Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 128
AIList Digest Tuesday, 27 May 1986 Volume 4 : Issue 128
Today's Topics:
Seminars - Intelligent Systems on Multiprocessors (CMU) &
Information Retrieval by Text Skimming (CMU) &
Theory of Nested Transactions (CMU) &
Intuitionist Logic & Constructive Type Theory (MIT) &
Learning to Construct Abstractions (MIT) &
Non-monotonicity in Probabilistic Logic (SU) &
Autoepistemic Logic, Stratified Programs, Circumscription (SU) &
Sequentialising Logic Programs (SU)
----------------------------------------------------------------------
Date: 13 May 1986 1533-EDT
From: Theona Stefanis@A.CS.CMU.EDU
Subject: Seminar - Intelligent Systems on Multiprocessors (CMU)
PS SEMINAR
Name: R. Bisiani, CMU
Date: Monday, 19 May
Time: 3:30
Place: WeH 5409
DEVELOPING INTELLIGENT SYSTEMS ON MULTIPROCESSOR ARCHITECTURES
Our long-term goal is to develop a software environment that meets the need
of application specialists to build and evaluate heterogeneous AI
applications quickly and efficiently. Speech and vision systems are typical
of this kind of AI applications. In these systems, @i(knowledge-intensive)
and conventional programming techniques must be integrated while observing
real time constraints and preserving good programmability characteristics.
State-of-the-art AI environments solve some but not all of the problems
raised by the systems we are interested in. Therefore, we are developing a
set of tools, methodologies and architectures called Agora that can be used
to implement custom programming environments. Agora can be customized to
support the programming model that is more suitable for a given application.
Agora has been designed explicitly to support multiple languages and highly
parallel computations. Systems built with Agora can be executed on a number
of general purpose and custom multiprocessor architectures.
------------------------------
Date: 19 May 86 15:18:26 EDT
From: Michael.Mauldin@cad.cs.cmu.edu
Subject: Seminar - Information Retrieval by Text Skimming (CMU)
What: Thesis Proposal: Information Retrieval By Text Skimming
Who: Michael L. Mauldin (MLM@CAD)
When: May 29, 1986 At 3pm
Where: In Wean Hall 5409
Most information retrieval systems today are word based. But simple word
searches and frequency distributions do not provide these systems with an
understanding of their texts. Full natural language parsers are capable of
deep understanding within limited domains, but are too brittle and slow for
general information retrieval.
The proposed dissertation attempts to bridge this gap by using a text skimming
parser as the basis for an information retrieval system that partially
understands the texts stored in it. The objective is to develop a system
capable of retrieving a significantly greater fraction of relevant documents
than is possible with a keyword based approach, without retrieving a larger
fraction of irrelevant documents. As part of my dissertation, I will
implement a full-text information retrieval system called FERRET (Flexible
Expert Retrieval of Relevant English Texts). FERRET will provide information
retrieval for the UseNet News system, a collection of 247 news groups covering
a wide variety of topics. Initially FERRET will cover NET.ASTRO, the
Astronomy news group, and part of my investigation will be to demonstrate the
addition of new domains with only minimal hand coding of domain knowledge.
FERRET will acquire the details of a domain automatically using a script
learning component.
FERRET will consist of a text skimming parser (based on DeJong's FRUMP
program), a case frame matcher that compares the parse of the user's query
with the parses of each text stored in the retrieval system, and a user
interface. The parser relies on two knowledge sources for its operation: the
sketchy script database, which encodes domain knowledge, and the lexicon. The
lexicon from FRUMP will be extended both by hand and automatically with syntax
and synonym information from an on-line English dictionary. The script
database from FRUMP will be extended both by hand and automatically by a
learning component that generates new scripts based on texts that have been
parsed. The learning component will evaluate the new scripts using feedback
from the user, and retain the best performers for future use.
The resulting information retrieval system will be evaluated by determining
its performance on queries of the UseNet database, both in terms of relevant
texts not retrieved and irrelevant texts that are retrieved. Over six million
characters appear on UseNet each week, so there should be enough data to study
performance on a large database.
The main contribution of the work will be a demonstration that a text skimming
retrieval system can make distinctions based on semantic roles and information
that word based systems cannot make. The script learning and dictionary
access are new capabilities that will be widely useful in other natural
language applications.
------------------------------
Date: 19 May 1986 1619-EDT
From: Theona Stefanis@A.CS.CMU.EDU
Subject: Seminar - Theory of Nested Transactions (CMU)
PS SEMINAR
Date: Thursday, 22 May
Time: 3:30
Place: WeH 7220
Prolegomenon to the Theory of Nested Transactions
Michael Merritt
A. T. and T. Bell Laboratories
Murray Hill, New Jersey
"The possibility of a thing can never be proved merely from the
fact that its concept is not self-contradictory, but only through
its being supported by some corresponding intuition." Immanuel Kant
This talk develops the foundation for a general theory of nested
transactions. Not without trepidation, it presents yet another formal
model for studying concurrency and resiliency in a nested environment.
This model has distinct advantages over the many alternatives, the
greatest of which is the unification of a subject replete with
formalisms, correctness conditions and proof techniques. The speaker
is presently engaged in an ambitious project to recast the
substantial amount of work in nested transactions within this single
intuitive framework. The talk focuses on preliminary results
of that project--a description of the model, and its use in stating
and proving correctness conditions of a lock-based concurrent scheduler.
This is joint work with Nancy Lynch, of the
Massachusetts Institute of Technology.
------------------------------
Date: Fri 9 May 86 09:49:15-EDT
From: Susan Hardy <SH@XX.LCS.MIT.EDU>
Subject: Seminar - Intuitionist Logic & Constructive Type Theory (MIT)
Friday, May 9, l986
TALK 1: 10:00 a.m., TALK 2: 2:00 p.m.
2nd Floor Lounge
David Turner
University of Kent at Canterbury, England
TALK 1: Intuitionist Logic and Functional Programming
Intuitionism is a heretical school of mathematics founded by L.
E. J. Brouwer in l907. The most outstanding characteristic of
intuitionists is that they reject the use of Boolean logic. Recent
discoveries have shown that there is deep underlying connection
between intuitionist logic and functional programming. This discovery
is likely to have profound consequences for the future of both
subjects. The talk will attempt to explain from the beginning what
intuitionist logic is about and how the coincidence with functional
programming arises.
TALK 2: Constructive Type Theory as a Programming Language
Constructive type theory is a formal logic and set theory which has
been developed by Per Martin-Lof as a foundation for constructive (or
intuitionist) mathematics. Curiously, it can also be read as a
(strongly typed) functional programming language, with a number of
unusual properties, including that all well typed programs terminate.
The talk will give an overview of the main ideas in constructive type
theory from the point of view of someone trying to use it as a
programming language.
HOST: Professors Arvind and Rishiyur S. Nikhil
------------------------------
Date: Wed, 14 May 1986 11:16 EDT
From: JHC%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU
Subject: Seminar - Learning to Construct Abstractions (MIT)
-- AI Revolving Seminar --
LEARNING TO CONSTRUCT ABSTRACTIONS
Rick Lathrop
MIT AI Lab
One useful trait of an intelligent agent is to construct higher-level
abstractions from a mass of detailed low-level information. This talk
will explore one way an agent might be taught how to construct such
abstractions, and why it might be a useful or interesting for an agent
to do so. A main motivation is the possibility of the use of these
abstractions to see similarities (between situations) that are
obscured by the mass of irrelevant details at the lower level.
Preliminary examples from the Rieger (causal) mechanism world, VLSI
circuit analysis, and protein structure analysis will be discussed.
Thursday, May 15, 4pm
NE-43, 8th floor playroom
------------------------------
Date: 13 May 86 1511 PDT
From: Vladimir Lifschitz <VAL@SU-AI.ARPA>
Subject: Seminar - Non-monotonicity in Probabilistic Logic (SU)
NON-MONOTONICITY IN PROBABILISTIC LOGIC
Benjamin Grosof
Computer Science Department, Stanford University
Thursday, May 15, 4pm
MJH 252
I will discuss how to formalize the notion of non-monotonicity in
probabilistic reasoning, using the framework of Probabilistic Logic
(cf. Nils Nilsson). I will give some motivating examples of types of
non-monotonic probabilistic reasoning that seem to be found in
practice. There seems to be a relationship to default inheritance,
i.e. prioritized defaults of the type used in classic example of
whether birds and ostriches fly. Next, I introduce the idea of
maximizing conditional independence, which can be thought of as
maximizing irrelevance. This can be described more simply in terms of
non-monotonic reasoning on Graphoids (cf. Judea Pearl).
I conjecture that an important type of non-monotonicity in probabilistic
reasoning may be concisely expressed in terms of conditional
independence and Graphoids. Finally, I pose as an open question how
to formulate in the above terms the non-monotonic behavior of
maximizing entropy, a widely-used technique in probabilistic
reasoning.
------------------------------
Date: 19 May 86 1322 PDT
From: Vladimir Lifschitz <VAL@SU-AI.ARPA>
Subject: Seminar - Autoepistemic Logic, Stratified Programs,
Circumscription (SU)
AUTOEPISTEMIC LOGIC, STRATIFIED PROGRAMS AND CIRCUMSCRIPTION
Michael Gelfond and Halina Przymusinska
University of Texas at El Paso
Thursday, May 22, 4pm
MJH 252
In Moore's autoepistemic logic, a set of beliefs of a rational agent
is described by a "stable expansion" of his set of premises T. If this
expansion is unique then it can be viewed as the set of theorems which
follow from T in autoepistemic logic. Marek gave a simple syntactic
condition on T which guarantees the existence of a unique stable
expansion. We will propose another sufficient condition, which is
suggested by the definition of "stratified" programs in logic
programming. The declarative semantics of such programs can be
defined using fixed points of non-monotonic operators (Apt, Blair and
Walker; Van Gelder) or by means of circumscription (Lifschitz). We
show how this semantics can be interpreted in terms of autoepistemic
logic.
------------------------------
Date: Fri 23 May 86 11:33:27-PDT
From: Richard Treitel <TREITEL@SU-SUSHI.ARPA>
Subject: Seminar - Sequentialising Logic Programs (SU)
PhD oral examination
Tuesday June 3rd 1986 at 3 p.m.
Building 200, Room 34
"Sequentialising Logic Programs"
Richard Treitel
In "expert systems" and other applications of logic programming, the issue
arises of whether to use rules for forward or backward inference, i.e. whether
deduction should be driven by the facts available to the rule or the goals that
are put to it. Often some mixture of the two is cheaper than using either
exclusively. I show that, under two restrictive assumptions, optimal choices
of directions for the rules can be made in time polynomial in the number of
rules in a recursion-free logic program. If either of these restrictions is
abandoned, the optimal choice is NP-complete. I present a search algorithm for
the easiest of the cases so obtained.
A related issue is the ordering of the terms in a rule, which can have a strong
effect on the computational cost of using the rule. Algorithms for ordering
terms optimally are known, but all of them rely on the direction of inference
being fixed in advance, and they apply only to a single rule considered in
isolation. A more general algorithm is developed, and a way is shown to
incorporate it into the choice of rule directions. This also leads to an
NP-complete problem.
Some attention is paid to the model of execution cost for logic programs on
which these results are based. Logic programs involving recursion are not
covered by this work because no satisfactory cost model exists for them.
------------------------------
End of AIList Digest
********************
