rutgers)
III SEMINAR
Title: Advanced Planning Systems
Speaker: Chitoor V. Srinivasan
Date: Friday, June 27, 2:50 PM
Place: Hill Center, Room 705
Dr. Srinivasan, a professor in our department, will present his current
research in an informal talk. Here is his abstract:
A new planning technique for planning in "dynamic worlds" is
introduced in this talk. It develops plans using a method of
approximate reasoning and plan refinements over abstraction spaces,
and is based on a formalization of the problem solving approach which
Navy planners use to design Naval Operational Plans.
A dynamic world is one in which changes occur not only in the
properties associated with the objects that exist in the world, but
the set of objects existing in the world itself may change. As the
world changes some objects may get destroyed and others may get newly
created. It is a world in which reasoning about multiple actions
occuring simultaneously over intervals of time is necessary to do
planning. Also, knowledge needed to do planning in such worlds may be
only incompletely known. Existing planning systems do not consider
worlds of this kind.
In the new planning technique plans are viewed as hierarchies of
"behaviors" to be realized by actions that occur in a world.
Behaviors are properties (usually dynamic ones), which (a). remain
invariant while worlds themselves change as a result of actions
occurring in them, and (b). are needed for the success of one or more
of those actions, or are intrinsic properties of the worlds
themselves. Of course, a given behavior may be the result of several
actions occurring simultaneously. Thus for example, "an object will
continue to move in a straight line, unless disturbed by force" is a
general behavior of movements which is an intrinsic property of the
world we live in. "Goods transported will eventually appear in
neighborhoods progressively closer to destination" is a general
behavior of transportation actions.
This concept of behavior is formally defined here and a formal
action language is introduced to describe actions in terms of
"[preconditions, behaviors, functions]." It gives rise to a new
"modal action calculus" which is quite different from both "situation
calculus" and calculus of "dynamic logic." It is shown how this
concept of \fIbehavior\fR makes it possible to develop plans in
dynamic worlds through a process of successive plan refinements.
------------------------------
Date: Mon, 30 Jun 86 12:49:03 pdt
From: eugene@AMES-NAS.ARPA (Eugene Miya)
Subject: Seminar - Real-Time Inferencing with Adaptive Logic Networks (NASA)
National Aeronautics and Space Administration
Ames Research Center
AMES AI FORUM
SEMINAR ANNOUNCEMENT
SPEAKER: Jacques J. Vidal
University of California, Los Angeles
TOPIC: REAL-TIME MULTISENSOR INFERENCING WITH ADAPTIVE LOGIC NETWORKS
The talk will present a general architecture model for special-purpose
parallel processing networks that perform logical inferences in real-time.
Operation is divided beween two complementary modes: Adaptation
(programming) and Processing. The data processing mode is a hierarchical,
asynchronous and completely parallel dataflow. Typically, logic operations
stored in a dynamically reconfigfurable combinatorial network, are performed
on sensor data. In the adaptation mode the network incrementally receives
goal information (either from a human user or directly from environment
sensors), and the node functions and/or connections self-adapt in order for
the output(s) to continually satisfy the externally defined goal. Adaptive
control is sequential, but performed in a distributed and largely concurrent
manner by the network nodes.
The target applications are event-detection, malfunction management and
similar robot functions, including vision.
DATE: Thursday, TIME: 1:00 - 2:00 pm BLDG. 239 Room B39
July 10, 1986 -------------- (Basement Conf. Room)
POINT(S) OF CONTACT: Lee Duke PHONE NUMBER: (805) 258-3802
NET ADDRESS: duke%ofe@ames-io.arpa
or Alison Andrews (415) 694-6741 andrews%ear@ames-io.arpa
(PLEASE NOTE ALISON'S EMAIL ADDRESS CHANGE! ^ )
VISITORS ARE WELCOME: Register and obtain vehicle pass at Ames Visitor
Reception Building (N-253) or the Security Station near Gate 18. Do not
use the Navy Main Gate.
Non-citizens (except Permanent Residents) must have prior approval [...]
------------------------------
Date: 26 Jun 86 17:07:07 EDT
From: Marcella.Zaragoza@isl1.ri.cmu.edu
Subject: Seminar - Overview of the MENTOR System (CMU)
SPECIAL SEMINAR
Topic: OVERVIEW OF THE MENTOR SYSTEM
Speaker: Bernard Lang, INRIA
Place: WeH 8220
Date: Monday, June 30
Time: 11:00am - 12:00noon
Mentor is a structured document manipulation system based on a
representation of documents as abstract syntax trees. After an
overview of the first implementation of Mentor and of the experience
acquired with its use for the development and maintenance of programs
and languages, we shall present some of the new developments underway.
A new version of the system is now being developed in a Lisp dialect
(Le_Lisp) in an object oriented style, with a strong emphasis on the
realisation of a complete kernel for abstract syntax tree manipulation
(user interfaces being developed independantly). The language Typol
for semantics specification, and the language PPML for pretty-printers
specification shall be briefly introduced.
------------------------------
Date: Mon, 23 Jun 86 12:12:07 edt
From: rba@petrus.bellcore.com (Robert B. Allen)
Subject: Conference - ACM Conference on Office Information Systems
ACM CONFERENCE ON OFFICE INFORMATION SYSTEMS
October 6-8, 1968, Providence, R.I.
Conference Chair: Carl Hewitt, MIT
Program Chair: Stan Zdonik, Brown University
Keynote Speaker: J.C.R. Licklider, MIT
Distinguished Lecturer: A. van Dam, Brown University
Panels and Sessions
Advanced Computational Models
AI in the Office
Impacts of Computer Technology on Employment
Organizational Analysis: Due Process
Future Directions in Office Technology
Comparison of Social Research Methods
Organizational Analysis: Organizational Ecology
Models of the Distributed Office
Interfaces
For more information, call the Conference Registrar at Brown U.
(401-813-1839), or send electronic mail to mhf@brown.CSNET.
------------------------------
End of AIList Digest
********************
