robina@castle.ed.ac.uk (Robin Allenson) (02/08/91)
I am trying to put together a short (15 min) talk on connectionism for some
people interested in AI, but not necessarily knowing a great deal about PDP.
Ideally, I want a short summary of the basic models, advantages, disadvantages,
and some applications. Can anyone suggest any books/articles that would do this
for me? If not, can anyone give email me what they think are the 'basics'.
The NN-contest asked for a def'n of NNs in a sentence. If anyone could
define connectionism in the above way in a paragraph, I'd be most interested.
If can you mail me, I'll mail the net a summary of the replies.
Thanks,
Robin Allenson
-------------------------------------------------------------------------------
| Robin Allenson | "If you find a fork in the road |
| | Take it. " |
| robina@castle.uk.ac.ed | "I'll do my .signature file tomorrow. |
| | No, really. " |
-------------------------------------------------------------------------------bhaskar@omega.cs.psu.edu (02/13/91)
In response to an earlier request for a summary of connectionism , I could find the following reference: "Connectionist Ideas and Algorithms" , Kevin Knight , CACM , Vol 33 , No 11, November , 1990 , pp 59-74. Hope this information is useful. Bhaskar Dasgupta.
robina@castle.ed.ac.uk (Robin Allenson) (04/19/91)
A couple of months back I posted a request for summaries of connectionism
for a talk I need to do on PDP and connectionism. Thanks to all those who
answered my mailing. I include here some of the (edited) replies, the
references that they gave. I did not include all the replies, but I'm pretty
sure that the references include all given to me. Sorry for the extreme delay
but things were very busy, followed by a month's Easter break when I had no
access to the net.
I will post on a script of my talk in due course, when I've got it all
typed up.
-------------------------------------------------------------------------------
The Replies
Date: Fri, 8 Feb 91 11:28:03 HST
From: "Ronald A. Amundson" <ronald@edu.hawaii.uhcc.uhunix>
Message-Id: <9102082128.AA10872@uhunix.uhcc.Hawaii.Edu>
Organization: University of Hawaii
There's a pretty nice summary written by the Churchlands in Scientific
American Jan 1990 p. 32 ff. It might not be what computer scientists
would want, because they're after the philosophical implications, but
it's a good clear intro to some main concepts.
Ron amundson
Date: Mon, 11 Feb 91 10:54:33 EST
From: Chester@edu.udel
Message-ID: <9102111054.aa20236@dewey.udel.edu>
Sender: Chester <Chester%edu.udel@edu.udel.dewey>
You might read the following article:
"Untangling Neural Nets" by Jeannette "Jet" Lawrence, in
Dr. Dobb's Journal, April, 1990, 38-44.
The commentary under the title on the Table of Contents reads
With more than 40 functioning neural net models to choose from,
it is important to understand their similarities and differences.
You also might be interested in the book
Neural Networks in Artificial Intelligence, by Matthew Zeidenberg,
Ellis Horwood Limited, 1990.
This book discusses neural networks, connectionism and their relationship
to AI. The chapter titles are:
Issues in Network Modeling
Neural Network Methods for Learning and Relaxation
Production Systems and Expert Systems
Knowledge Representation
Speech Recognition and Synthesis
Visual Perception and Pattern Recognition
Language Understanding
I hope these suggestions help.
Daniel Chester
University of Delaware
Department of Computer and Information Sciences
Newark, Delaware 19716 USA
chester@udel.edu
Date: Tue, 12 Feb 91 10:12:41 EST
From: Duane D Morin <dmorin@edu.wpi.wpi>
Message-Id: <9102121512.AA15727@wpi.wpi.edu>
Organization: Worcester Polytechnic Institute
Not being in the field myself I cant really give this one a review, but
AI Expert magazine just released (2 months ago?) a special issue dedicated to
Neural Networks. Seems they had like a 10part series running over the past
year in their magazine, and put together a special issue just for NN people.
I have it, and the first chapter or two are quite easy for a layman in the
field to understand.
Hope that helps.
Duane Morin WPI Worcester, MA 01609
Date: Tue, 12 Feb 91 17:38:03 -0800
From: Christopher N Toomey <toomey@edu.stanford.neon>
Message-Id: <9102130138.AA05299@Neon.Stanford.EDU>
Organization: Computer Science Department, Stanford University
There's an excellent overview of connectionism by Kevin Knight in
last November's Communications of the ACM. It sounds like about what
you're looking for.
Chris
Date: Tue, 12 Feb 91 18:51:09 cst
From: afreeman@edu.utexas.cs
Message-Id: <9102130051.AA20742@rub-al-khali.cs.utexas.edu>
Organization: U. Texas CS Dept., Austin, Texas
I'm relatively new to AI, but assuming that the term "connectionism"
refers to the study of massively parallel computer architectures, then a
good book would be The Connection Machine, which I belive is published
by MIT Press. I can find the call number for you if you have trouble
finding it.
--------------------------------------------------------------------------------
References
Scientific American January 1990, p32 ff
"Untangling Neural Nets" by Jeannette "Jet" Lawrence, in
Dr. Dobb's Journal, April, 1990, 38-44.
As Dr. D's puts it:
With more than 40 functioning neural net models to choose from,
it is important to understand their similarities and differences.
Neural Networks in Artificial Intelligence, by Matthew Zeidenberg,
Ellis Horwood Limited, 1990.
Discusses neural networks, connectionism and their relationship
AI Expert magazine (4 months ago?)
A special issue dedicated to Neural Networks. Seems they had like a 10part
series running over the past year in their magazine, and put together a
special issue just for NN people.
Communications of the ACM, Nov 1989
Article by Kevin Knight, contains an excellent overview of connectionism.
The Connection Machine
by Daniel Hillis, MIT Press.
All about the study of massively parallel computer architectures, and one in
particular. Originally Daniel Hillis' PhD thesis I think. Pretty easy read, with
lots of useful stuff about the failures of the von Neumann computer, and how
parallel processing gets away from them.
Parallel Distibuted Processing
by PDP Research Group, MIT Press.
A two volume work - if you want a good intro into PDP, with quite a bit of
theory as well, then I'd recommend this. The collection of papers within covers
most PDP research, although it might be a bit out of date in a few areas now.
--------------------------------------------------------------------------------
Robin Allenson \ Yesterday I had a premonition that today I would
robina@uk.ac.ed.castle \ have a sense of deja-vu.
--------------------------------------------------------------------------------