Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 102
AIList Digest Tuesday, 21 Apr 1987 Volume 5 : Issue 102
Today's Topics:
Seminars - Problems in Nonmonotonic Reasoning (MCC) &
Constraints, Planning, and Design (TI) &
Order-Sorted Unification (SRI) &
Conceptual Thinking for Restructuring and Insight (CMU) &
A Synthesis of Higher-Order Unification (CMU) &
Graphical Access to an Expert System (Rutgers) &
Reactive Learning (CMU) &
Equivalences of Logic Programs (Rutgers)
----------------------------------------------------------------------
Date: Tue 14 Apr 87 15:24:13-CDT
From: Charles Petrie <AI.PETRIE@MCC.COM>
Reply-to: Petrie@MCC
Subject: Seminar - Problems in Nonmonotonic Reasoning (MCC)
Michael Reinfrank
NML/RMS-Group, Dept. of CS, Univ. of Linkoeping
XPSLAB ZTI INF 31, SIEMENS AG, Munich
PROBLEMS IN NONMONOTONIC REASONING
MCC Auditorium
10:30 April 16
Nonmonotonic reasoning now has been a topic of interest for more than
fifteen years, although substantial research into its theoretical
foundations did not begin until the late seventies. Recently, some doubts
arose concerning the achievements in this field, in particular concerning
the question whether the techniques developed so far can solve those
problems they are intended to solve. A survey of the history of nonmonotonic
reasoning will be given and major unresolved issues in the current theories
identified.
------------------------------
Date: Wed, 15 Apr 87 13:27:12 cdt
From: "Michael T. Gately" <gately%resbld%ti-csl.csnet@RELAY.CS.NET>
Subject: Seminar - Constraints, Planning, and Design (TI)
From: NGSTL1::LINDAHL "Multihack -- lindahl%ngstl1@ti-eg.csnet"
From: TILDE::"BEF@HOME"
Texas Instruments Computer Science Center Lecture Series
CONSTRAINTS, PLANNING, AND DESIGN:
THERE IS A REASON FOR EVERYTHING UNDER THE SUN
PROF. DANIEL WEISE (STANFORD UNIVERSITY)
10:00 am, Friday, 1 May 1987
Semiconductor Building Main Auditorium
Every design decision has a set of rationales and a set of
ramifications. These ramifications affect other design decisions. I
believe that algorithms and expert systems fail for automatic design
synthesis because they do not explicitly reason about rationales or
ramifications. They also fail because they do not react to the design
being built. In this talk I will outline the problems of
automatically designing hardware, show why current approaches must
fail, and describe a new methodology, based on communicating
constraint based problem solvers, which might succeed.
BIOGRAPHY
Daniel Weise received both his Masters and Ph.D. degrees at the MIT
Artificial Intelligence Laboratory. He did his dissertation on
verifying MOS circuits. He is now an assistant professor at Stanford
University working on design automation and silicon compilation.
----------------------------------------------------------------------
The lecture will be given in the Semiconductor Building Main
Auditorium at the Dallas Expressway site. Visitors to TI should
contact Dr. Bruce Flinchbaugh (214-995-0349) in advance and meet in
the north entrance lobby of the Semiconductor Building by 9:45am.
------------------------------
Date: Thu, 16 Apr 87 10:27:48 PDT
From: lunt@april.csl.sri.com (Teresa Lunt)
Subject: Seminar - Order-Sorted Unification (SRI)
ORDER-SORTED UNIFICATION
Jose Meseguer
SRI International Computer Science Laboratory
Monday, April 27 at 4:00 pm
SRI International, Computer Science Laboratory, BN182
Jose Meseguer will speak on his work with Joseph Goguen of
SRI International and Gert Smolka of Universitat Kaiserslautern.
Order-sorted logic is the logic of multiple inheritance and
overloading polymorphism. It provides a rich type theory
that permits easy and natural expression of many problems in
knowledge representation, natural language processing,
theorem proving, etc. Order-sorted logic is also the basis
for the logical languages OBJ3 and Eqlog. Despite its
considerable expressive power, all the usual results of
equational and first-order logic generalize to order-sorted
logic. The present work develops a general theory of
order-sorted E-unification, and characterizes the cases
where there is a minimal family of unifiers, a finite family
of unifiers, and a unique most general unifier. The latter
case has a simple syntactic characterization and also a
quasi-linear unification algorithm a la Martelli-Montanari
that is in fact more efficient than ordinary unification,
due to its type-checking.
------------------------------
Date: 16 Apr 1987 1049-EDT
From: Elaine Atkinson <EDA@C.CS.CMU.EDU>
Subject: Seminar - Conceptual Thinking for Restructuring and Insight
(CMU)
SPEAKER: Dr. Stellan Ohlsson, LRDC, University of Pittsburgh
TITLE: "A theory of conceptual thinking applied to the phenomena of
restructuring and insight"
DATE: Tuesday, April 21
TIME: 12:00 - 1:20 p.m.
PLACE: Adamson Wing, Baker Hall
ABSTRACT: Current theories of problem solving focus on the nature and function
of problem solving strategies. However, novel problems cannot, by definition,
be solved by applying a pre-existing strategy; rather, they are solved by
trying to understand the problem situation, a process to be called "conceptual
thinking". According to studies by the Gestalt psychologists, conceptual
thinking exhibits the phenomena of restructuring and insight. A first
approximation theory of restructuring and insight and some relevant data
will be discussed.
------------------------------
Date: 15 Apr 87 10:08:23 EDT
From: Conal.Elliott@theory.cs.cmu.edu
Subject: Seminar - A Synthesis of Higher-Order Unification (CMU)
Area Qualifier Talk
Speaker: Conal Elliott
Date: April 21
Time: 10:00-11:30
Place: WeH 7220
Topic: A Synthesis of Higher-Order Unification
Program synthesis is the derivation of implementations from noneffective
specifications.
Higher-order unification is unification in the typed lambda calculus with
alpha, beta, and eta conversion. It has been used in
- program manipulation,
- theorem proving in higher-order logic,
- logic programming, and
- mechanizing natural deduction.
In this talk, we
- give a new, useful conceptualization of the unification problem,
- synthesize a family of ``pre-algorithms'' for unification, unifiablity,
matching, matchability, with some efficiency improvements, and
- present a new synthesis methodology, which may be viewed as a new
interpretation, justification, and generalization of Burstall &
Darlington's methodology.
------------------------------
Date: 16 Apr 87 14:14:26 EDT
From: KALANTARI@RED.RUTGERS.EDU
Subject: Seminar - Graphical Access to an Expert System (Rutgers)
RUTGERS COMPUTER SCIENCE COLLOQUIUM SCHEDULE - SPRING 1987
DATE: Thursday, April 23, 1987
SPEAKER: Ted Shortliffe
AFFILIATION:
Visiting Professor of Computer and Information Science
University of Pennsylvania
and
Associate Professor of Medicine and Computer Science
Medical Computer Science Group
Knowledge Systems Laboratory
Stanford Medical School
TITLE: GRAPHICAL ACCESS TO AN EXPERT SYSTEM: THE EVOLUTION OF THE ONCOCIN
PROJECT
TIME: 2:50 (Coffee and Cookies will be setup at 2:30)
PLACE: Hill Center, Room 705
ABSTRACT
The research goals of Stanford's Medical Computer Science group are directed
both toward the basic science of artificial intelligence and toward the
development of clinically useful consultation tools. Our approach has been
eclectic, drawing on fields such as decision analysis, interactive graphics,
and both qualitative and probabilistic simulation as well as AI. In this
presentation I will discuss ONCOCIN, an advice system designed to suggest
optimal therapy for patients undergoing cancer treatment, as well as to
assist in the data management tasks required to support research treatment
plans (protocols). A prototype version, developed in Interlisp and SAIL
on a DEC-20, was used between May 1981 and May 1985 by oncology faculty and
fellows in the Debbie Probst Oncology Day Care Center at the Stanford
University Medical Center. In recent years, however, we have spent much
of our time reimplementing ONCOCIN to run on Xerox 1100 series workstations
and to take advantage of the graphics environment provided on those
machines. The physician's interface has been redesigned to approximate the
appearance and functionality of the paper forms traditionally used for
recording patient status. The Lisp machine version of ONCOCIN was introduced
for use by Stanford physicians earlier this year.
In response to the need for an improved method for entering and maintaining
the rapidly expanding ONCOCIN protocol knowledge base, we have also developed
a graphical knowledge acquisition environment known as OPAL. This system
allows expert oncologists to directly enter their knowledge of protocol-
directed cancer therapy using graphics-based forms developed in the
Interlisp-D environment. The development of OPAL's graphical interface led
to a new understanding of the natural structure of knowledge in this domain.
ONCOCIN's knowledge representation was accordingly redesigned for the Lisp
machine environment. This has involved adopting an object-centered knowledge
base design which has provided an increase in the speed of the program while
providing more flexible access to system knowledge.
------------------------------
Date: 17 Apr 87 15:59:21 EDT
From: Patricia.Mackiewicz@isl1.ri.cmu.edu
Subject: Seminar - Reactive Learning (CMU)
TOPIC: Reactive Learning: Experimentation and Decompilation
SPEAKER: Jaime Carbonell, CMU
WHEN: Tuesday, April 21, 1987, 3:30 p.m.
WHERE: Wean Hall 5409
Most symbolic learning approaches have been purely empirical (inductive)
or purely analytical. The former extracts a general concept from a set of
empirical observations, whereas the latter composes primitive concepts
into larger units (chunks, macro-operators, "explanations", etc.).
Analytical methods include explanation-based learning, capable of
exploiting a complete domain theory to learn complex concepts from very few
instances. However, the domain theory may be partial, and judicious
integration of empirical and analytical methods may prove far superior
to either method alone. Reactive experimentation is a case in point:
partial domain knowledge is used to formulate hypotheses, and empirical
data from the experiments is used to formulate new concepts or modify
existing ones. Decompilation maps complex empirical observations into
comprehensible operational units using analytical techniques. Both
methods for combining analytical and empirical approaches are
explored with the objective of creating robust learning systems.
------------------------------
Date: 19 Apr 87 04:18:08 EDT
From: KALANTARI@RED.RUTGERS.EDU
Subject: Seminar - Equivalences of Logic Programs (Rutgers)
SPECIAL RUTGERS COMPUTER SCIENCE COLLOQUIUM
DATE : Tuesday, April 21st
SPEAKER: Michael J. Maher
TITLE: Equivalences of Logic Programs
AFFILIATION: IBM T.J. Watson Center
TIME: 1:30 pm
PLACE: Hill 423
One of the most important relationships between programs in any
programming language is the equivalence of such programs. This
relationship is at the basis of most, if not all, programming
methodologies. This talk provides a systematic comparison of
the relative strengths of various formulations of equivalence
for logic programs. These formulations arise naturally from
several well-known formal semantics. These comparisons are
useful in reasoning about program behavior, verification of
correctness and termination of programs, the correctness of ad
hoc source-to-source transformations such as occur in program
development, and, at a more abstract level, the establishment of
the correctness and other properties of automated transformation
systems which can be used both in program development and as a
pre-compilation optimization.
---------------------------------------------------------------------------
DATE : Friday, April 24
SPEAKER: Dr. Susan Epstein
TITLE: An Introduction to GT, the Graph Theorist
AFFILIATION: Hunter College (CUNY)
TIME: 1:30 (Coffee and Cookies will be setup after the talk at 2:30)
PLACE: Hill 705
ABSTRACT
GT, the Graph Theorist, is a knowledge-intensive, domain-specific
learning system which uses algorithmic class descriptions to
discover new mathematical concepts and relations among them.
GT is based upon a set of powerful
representation languages for object classes. The definition of a graph theory
concept is an expression in one of these languages.
GT generates correct examples of any of its concepts, constructs new
concepts, and conjectures and proves relations among concepts. Beginning
from only the concept of "graph," GT has developed its own version of graph
theory and discovered such concepts as "tree," "acyclic," "connected,"
and "bipartite." GT has also conjectured and then proved such theorems as
"The set of acyclic, connected graphs is precisely the set of trees" and
"There is no odd-regular graph on an odd number of vertices."
This talk presents initial results and outlines the theoretical
foundations for this work.
------------------------------
End of AIList Digest
********************
