Copy Link
Add to Bookmark
Report
Neuron Digest Volume 02 Number 21
NEURON Digest Wed Sep 16 13:46:45 CDT 1987 Volume 2 / Issue 21
Today's Topics:
Copies of neural network letters from AIList Digest
Query: Connectionist Simulator - AIList V5 #196
Neural Networks - AIList V5 #169
Re: Neural Networks - AIList V5 #196
Re: Neural Networks - AIList V5 #196
Responses to connectionist simulator/inference query - AIList V5 #207
Neural Network Simulator - AIList V5 #207
Neural Network Simulator - AIList V5 #207
----------------------------------------
From: GATELY@TI-CSL.CSNET
Subject: Copies of neural network letters from AIList Digest
This whole Neuron Digest is devoted to resending a number of messages
which have been found in recent issues of AIList Digest. It is
quite difficult for me to decide whether to relay a message to
the Neuron subscribers which I find on AIList Digest because I
know that there is a large intersection of readership. However,
there are those who do not receive AIList Digest - and since this
is a digest on neural networks...
I will probably do this from time to time, with thanks to Ken Laws
(the moderator of AIList Digest [AIList@KL.SRI.COM]). For those
who think it is wasteful of resources to do this, my only suggestion
is that I will put some form of notice that this (and other issues
of this nature in the future) will be properly labeled.
------------------------------
Date: Fri, 7 Aug 87 08:37 EDT
From: Andre Marquis <Bodick@cis.upenn.edu>
Subject: Query: Connectionist Simulator - AIList V5 #196
I would like to experiment with connectionist inference mechanisms. Is
there a publically avaialble connectionist simulator? I have a Sun-3/160
with C and Common Lisp, among other things. I'm willing to port code from
other machines.
Also, if you have any good references on inference using connectionist
networks, please send them. Shastri's PhD thesis is the only thorough
treatment I've seen so far.
Thank you.
Andre Marquis
bodick@cis.upenn.edu
------------------------------
Date: 7 Aug 87 20:13:06 GMT
From: ihnp4!inuxc!iuvax!ndmath!milo@ucbvax.Berkeley.EDU (Greg Corson)
Subject: Neural Networks - AIList V5 #169
I am looking for some information and/or demo programs on Neural Networks
and how to simulate them on a computer. Any demo programs would be greatly
appreciated even if they don't do much.
I've heard that someone is selling a Neural Network simulator for the Mac
but I haven't heard much about it. If anyone has details on what the
program can do I would be interested to hear about it.
Also, I could use a few good introductory references on Nerual networks,
how they can be simulated and how to use them for pattern matching type
operations.
Thanks for the help...
Greg Corson
19141 Summers Drive
South Bend, IN 46637
(219) 277-5306 (weekdays till 6 central)
------------------------------
Date: 8 Aug 87 15:14:59 GMT
From: unc!hch@mcnc.org (Hong John Hsieh)
Subject: Re: Neural Networks - AIList V5 #196
In the 1st International Conference on Neural Network held in June,
there were a number of software/hardware products in exhibition.
Below is a short summary. Among the Macintoch-based systems,
Neuronics, inc. gave their contact phone number as 617-367-9254.
_______________________________________________________________________________
organization product host functions
_______________________________________________________________________________
(1) TRW Mark III coprocessor to vax hardware nn
(2) U. Colorodo MacTivation Mac simulator nn
(3) Nestor inc. Nestor PC-AT handwritten char rec.
(4) AIWARE inc. AINET PC process control
(5) Gen. Dynamics - - situation assessment,
weapon control
(6) HNC ANZA ANZA + PC-AT PC-AT/coprocssor
(7) MEIKO Computer net of transputers TSP, Image restoration
surface
(8) Neural Systems AWARENESS PC nn simulation
(9) Neuraltech inc. PLATO/ PC to Cray knowledge processor
ARISTOTLE
(10) Neuronics MacBrain Mac nn simulation
(11) SAIC,AI Tech. GINNI Symbolics 3670/CM,.. nn development tool
(12) SAIC,Tech. Res. Sigma-1 PC-AT nn simulation
(13) TI - Explorer+Odyssey nn simulation
(14) VERAC - BAMS simulation Assoc. Mem.
_______________________________________________________________________________
Cheng-Hong Hsieh (hch@unc.UUCP)
------------------------------
Date: 8 Aug 87 17:54:18 GMT
From: ramones.rutgers.edu!pratt@RUTGERS.EDU (Lorien Y. Pratt)
Subject: Re: Neural Networks - AIList V5 #196
At the Neural Networks tutorial at AAAI in Seattle in July, Terry
Sejnowski said that the third volume of Parallel Distributed Processing
will be released this coming fall. I have confirmed this with the
publisher. Sejnowski also said that the book will come with a disk
containing some sort of a neural network simulator which will run under
UNIX.
--Lorien Pratt
------------------------------
Date: Thu, 27 Aug 87 10:01 EDT
From: Andre Marquis <Bodick@cis.upenn.edu>
Subject: Responses to connectionist simulator/inference query - AIList V5 #207
Here is an edited summary of the replies I received for my query:
In article <8708071233.AA29422@linc.cis.upenn.edu> you write:
I would like to experiment with connectionist inference mechanisms.
Is there a publically avaialble connectionist simulator? I have a
Sun-3/160 with C and Common Lisp, among other things. I'm willing to
port code from other machines.
Also, if you have any good references on inference using connectionist
networks, please send them. Shastri's PhD thesis is the only thorough
treatment I've seen so far.
SIMULATORS:
>From: C Lynne D'Autrechy <mimsy!lynne@mimsy.umd.edu> on Mon 10 Aug
> 1987 at 9:49, 20 lines
>Stat: Read/Answered/Network
>To: bodick@cis.upenn.edu
>Subj: Re: Query: Connectionist Simulator
>A group of us here have developed a general-purpose simulator for developing
>and evaluating connectionist models. The system is currently written in Franz
>Lisp and runs on a MicroVAX (and should also run on a SUN supporting Franz
>Lisp).
>Lynne D'Autrechy
>From: mcvax!unizh!fuchs@seismo.CSS.GOV on Tue 11 Aug 1987 at 12:44, 19 lines
>Stat: Read/Network
>To: Bodick@cis.upenn.edu
>Subj: Re: Query: Connectionist Simulator
>Zoltan Schreter, Genetic Artificial Intelligence and Epistemics Laboratory,
>Faculty of Psychology and Educational Sciences, University of Geneva,
>Switzerland, has what he calls "a poor man's connectionist system" running -
>if I remember correctly - on an IBM PC.
> --- nef
>From: "John C. Akbari" <AKBARI@CS.COLUMBIA.EDU> on Thu 13 Aug 1987 at
> 22:46, 65 lines
>Stat: Read/Answered/Network
>To: bodick@cis.upenn.edu
>Subj: [Mail Delivery Subsystem <mailer-daemon@columbia.edu>: Returned mail:
> Host unknown]
> univ. rochester has a simulator for the sun & butterfly. you need to
>fill out a license agreement. costs about $150 (as of a few montsh ago,
>anyway). looks pretty neat from the description they sent. writen in c; uses
>sun graphics pacakge. contact: john costanzo (costanzo@cs.rochester.edu).
> univ. california (san diego) has a spiffy-looking simulator
>for the symbolics running release 7.1. costs $500 (educational rate). i
>believe the commericial rate is $2000 (!). for "quickie" overview, see
>chapter 13 in PDP. written in flavors, i think, so probably not too portable.
>contact
> kathy farrelly
> institute for cognitive science, c-015
> cognitive mechanisms group
> ucsd
> la jolla ca 92093
> 619.534.3359
> sorry, no email address
>please be advised that i haven't worked with either of these simulators.
>
>ainet-1 & ainet-2: call andy chun at brandeis (617.671.3652 or
>hon@brandeis.csnet) regarding these. about $150 for ainet & $300 for ainet-2.
>andy says ainet-1 is primarily a graphics tool & ainet-2 is for development.
>neither supports learning. also, THESE RUN UNDER RELEASE 6.1 ONLY.
REFERENCES:
My own:
%A Jerome A. Feldman
%A Dana H. Ballard
%A Christopher M. Brown
%A Gary S. Dell
%T Rochester Connectionist Papers: 1979-1985
%R University of Rochester Department of Computer Science Technical Report 172
%D December, 1985
%A Lokendra Shastri
%T Evidential Reasoning in Semantic Netowrks: A Formal Theory and its Parallel
Implementation
%R University of Rochester Department of Computer Science Technical Report 166
%D September, 1985
%A Lokendra Shastri
%A Jerome A. Feldman
%T Semantic Networks and Neural Nets
%R University of Rochester Department of Computer Science Technical Report 131
%D June, 1984
%A David E. Rumelhart
%A James L. McClelland
%T Parallel distributed processing
%V 1 & 2
%C Cambridge, Massachusetts
%P The MIT Press
%D 1987
%K Human information processing, cognition
>From: George Berg <berggeo@eecs.NWU.EDU> on Mon 10 Aug 1987 at 23:24, 74 lines
>Stat: Read/Answered/Network
>To: bodick@cis.upenn.edu
>Subj: Inference on Connectionist Networks
> I saw your posting to the net this afternoon. I don't know if this
>is exactly what you want but I saw two papers recently which roughly
>fit the category of "inference in a connectionist framework". Both of
>them are by Mark Derthick of the Department of Computer Science at
>Carnegie-Mellon University.
> "Counterfactual Reasoning with Direct Models," Proceedings of AAAI-87.
> "A Connectionist Architecture for Representing and Reasoning about
>Structured Knowledge," Proceedings of the Ninth Annual Conference of
>the Cognitive Science Society (1987).
>
> George Berg
> EECS Dept.
> Northwestern University
> Evanston, IL 60208
> berggeo@alpha.eecs.nwu.edu
> or
> berggeo@nucsrl.UUCP
> or
> ...!ihnp4!nucsrl!berggeo
>From: "John C. Akbari" <AKBARI@CS.COLUMBIA.EDU> on Thu 13 Aug 1987 at
> 22:46, 65 lines
>Stat: Read/Answered/Network
>To: bodick@cis.upenn.edu
>Subj: [Mail Delivery Subsystem <mailer-daemon@columbia.edu>: Returned mail:
> Host unknown]
>Two books not yet published:
>mcClelland & rumelhart. explorations in pdp: a handbook
>of models, programs & exercises. mit press.
>supposed to have some c programs to accompany the pdp books.
>
>macgregor, r.j. neural & brain modelling. academic press, nov. 87.
>supposed to have some fortran programs to simulate neuroelectrical activity.
>From: sdcsvax!ics.UCSD.EDU!cottrell@ucbvax.Berkeley.EDU on Fri 21 Aug
> 1987 at 0:01, 33 lines
>Stat: Read/Network
>To: CIS.UPENN.EDU!Bodick@ucbvax
>Subj: Re: Query: Connectionist Simulator
>Try my IJCAI 85 paper for an alternative way of doing inheritance inferences.
Andre Marquis
Department of Pathology and Lab Medicine
2 Gibson Building
Hospital of the University of Pennsylvania
3600 Spruce Street
Philadelphia, PA 19106
bodick@cis.upenn.edu
------------------------------
Date: 26 Aug 87 20:25:11 GMT
From: hao!boulder!mikek@ames.arpa (Mike Kranzdorf)
Subject: Neural Network Simulator - AIList V5 #207
I have seen inquiries around here about neural net simulators. I have
written a program called Mactivation which simulates single and double
layer networks which can be viewed as matrix-vector multipliers. If this
doesn't make sense you can still probably use the program. It assumes
minimal knowledge of things like activation values, state vectors, and
connection matrices. Things like delta rule, normalization, etc. are
included. Activation curves are (almost) completely under your control.
A nine page manual is included in the form of a MacWrite 4.5 document.
Version 2.0c was distributed at the ICNN conference in San Diego in June.
Since then I have made many improvements but not completed it as a product.
Since I have to get back to my research, I am releasing version 2.01 as is.
The program runs on any Mactintosh, as far as I know. It is public domain,
so you may be able to find it around. If you would like a copy, send either
1) a blank 3.5" disc and a 39 cent stamp or
2) a check for five dollars made out to me
to the address below. I format single-sided to be safe. Foriegn orders
are fine. I will probably upload it one of these days, or someone else
can when they get one. Source code is not available now, but it might
be someday. Happy trails.
Mike Kranzdorf
University of Colorado
Center for Opto-electronic Computing Systems
Campus Box 425
Boulder, Colorado 80309
(303) 492-8238
CSNET: mikek@boulder.colorado.edu
ARPA: mikek%boulder.colorado.edu@relay.cs.net
BITNET: mikek%boulder.colorado.edu@wiscvm.bitnet
UUCP: {hao, nbires}!boulder!mikek
Mactivation(tm) copyright 1987 Joint Optical Laboratory
"Once in while you get shown the light
in the strangest of places
if you look at it right"
------------------------------
Date: 25 Aug 87 01:56:52 GMT
From: cbosgd!mandrill!hal!uccba!finegan@ucbvax.Berkeley.EDU (Mike
Finegan)
Subject: Neural Network Simulator - AIList V5 #207
An Adaptive Template Matching Image Categorizer
(An Experimental Computer Vision Program)
[I can't forward the entire 72K file, particularly since I had to
set an AIList policy against sending out code. Those who want the
simulator can probably get it from the message author. -- KIL]
[NEURON Digest readers - the C code which was originally included
with this message was taken off by the AIList Digest moderator
due to its length. Although I hate to think of the amount of mail
requests he might get - you should probably 'order' it directly
from the original author. - MTG]
Hi,
Thanks for the response, this will be brief, but feel
free to send mail about neural nets, or anything related. It
may be that different machines require other header files ...
I have since been told that the original source is on Compuserve
in DDJFORUM -> C Chest DL. I have never used Compuserve ...
- Mike ...hal!uccba!finegan
I don't have a shell archiver, or anything like that, so
just cut out file 1, and then file 2. The remembered pattern
is identical to the training pattern format except for no
first (hdr) line. It will become clear if you read the Dr. Dobbs
article and this (or the original) program. Modified program
is followed by the original. I typed the original in by hand -
if there are uncorrected typo's, I apologize. I copied it and
then modified it (there were some typo's originally).
This file is ~69K, ~1/2 for each ...
#define PGM_ID "SILOAM CI-C86 Ver. 0f 11/22/86 for PC-DOS 2.x+"
/*
* An Adaptive Template Matching Image Categorizer
* (An Experimental Computer Vision Program)
*
* This program implements a trainable pattern classifier as
* a committee network of threshold logic units. It learns to
* recognize patterns by being trained from a set of prototype
* patterns presented in a training file. The training file is
* organized as a set of visual images represented as an orthogonal
* array of picture elements, or pixels. Each pixel is a number
* representing the gray-scale value of that point in the image.
* Associated with each pattern is a number, or tag, that
* represents the category to which that pattern belongs.
*
* R. J. Brown
* Elijah Laboratories International
* 5225 N.W. 27th Court
* Margata, Fl. 33063
* (305) 979-1567
*
* Ownership: I hereby place this program in the public domain.
*
* System: Red River ATlas 10 Mhz 80286 IBM-PC/AT clone
*
* Compiler: C86 Version 2.30H; Computer Innovations, Inc.
*
*/
------------------------------
End of NEURON-Digest
********************
