Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 252
AIList Digest Thursday, 29 Oct 1987 Volume 5 : Issue 252
Today's Topics:
Journal - Neural Network Review,
Binding - Transatlantic Netmail to UK,
Seminars - Dependency-Directed Prolog (BBN) &
Speech Recognition Using Connectionist Networks (UNISYS),
Conference - Expert Systems in Business and Finance
----------------------------------------------------------------------
Date: Wed, 28 Oct 87 15:30:08 EST
From: csed-1!will@hc.dspo.gov (Craig Will)
Subject: Announcing Neural Network Review
Announcing a new publication
NEURAL NETWORK REVIEW
The critical review journal
for the neural network community
Neural Network Review is intended to provide a forum
for critical analysis and commentary on topics involving
neural network research, applications, and the emerging
industry. A major focus of the Review will be publishing
critical reviews of the neural network literature, including
books, individual papers, and, in New York Review of Books
style, groups of related papers.
The Review will also publish general news about events
in the neural network community, including conferences,
funding trends, and announcements of new books, papers,
courses, and other media, and new hardware and software pro-
ducts.
The charter issue, dated October, 1987, has just been
published, and contains a review and analysis of 11 articles
on neural networks published in the popular press, a report
on the San Diego conference, a report on new funding initia-
tives, and a variety of other information, a total of 24
pages in length. The next issue, due in January, 1988, will
begin detailed reviews of the technical literature. Neural
Network Review is aimed at a national audience, and will be
published quarterly. It is published by the Washington
Neural Network Society, a nonprofit organization based in
the Washington, D.C. area.
Subscriptions to Neural Network Review are $ 10.00 for
4 issues, or $ 2.50 for a single copy. International rates
are slightly higher. Rates for full-time students are $5.00
for 4 issues. (Checks should be payable to the Washington
Neural Network Society). Subscription orders and inquiries
for information should be sent to:
Neural Network Review
P. O. Box 427
Dunn Loring, VA 22027
For more information on Neural Network Review, send your
physical, U. S. Postal mail address in a message to
will@hc.dspo.gov (Craig Will).
------------------------------
Date: 28 Oct 87 16:46 PST
From: hayes.pa@Xerox.COM
Subject: transatlantic netmail to UK
I recently had some correspondence about this with an informed UK
source, and here is his statement about what is going on and why, and
what the future should hold. Looks good.
Pat Hayes
----------
The costs of transatlantic traffic, in both directions, through the UCL
Arpanet gateway are borne by a UK funding agency, the Alvey
Directorate/SERC . Darpa does not pay for messages originating in the
USA and sent to the UK gateway, and UCL ( University College, part of
London University ) has no way of charging individual American
originators of messages.
Some time ago, UCL needed to get more accurate statistics about UK usage
to strengthen its case for more money to run the transatlantic link. To
show that the gateway was a vital facility, UCL instigated the policy of
requiring UK users to be properly authorised, ie officially registered
as users.
The cost of this bi-directional transatlantic traffic now exceeds the
budget granted by Alvey/SERC to UCL. Appeals by UCL to SERC brought
to light that much net traffic originating in the UK was being
channelled through the very few `official' accounts. Moreover, UCL
had no data on the number of US customers it serves.
More recently, the increased cost of running the link has meant that UCL
now wishes to track traffic originating in the USA, to help show the
importance of the link. As USA to UK messages are not funded by any USA
agency, then either the SERC pays for it all, via the UCL budget, or
some form of charge-back to UK recipients must be instigated. This is
the origin of the recent change in operating policy requiring USA users
to be registered as collaborating with some specific UK group ie
charging centre. Such a charge would then be allowable against
individual SERC grants, rather than UCL picking up the total cost.
There is no suggestion that any USA user will be refused authorisation.
It is clear to all parties that this is not a satisfactory mechanism,
either now or for the future. I am pleased to tell you that negotiations
are now well advanced for a more permanent and sensible solution.
The proposal is that the UK's SERC and the USA's NSF (more natural
counterparts than Darpa) will instigate a new USA-UK link, properly
jointly organised and funded for the benefit of academics. This link
will, on the USA side, gateway the UK's Janet (the official name of the
UK academic net) into most of the USA nets (arpa, NSF's own, Usenet
etc. ) The present Arpanet-Janet link will continue until this improved
NSF-Janet comes into service. There is no firm date for this yet, but I
think that if people in the USA cooperate with UCL in the short term,
and have a little patience and sympathy for UCL's predicament, then we
should all be able to keep communicating via UCL until the next
generation gateway comes into service.
------------------------------
Date: Tue 27 Oct 87 10:38:24-EST
From: Marc Vilain <MVILAIN@G.BBN.COM>
Subject: Seminar - Dependency-Directed Prolog (BBN)
BBN Science Development Program
AI Seminar Series Lecture
DEPENDENCY DIRECTED PROLOG
Jeffrey Mark Siskind
MIT Laboratory for Computer Science
(also: summer intern at Xerox PARC)
(Qobi@ZERMATT.LCS.MIT.EDU)
BBN Labs
10 Moulton Street
2nd floor large conference room
10:30 am, Tuesday November 3
In this talk I will describe an implementation of pure Prolog which uses
dependency directed backtracking as a control strategy for pruning the
search space. The implementation uses a strategy whereby the Prolog
program is compiled into a finite set of templates which characterize a
potentially infinite boolean expression which is satisfiable iff there
is a proof of the goal query. These templates are incrementally
unraveled into a sequence of propositional CNF SAT problems and
represented in a TMS which is used to find solutions using dependency
directed backtracking. The technique can be extended to use ATMS-like
strategies for searching for multiple solutions simultaneously.
Two different strategies have been implemented for dealing with
unification. The first compiles the unification constraints into SAT
clauses and integrates them in the TMS along with the and/or goal tree
produced by unraveling the templates. The second uses a separate module
for doing unification at run time. This unifier is novel in that it
records dependencies and allows nonchronological retraction. The
interface protocol between the TMS and the unifier module has been
generalized to allow integration of other "domains" of predicates, such
as linear arithmetic and simple linear inequalities, to be built into
the system while still preserving the soundness and completeness of the
pure logical interpretation of Prolog.
In the talk, time permitting, I will discuss the search prunning
advantages of this approach and its relation to previous approaches, the
implementation mechanism, and some recent work indicating the potential
applicability of this approach to parsing with disjunctive feature
structures, such as done with the LFG and related grammar formalisms.
------------------------------
Date: Tue, 27 Oct 87 15:35:57 EST
From: finin@bigburd.PRC.Unisys.COM (Tim Finin)
Subject: Seminar - Speech Recognition Using Connectionist Networks
(UNISYS)
AI Seminar
UNISYS Knowledge Systems
Paoli Research Center
Paoli PA
SPEECH RECOGNITION USING CONNECTIONIST NETWORKS
Raymond Watrous
Siemens Corporate Research
and
University of Pennsylvania
The thesis of this research is that connectionist networks are
adequate models for the problem of acoustic phonetic speech
recognition by computer. Adequacy is defined as suitably high
recognition performance on a representative set of speech recognition
problems. Six acoustic phonetic problems are selected and discussed
in relation to a physiological theory of phonetics. It is argued that
the selected tasks are sufficiently representative and difficult to
constitute a reasonable test of adequacy.
A connectionist network is a fine-grained parallel distributed
processing configuration, in which simple processing elements are
interconnected by simple links. A connectionist network model for
speech recognition has been defined called the TEMPORAL FLOW MODEL.
The model incorporates link propagation delay and internal feedback to
express temporal relationships.
It has been shown that temporal flow models can be 'trained' to
perform successfully some speech recognition tasks. A method of
'learning' using techniques of numerical nonlinear optimization has
been demonstrated for the minimal pair "no/go", and voiced stop
consonant discrimination in the context of various vowels. Methods for
extending these results to new problems are discussed.
10:00am Wednesday, November 4, 1987
Cafeteria Conference Room
Unisys Paloi Research Center
Route 252 and Central Ave.
Paoli PA 19311
-- non-UNISYS visitors who are interested in attending should --
-- send email to finin@prc.unisys.com or call 215-648-7446 --
------------------------------
Date: Thu 29 Oct 87 01:08:12-PST
From: Ken Laws <Laws@KL.SRI.Com>
Reply-to: AIList-Request@SRI.COM
Subject: Conference - Expert Systems in Business and Finance
John Feinstein [(703) 934-3280] asked me to send out a notice
about the first annual conference on expert systems in business
and finance -- but I see that John Akbari submitted a description
in AIList V5 N246, Oct. 25. I'll just repeat that it's at the
Penta Hotel in New York City, November 10-12, 1987, $525. Call
(609) 654-6266.
------------------------------
End of AIList Digest
********************
