Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 088
AIList Digest Thursday, 26 Mar 1987 Volume 5 : Issue 88
Today's Topics:
Conference - Genetic Algorithms,
Seminars - NUPRL as a Framework for Defining Logics (UPenn) &
Natural Deduction Meets Schubert's Steamroller (SMU) &
Parallel Production System Algorithms (UTexas)
----------------------------------------------------------------------
Date: Tue, 24 Mar 87 14:04:55 est
From: John Grefenstette <gref@nrl-aic.ARPA>
Subject: Genetic Algorithms
Copies of the Proceedings of the First International Conference
on Genetic Algorithms, held at Carnegie-Mellon in 1985 can be
obtained by sending me your US Mail address.
The 2nd GA Conference, sponsored by AAAI, the Navy Center for
Applied research in AI, and Bolt Beranek and Newman, will be
held July 28-31, 1987, at MIT. For registration forms and info
concerning local arrangements, contact:
Mrs. Gayle M. Fitzgerald
Conference services Office
Room 7-111
MIT
77 Massachusetts Ave.
Cambridge, MA 02139
If you would like to submit a paper to the Conference, please send
three copies of the paper to:
John J. Grefenstette
Navy Center for Applied Research in AI
Naval Research Lab
Washington, DC 20375-5000
(202) 767-2685
Arpanet: gref@NRL-AIC.ARPA
The program committee will review papers starting April 10.
Final camera ready versions will be due May 30.
-- JJG
------------------------------
Date: Mon, 23 Mar 87 08:24:18 EST
From: tim@linc.cis.upenn.edu (Tim Finin)
Subject: Seminar - NUPRL as a Framework for Defining Logics (UPenn)
From: dale%linc.cis.upenn.edu@cis.upenn.edu
Math/CS Logic Seminar
University of Pennsylvania
RECENT RESULTS ABOUT NUPRL:
USING NUPRL AS A FRAMEWORK FOR DEFINING LOGICS.
Robert Constable
Cornell University
Abstract: Nuprl can be used to define natural deduction style logic.
We will also mention other recent results about the Nuprl type theory
such as those about representing partial functions.
Math/Physics Building (DRL)
4th floor Math Seminar Room
Monday 23 March 87, 10:30am
------------------------------
Date: Tue, 24 Mar 1987 18:21 CST
From: Leff (Southern Methodist University)
<E1AR0002%SMUVM1.BITNET@wiscvm.wisc.edu>
Subject: Seminar - Natural Deduction Meets Schubert's Steamroller
(SMU)
Past Seminar, Southern Methodist University, Department of Computer Science
Natural Deduction meets Schubert's Steamroller
Frank Vlach
Texas Instruments
Schubert's Steamroller is a test problem for automatic theorem provers
that has attracted a lot of attention recently, and has proved
difficult for resolution theorem provers. A human theorem prover
would prove Schubert's Steamroller using a `natural' but mechanical
and totally non-creative method that is readily programmable and quite
different from resolution. Hand computations indicate that this
strategy is much less complex than resolution for Schubert's
Steamroller and a number of similar problems. An implementation is in
progress in order to compare this method with resolution (and other
methods) over a wide range of problems.
This strategy also has the advantage that it requires no preprocessing
of formulas (such as Skolemization or conversion to clausal form), and
lends itself to the generation of natural proofs, readable by humans.
------------------------------
Date: Tue 24 Mar 87 14:23:26-CST
From: Adam Farquhar <AI.FARQUHAR@R20.UTEXAS.EDU>
Subject: Seminar - Parallel Production System Algorithms (UTexas)
The COMPUTER SCIENCES GRADUATE STUDENT COUNCIL
PRESENTS
Daniel P. Miranker
Recent Developments in
Parallel Production System Algorithms
at the CSGSC BROWN-BAG SEMINAR
Friday, March 27, 12:00 Noon
Tay 2.106
All Students and Faculty are invited.
Okay to bring your lunch.
The development of a parallel production system interpreters may be
seperated into three nearly independent facets, low-level matching,
partitioning of the rule base and synchronizing the partitions. This
talk will address the partitioning issue.
A problem associated with parallelizing production system execution is
that on any given production system cycle only a small subset of the
rules require processing. Worse, on a given cycle, often the
processing requirements for a single rule will completely dominate the
execution time. "Copy and constrain" is a method by which the
processing requirements for matching a single rule may be distributed
over many processors. This method has been shown to very effectively
reduce the variance of the match times of different rules. Further,
this method has implications for the fault tolerant execution of
production systems. It appears, due to increased processor
utilization, that fault tolerance may be introduced into a parallel
production system interpreter without modification of the hardware and
without significant performance degradation.
------------------------------
End of AIList Digest
********************