Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 083
AIList Digest Thursday, 19 Mar 1987 Volume 5 : Issue 83
Today's Topics:
Seminars - EFFIGY: Symbolic Execution of Programs (IBM) &
Applying Precedents in Case-Based Reasoning (Rochester) &
Learning Decomposition Methods (CMU) &
Circumscriptive Theories (SU),
Conferences - TI's AI Satellite Symposium III &
AAAI Workshop on Battle Management
----------------------------------------------------------------------
Date: Wed, 18 Mar 87 18:21:19 PST
From: IBM Almaden Research Center Calendar <CALENDAR@IBM.COM>
Subject: Seminar - EFFIGY: Symbolic Execution of Programs (IBM)
IBM Almaden Research Center
650 Harry Road
San Jose, CA 95120-6099
EFFIGY: SYMBOLIC EXECUTION OF PROGRAMS
J. C. King, IBM Almaden Research Center
Computer Science Coll. Thurs., March 26 3:00 P.M. Room: Front Aud.
Long ago and far away, a group in IBM Yorktown Heights devised a
computer system called "EFFIGY" which executed computer programs
"symbolically." On a recent archeological dig in musty old CMS files,
I stumbled upon what appeared to be a genuine EFFIGY MODULE. After a
time, with a new FILEDEF and a long forgotten LINK, I was able to
execute the model, just as the ancients did. It was amazing for me to
remember how advanced civilization was, even then (12-15 years ago).
For some reason, the art of symbolic execution never caught on in a
big way, and it has nearly been lost. For those of the younger
generation, who have never seen the chanting and chest beating of the
symbolic executors (sexers for short), I will try to recreate some of
that ancient spirit. Especially with the new projection system in the
Front Auditorium, which is capable of showing computer terminal output
on-line, I can demonstrate this EFFIGY system, as it was only possible
to do before in a one-on-one situation in a Yorktown cave. The sexers
had discovered that the same leverage obtained by using algebra to
understand and prove things about arithmetic can be applied to
computer programs. If one executes a program using mathematical
symbols, instead of numbers, as program inputs, the same algebraic
leverage can be obtained. Of course, the dynamic aspects of program
execution makes this process tantalizingly non trivial. Combining the
well-known concepts of program execution and algebra, the notions of
"proving the correctness of programs" and "inductive assertions" can
be easily understood without knowingly resorting to heavy mathematical
concepts.
Host: R. Williams
(Refreshments at 2:45 P.M.)
------------------------------
Date: Mon, 16 Mar 87 16:46:45 EST
From: tim@linc.cis.upenn.edu (Tim Finin)
Subject: Seminar - Applying Precedents in Case-Based Reasoning
(Rochester)
Colloquium
Computer and Information Science
University of Pennsylvania
"Applying Relevant Precedents in a Case-Based Reasoning System"
Kevin D. Ashley
Department of Computer and Information Science
University of Massachusetts at Amherst
The law is an excellent domain to study Case-Based Reasoning (``CBR")
problems since it espouses a doctrine of precedent in which prior
cases are the primary tools for justifying legal conclusions. The law
is also a paradigm for adversarial CBR; there are ``no right answers",
only arguments pitting interpretations of cases and facts against each
other.
This talk will demonstrate techniques employed in the HYPO program for
representing and applying case precedents and hypothetical cases to
assist an attorney in evaluating and making arguments about a new fact
situation. HYPO performs case-based reasoning and, in particular,
models legal reasoning in the domain of trade secrets law. HYPO's key
elements include: (1) a structured case knowledge base (``CKB") of
actual legal cases; (2) an indexing scheme (``dimensions") for
retrieval of relevant precedents from the CKB; (3) techniques for
analyzing a current fact situation (``cfs"); (4) techniques for
``positioning" the cfs with respect to relevant precedent cases in the
CKB and finding the most on point cases (``mopc"); (5) techniques for
manipulating cases (e.g., citing, distinguishing, hybridizing); (6)
techniques for perturbing the cfs to generate hypotheticals that test
the sensitivity of the cfs to changes, particularly with regard to
potentially adverse effects of new damaging facts coming to light and
existing favorable ones being discredited; and (7) the use of ``3-ply"
argument snippets to dry run and debug an argument.
An extended example of HYPO in action on a sample trade secrets case
will be presented. The example will demonstrate how HYPO uses
``dimensions", ``case-analysis-record" and ``claim lattice" mechanisms
to perform indexing and relevancy assessment of precedent cases
dynamically and how it compares and contrasts cases to come up with
the best precedents pro and con a decision.
March 20, 1987
3:00 to 4:30
Room 216
Refreshments Available
2:30-3:00
Faculty Lounge
------------------------------
Date: 18 Mar 87 01:11:28 EST
From: Steven.Minton@cad.cs.cmu.edu
Subject: Seminar - Learning Decomposition Methods (CMU)
This week's speaker is Sridhar Mahadevan. As usual, the seminar is
in 7220 Wean on Friday at 3:15. Come one, come all.
LEARNING DECOMPOSITION METHODS TO IMPROVE HIERARCHICAL
PROBLEM-SOLVING PERFORMANCE
Previous work in machine learning on improving problem-solving
performance has usually assumed a @i(state-space) or "flat"
problem-solving model. However, problem-solvers in complex domains,
such as design, usually employ a hierarchical or problem-reduction
strategy to avoid the combinatorial explosion of possible operator
sequences. Consequently, in order to apply machine learning to
complex domains, hierarchical problem-solvers that automatically
improve their performance need to designed. One general approach is
to design an @i(interactive) problem-solver -- a @i(learning
apprentice) -- that learns from the problem-solving activity of expert
users. In this talk we propose a technique, VBL, by which such a
system can learn new problem-reduction operators, or @i(decomposition
methods), based on a verification of the correctness of example
decompositions. We also discuss two important limitations of the VBL
technique -- intractability of verification and specificity of
generalization -- and propose solutions to them.
------------------------------
Date: 18 Mar 87 1142 PST
From: Vladimir Lifschitz <VAL@SAIL.STANFORD.EDU>
Subject: Seminar - Circumscriptive Theories (SU)
CIRCUMSCRIPTIVE THEORIES
Vladimir Lifschitz
Thursday, March 19, 4pm
Bldg. 160, Room 161K
The use of circumscription for formalizing commonsense knowledge and
reasoning requires that a circumscription policy be selected for each
particular application: we should specify which predicates are
circumscribed, which predicates and functions are allowed to vary,
what priorities between the circumscribed predicates are established,
etc. The circumscription policy is usually described either informally
or using suitable metamathematical notation. In this talk a simple and
general formalism will be proposed which permits describing circumscription
policies by axioms, included in the knowledge base along with the axioms
describing the objects of reasoning.
------------------------------
Date: Mon, 16 Mar 87 14:33:02 cst
From: "Michael T. Gately" <gately%resbld%ti-csl.csnet@RELAY.CS.NET>
Subject: Conference - TI's AI Satellite Symposium III
This is a short extension to Dan Cerys' message of 12-MAR-87
regarding the TI Artificial Intelligence Satellite Symposium.
First, the phone number, 1-800-527-3500 can be used to answer
many questions; such as how to rent a satellite antenna, what
type of video equipment is necessary for different audience
sizes, etc. Second, take note of the unusual time shifting for
different time zones across North America. Finally, the
following is a list of cities which already have public sites
planned. Please call the toll-free number as soon as possible to
reserve a seat.
AI Satellite Symposium III
"AI Productivity Roundtable"
April 8, 1987
Eastern/Rocky Mountain Time Zones (Daylight Times)
9:00 - 13:00
Pacific/Central Time Zones (Daylight Times)
8:00 - 12:00
AI Symposium II condensation
April 8, 1987
Eastern/Rocky Mountain Time Zones (Daylight Times)
14:00 - 15:30
Pacific/Central Time Zones (Daylight Times)
13:00 - 14:30
Atlanta GA Austin TX Boston MA
Chicago IL Cleveland OH Dallas TX
Dayton OH Denver CO Detroit MI
Hartford CT Houston TX Huntsville AL
Kansas City KS Los Angeles CA Miami FL
Milwaukee WI Minneapolis MN Montreal Canada
New York NY Philadelphia PA Raleigh/Durham NC
San Diego CA San Francisco CA San Jose CA
Seattle WA St. Louis MO Summit NJ
Toronto Canada Washington DC
------------------------------
Date: Mon, 16 Mar 87 12:26:17 est
From: elsaesser%mwcamis@mitre.ARPA
Subject: Conference - AAAI Workshop on Battle Management
Issues Concerning AI Applications To Battle Management
University of Washington
Thursday, July 16, 1987
Sponsored by AAAI
Success in applying AI technologies to battle management (e.g., production
and blackboard systems for sensor fusion, constraint propagation for
non-temporal planning tasks) has generated growing interest in the defense
community in developing intelligent battle management aids, workstations,
and systems. Along with this growing interest, there has been an order of
magnitude increase in funding for battle management AI projects (e.g.,
Army-DARPA's Air-Land Battle Management, SAC-JSTPS-RADC-DARPA's
Survivable Adaptive Planning Experiment).
Past successes belie the lag of the AI community in solving technical
issues associated with these projects. These issues include those
associated with cooperating knowledge-based systems, distributed problem
solving, uncertainty management, non-monotonic reasoning, planning,
real-time performance requirements (i.e., the need for parallel or other
advanced architectures), and the ability of users to maintain understanding
and control of the automation.
The purpose of this workshop is to gather together researchers who are
attempting to find solutions to these and related issues and to discuss the
current state of these arts. We believe that not enough has been done in
these key areas areas, and that one result of the workshop might be some
road map of how the community ought to proceed. The issues are so numerous
and the area is large enough that we feel the initial workshop will only
allow us to delineate how much has been done and what needs to be done in
key areas. Thus, the goal is both to articulate where the major gaps are
and which ones have a reasonable chance of solution in some believable
time-frame.
Interested persons should submit an extended abstract of not more than six
pages to either person listed below (no on-line submissions please) on an AI
subject of relevance to the above workshop objectives not later than 1 May
1987. Authors will be notified of acceptances by 1 June 1987, along with
information relative to the workshop administration.
R. Peter Bonasso Chris Elsaesser
(703) 883 6908 (703) 883 6563
bonasso@mitre elsaesser%mwcamis@MITRE
MITRE Washington AI Center
Mail Stop W410
7525 Colshire Drive
McLean, VA 22102
------------------------------
End of AIList Digest
********************
