neuron-request@HPLMS2.HPL.HP.COM ("Neuron-Digest Moderator Peter Marvit") (08/24/90)
Neuron Digest Thursday, 23 Aug 1990 Volume 6 : Issue 50 Today's Topics: neuron digest submission (protein folding) Help with reference. Re: Help with reference. Papers on applications of N.N.s to constraint satisfaction problem Re NN-Definition Language SF Bay area - AI Forum Meeting Post-Doc positions in U.K. Re: request for simple NN program (LONG) Send submissions, questions, address maintenance and requests for old issues to "neuron-request@hplabs.hp.com" or "{any backbone,uunet}!hplabs!neuron-request" Use "ftp" to get old issues from hplpm.hpl.hp.com (15.255.176.205). ------------------------------------------------------------ Subject: neuron digest submission (protein folding) From: smuskal%calv01.hepnet@Csa2.LBL.Gov Date: Sun, 19 Aug 90 15:24:18 -0700 Recently, the world of neural networks has cast itself upon molecular biology's most challenging computational problems, "The protein Folding Problem." Our lab and others have applied feed-forward, supervised networks to the most obvious subproblems such as secondary structure predictions, solvent accessibility predictions, disulfide-bonding state predictions, and small tertiary structure predictions. Aside from using similar type neural networks, each approach to these subproblems has one thing in common, mapping LOCAL amino acid sequence to some protein structural feature. And while people have experimented with representation of the 20 common amino acids, still no one has provided any good method of representing the ENTIRE amino acid sequence for predictive purposes. This is, in essence, the key problem since a protein's folded structure depends on ALL of the competing interactions within and surrounding the COMPLETE amino acid sequence. Being experts in representation, perhaps some of you have ideas on describing the complete amino acid sequence of a protein in as few computational nodes as possible. Keep in mind that there are 20 common amino acids and that most proteins of interest are over 100 amino acids long. Steve smmuskal@lbl.gov ------------------------------ Subject: Help with reference. From: "Paulo V. Rocha" <P.Rocha@cs.ucl.ac.uk> Date: Tue, 21 Aug 90 09:15:55 +0100 I wonder if anyone could help me finding a journal. The reference I have is somewhat incomplete but ... Bulletin of the Electrotechnical Laboratory Vol 53, No 10 (1989) I suppose the referred lab is one of the NTT labs in Japan. The article I am interested on is Toward Soft Logic for the Foundation of Flexible Information Processing Nobuyuki OTSU I would appreciate if someone could supply the author's address, email or any means by which I could reach him or even better :-), where I could find the journal itself (here in London it is being very hard, none of the Univ. libraries seem to have it) The article is said to be an edited English Version of the Japanese papers (if it helps...) Towards Soft Logic N. Otsu & K. Tamura Journal of IPS (Information Processing Society?), 28 629-636 (1987) Soft Logic for Recognition and Understanding - Recapitulation of pattern Recognition Journal of IEICE (?), 71, 1231-1240 (1988) Thanks for your time, P. +-----------------------------+---------------------------------------------+ Paulo Valverde de L. P. Rocha | JANET:procha@uk.ac.ucl.cs Department of Computer Science| BITNET:procha%uk.ac.ucl.cs@UKACRL University College London |Internet:procha%cs.ucl.ac.uk@nsfnet-relay.ac.uk Gower Street | ARPANet:procha@cs.ucl.ac.uk London WC1E 6BT | UUCP:...!mcvax!ukc!ucl-cs!procha England | tel: +44 (071) 387 7050 x 3719 | fax: +44 (071) 387 1397 +-----------------------------+---------------------------------------------+ ------------------------------ Subject: Re: Help with reference. From: yakiyama@etl.go.jp (Yutaka Akiyama) Organization: Electrotechnical Laboratory, Tsukuba Science City Date: 22 Aug 90 01:19:14 +0000 Dear Dr. Rocha, I'm writing from the Electrotechnical Laboratory... In article <1128@ucl-cs.UUCP> P.Rocha@ucl-cs.UUCP writes: > I wonder if anyone could help me finding a journal. > The reference I have is somewhat incomplete but ... > > Bulletin of the Electrotechnical Laboratory > Vol 53, No 10 (1989) You can reach the author by: e-mail: otsu@etl.go.jp snail: Dr. Nobuyuki Otsu Mathematical Informatics Section, Information Science Division, ETL Umezono 1-1-4, Tsukuba, 305, JAPAN > I suppose the referred lab is one of the NTT labs in Japan. The Electrotechnical Laboratory (ETL) belongs to the Agency of Industrial Science and Technology (AIST) of the Ministry of International Trade and Industry (MITI). The largest national research institute in Japan was founded in 1891 and is dedicated to basic research and development in the fields such as electronics, information processing, energy technology and standards and measurements. Dr. Otsu is currently the chief of the Math. Info. Section, and serves concurrently as the head researcher of the ETL. I'm sure to relay your request to him. (^_^) Yutaka Akiyama (yakiyama@etl.go.jp) Computation Models Section, Computer Science Division, Electrotechnical Laboratory, Umezono, Tsukuba Science City, 305 JAPAN Interest: Optimization by Neural Networks, VLSI impl. of Gaussian Machines ------------------------------ Subject: Papers on applications of N.N.s to constraint satisfaction problem From: qian@icopen.ICO.OLIVETTI.COM (DA QUN QIAN) Date: Tue, 21 Aug 90 15:58:29 +0100 I am looking for the papers on how to use neural nets to solve constraint satisfaction problem and how to generate production rules or other represenations of knowledge from neural networks. I would be grateful if you can offer me such information. Best regards. Qian Da Qun Email: qian@icopen.ico.olivetti.com Olivetti Artificial Intelligence Center Nuova ICO 3 Piano Via Jervis 77, 10015 Ivrea(TO) Italy ------------------------------ Subject: Re NN-Definition Language From: Gary Fleming <72260.2544@compuserve.com> Date: 21 Aug 90 13:13:32 -0400 Date: 21 August 1990 From: Gary Fleming 72260.2544@CompuServe.COM Organization: American Electronics, Inc., and International Neural Network Society / SIG Washington Dear Ms. Thalmann and Mr. Almassy, One item I found missing from your definition language was any mention of the nodal activation (or transfer function). I assume you have implemented your network with the logistic activation function f(x) = 1/(1+exp(-x)). However, is this true of the nodes comprising the input and output layers? You might investigate the necessity (and desirability) of using anything but a linear activation function for the input-output layers. It is my belief that the logistic function is not a particularly good choice for the activation function of the hidden layer nodes, either. It is this premise that has forced me to generate my own NN software rather than purchase an off-the-shelf product. You might consider how the user could specify an arbitrary activation function (and its derivative) for each of the nodes comprising your network. I think the observation that every researcher wishes to do something a little differently will quickly render any NN-definition language non-universal. However, on a broader stance, I must say that a general purpose language is premature. A more desirable level of organization is the establishment of a common artificial neural network (ANN) vocabulary to be shared by all of the researchers and practitioners who represent many diverse fields. As you know, electrical engineers, neurophysiologists, cognitive psychologists, mathematicians, et. al., can easily use different terms for the same concept or (worse) use the same term for differing concepts. This intellectual Tower of Babel hobbles the sister field of nonlinear systems theory (chaos) as well. There are several members of the INNS/SIG Washington interested in a resolution to this problem. At any rate, a universally accepted NN-definition language should implement universally accepted ANN concepts and methodologies. Gary Fleming 72260.2544@CompuServe.COM ------------------------------ Subject: SF Bay area - AI Forum Meeting From: kingsley@hpwrc02.hp.com Date: Tue, 21 Aug 90 15:52:37 -0700 ************************************************************** * * * A I F O R U M M E E T I N G * * * * * * SPEAKER: Andreas Wiegend * * TOPIC: Pruning neural nets and predicting * * Sun Spots * * WHEN: 7PM Tuesday 8/28/90 * * WHERE: Lockheed building 202, auditorium * * 3251 Hanover Street * * Palo Alto, CA * * * * AI Forum meetings are free, open and monthly! * * Call (415) 594-1685 for more info * ************************************************************** ------------------------------ Subject: Post-Doc positions in U.K. From: PSS001%VAXA.BANGOR.AC.UK@VMA.CC.CMU.EDU Date: Wed, 22 Aug 90 18:47:17 +0000 Department of Psychology, University of Wales, Bangor and Department of Psychology, University of York CONNECTIONISM AND PSYCHOLOGY THREE POST-DOCTORAL RESEARCH FELLOWSHIPS Applications are invited for three post-doctoral research fellowships to work on the connectionist and psychological modelling of human short-term memory and spelling development. Two Fellowships are available for three years, on an ESRC- funded project concerned with the development and evaluation of a connectionist model of short-term memory. One Fellow will be based with Dr. Gordon Brown in the Cognitive Neurocomputation Unit at Bangor and will be responsible for implementing the model. The other Fellow, based at York with Dr. Charles Hulme, will be responsible for undertaking psychological experiments with children and adults to evaluate the model. Starting salary for both posts on research 1A grade up to # 13,495. One two-year Fellowship is available to work on an MRC-funded project to develop a sequential connectionist model of the development of spelling and phonemic awareness in children. This post is based in Bangor with Dr. Gordon Brown. Starting salary on research 1A grade up to # 14,744. Applicants should have postgraduate research experience or interest in cognitive psychology/cognitive science or connectionist/ neural network modelling and computer science. Good computing skills are essential for the posts based in Bangor, and experience in running psychological experiments is required for the York-based post. Excellent computational and research facilities will be available to the successful applicants. The appointments may commence from 1st. October 1990, but start could be delayed until 1st. January 1991. Closing date for applications is 7th. September 1990, but intending applicants should get in touch as soon as possible. Informal enquiries regarding the Bangor-based posts, and requests for further details of the posts and host departments, to Gordon Brown (0248 351151 Ext 2624; email PSS001@uk.ac.bangor.vaxa); informal enquiries concerning the York-based post to Charles Hulme ( 0904 433145; email ch1@uk.ac.york.vaxa). Applications (in the form of a curriculum vitae and the names and addresses of two referees) should be sent to Mr. Alan James, Personnel Office, University of Wales, Bangor, Gwynedd LL57 2DG, UK. (Apologies to anyone who receives this posting through more than one list or newsgroup) ------------------------------ Subject: Re: request for simple NN program (LONG) From: chuck@utkux1.utk.edu (chuck) Organization: University of Tennessee Computing Center, Knoxville Date: 22 Aug 90 21:05:06 +0000 [[ Editor's Note: Every so often, a series of requests comes for public domain or other software to run neural network simulations. The list below is primarily for UNIX machines, though some DOC and MAC code is included. No commercial packages are listed. I'm sure we all thanks Chuck for the effort he put in. -PM ]] I had occasion recently to compile this list as an appendix to my MS thesis. Enjoy, but please try to be considerate of the authors. If you want one at your site, have it installed by your system folks (really only true of the larger ones. . .) to prevent multiple copies and mass FTP loads! Enjoy. It's been fun! chuck ________ A list of publicly available simulators for artificial neural networks. The following list is by no means to be considered complete or exhaustive, it merely represents those simulators which have come to my attention through postings on the internet and articles in various journals. Inclusion in this list represents no particular endorsement, and ommission from the list should be regarded as unfortunate oversight on the author's part. The list is provided as a starting point for individuals and institutions who desire to obtain software suitable to their needs who wish to examine these offerings. The term "publically available" may be taken to mean anything available at nominal or no cost through channels other than those conventionally regarded as commercial software products. Since the term 'public domain' has become synonimous with free and unrestricted software it should be noted in passing that much of what is available is, in fact, copyrighted work which must be licensed --albeit at no cost. The term 'ftp' used below stands for file transfer protocol, which is a means of obtaining and transfering files between computers linked by an international computer network collectively know as the Internet. To obtain more information concerning ftp or the Internet, please consult operators of your local computing facility. The format used for site addresses and electronic mail addresses below conforms to that currently used on the Internet. I have included alternate means of obtaining the simulators where they have come to my attention. ++++++++++++++++++++++++ bps - George Mason University Back Prop Simulator Current version is 1.01 (Nov., 1989) A special-purpose simulator for Back Propagation and a BP speedup technique called 'gradient correlation' [IJCNN, Jan., 1990]. Available via anonymous ftp from gmuvax2.gmu.edu (129.174.1.8). Distributed as executables for VAX 8530 under Ultrix 3.0, and versions for 8088 based IBM PC, and 80286/386 IBM PC machines. Includes examples and a tutorial document. Source code license is available. Contact: Eugene Norris (703) 323-2713 Computer Science Department enorris@gmuvax2.gmu.edu George Mason University Fairfax, Virginia 22032 ++++++++++++++++++++++++ Back Propagation simulator [Name and version not known.] A special-purpose simulator for back propagation using several training methods. Distributed on disk for the IBM PC, with mouse and VGA or EGA display. To obtain it, send a stamped, self-addressed envelope and a floppy disk (3.5" or 5.25") to: Universitt Kassel Fachbereich Mathematik Forschungsgruppe Neuronale Netzwerke Heinrich-Plett-Str 40 3500 Kassel West Germany [Report taken from Internet News, October, 1989. ] ++++++++++++++++++++++++ MIRRORS/II -- Maryland MIRRORS/II Connectionist Simulator A general-purpose connectionist simulator. To obtain this simulator you must sign an institutional site license. A license for individuals is not acceptable. The only costs incurred are for postage for a printed copy of the manual and tape cartridge (you send your own 1/4" cartridge or TK50 cartridge to them, if desired.) Instructions for obtaining the software via ftp are returned to you upon receipt of the license agreement. To obtain a copy of the license send your physical mail address via e-mail to: mirrors@cs.umd.edu or by U.S. Mail to: Lynne D'Autrechy University of Maryland Department of Computer Science College Park, MD 20742 MIRRORS/II is implemented in Franz Lisp and will run under Opuses 38, 42, and 43 of Franz Lisp on UNIX systems. It is currently running on a MicroVAX, VAX and SUN 3. ++++++++++++++++++++++++ NeurDS -- The Neural Design and Simulation System. Current Version is 3.1 (May, 1989.) A general purpose simulator. The system is licensed on a no-fee basis to educational instutions by Digital Equipment Corporation. To obtain information, send your physical or electronic mail address to: Max McClanahan mcclanahan%cookie.dec.com@decwrl.dec.com Digital Equipment Corporation 1175 Chapel Hills Drive Colorado Springs, Colorado 80920-3952 You should receive instructions on how to obtain a copy of the manual and copies of the license agreement. [Beyond receipt of the license agreement, I do not know the details of distribution.] The NeurDS system will run on any Digital platform including Vax/VMS, Vax/Ultrix, and DECsystem/Ultrix. A graphics terminal is not required to support the window interface. Specific models are described using a superset of the C programming language, and compiled into a simulator form. This simulator can accept command scripts or interactive commands. Output can take the form of a window-type envirionment on VT100 terminals, or non-window output on any terminal. ++++++++++++++++++++++++ FULL -- Fully connected temporally recurrent neural networks. A demonstration network described in "Learning State Space Trajectories in Recurrent Neural Networks" [no other reference material!] The author (whose name is Barak Pearlmutter of the Journal of Neural Computation, 'bap@f.gp.cs.cmu.edu') describes this as "a bare bones simulator for [. . .] temporally recurrent neural networks" and claims that it should vectorize and parrallelize well. It is available for ftp from doghen.boltz.cs.cmu.edu. Login as 'ftpguest' password 'oaklisp'. Be sure to ftp as binary for the file 'full/full.tar.Z' (you must either use a directory named full on your local machine, or use 'get' and let it prompt you for remote and local file names.) Do not attempt to change directories. It is copyrighted and is given out for academic purposes. [The information dates from November of 1988.] ++++++++++++++++++++++++ GRADSIM Connectionist Network Simulator. A special-purpose simulator specifically designed for experiments with the temporal flow model. Latest Version 1.7 The simulator is available for anonymous ftp from ai.toronto.edu (128.100.1.65). For information contact: Raymond Watrous watrous@ai.toronto.edu Department of Computer Science University of Toronto Toronto, Ontario M5S 1A4 In C, implementations on VAX (VMS & Ultrix), Sun and CYBER are mentioned. A graphical interface is 'under development.' (March, 1988.) Includes an excellent article with references. ++++++++++++++++++++++++ opt -- A Neural-Net Training Program Based on Conjugate-Gradient Optimization. A special-purpose simulator. [Current version unknown.] Available for anonymous ftp from cse.ogc.edu (129.95.40.2) from the directory "/ogc2/guest/ftp/pub/nnvowels". Consult the file 'README' for more instructions. For further information, you might contact 'opt-dist@cse.ogc.edu'. Basically C code to be compiled under BSD Unix, with no graphic interface. They do maintain a list of users, perhaps a mailing list. An unsigned paper describing the technique and use of the simulator is included. ++++++++++++++++++++++++ CasCor1 -- Cascade-Correlation Simulator A special-purpose simulator for experimenting with the Cascade-Correlation algorithm described in: Fahlman, Scott, and C. Lebiere. "The Cascade-Correlation Learning Architecture." In _Advances in Neural Information Processing Systems 2_, edited by D. S. Touretzky. New York : Morgan Kaufman Publishers, 1990. It is available for anonymous ftp from pt.cs.cmu.edu (128.2.254.155) in the directory "/afs/cs/project/connect/code" (subdirectories may be available, but parent directories may not be.) There are Lisp and C versions available, as well as several other programs. The simulator is placed in the public domain. For information contact: Scott E. Fahlman fahlman@cs.cmu.edu School of Computer Science Carnegie-Mellon University Pittsburgh, PA 15217 The original version by Scott Fahlman was written in Common Lisp and has been tested on CMU Common Lisp on the IBM RT, Allegro Common Lisp (beta test) for Decstation 3100, and Sun/Lucid Common Lisp on the Sun 3. This program was translated into C by Scott Crowder. ++++++++++++++++++++++++ GENESIS - GEneral NEural SImulation System with XODUS - X-windows Output and Display Utility for Simulations A general simulator. Currently Beta-Test Version, 1990. From the release announcement ( January, 1990 by Jim Bower ): " Full source for the simulator is available via FTP from genesis.cns.caltech.edu (131.215.135.64). To acquire FTP access to this machine it is necessary to first register for distribution by using telnet or rlogin to login under user "genesis" and then follow the instructions (e.g. 'telnet genesis.cns.caltech.edu' and ' login as 'genesis'). When necessary, tapes can be provided for a small handling fee ($50). Those requiring tapes should send requests to genesis-req@caltech.bitnet. Any other questions about the system or its distribution should also be sent to this address. GENESIS and XODUS are written in C and run on SUN and DEC graphics work stations under UNIX (version 4.0 and up), and X-windows (version 11). The software requires 14 meg of disk space and the tar file is approximately 1 meg. The current distribution includes full source for both GENESIS and XODUS as well as three tutorial simulations (squid axon, multicell, visual cortex). Documentation for these tutorials as well as three papers describing the structure of the simulator are also included. As described in more detail in the "readme" file at the FTP address, those interested in in developing new GENESIS applications are encouraged to become registered members of the GENESIS users group (BABEL) for an additional one time $200 registration fee. As a registered user, one is provided documentation on the simulator itself (currently in an early stage), access to additional simulator components, bug report listings, and access to a user's bulletin board. In addition we are establishing a depository for additional completed simulations. " ++++++++++++++++++++++++ SunNet A generalized simulator. Version 5.5.2.4 Available for anonymous ftp from boulder.colorado.edu (128.138.240.1). While this program was obviously written for Sun workstations (versions for suntools and the X-window envirionment), the documents list other configurations. These include a non-graphic version which runs on "any UNIX machine", and versions which run on an Alliant or UNIX machine and send data to a graphics support program running on a Sun workstation. It is very easy to install. A mailing list exists for users of the simulator. ++++++++++++++++++++++++ RCS - The Rochester Connectionist Simulator A general simulator. Version 4.2 Available for anonymous ftp from cs.rochester.edu (192.5.53.209). Tapes may be purchased (1600 bpi 1/2" reel or QIC-24 Sun 1/4" cartridge) from: Peg Meeker Computer Science Department University of Rochester Rochester, New York 14627 C source code is provided, including a graphic interface which may function under X Windows or SunView on Sun Workstations. A wide variety of Unix machines are supported, and the simulator may be used without the graphics interface. A version for the MacIntosh is included in the distribution. Mailing lists exist for users and bug reports. ++++++++++++++++++++++++ SFINX -- Structure and Function in Neural Connections A General Simulator. Version 2.0 ( November, 1989 ) In order to ftp this simulator, a license agreement must be submitted. Upon receipt of this agreement, instructions and the password to ftp the software are made available. To obtain the license write: Machine Perception Laboratory Computer Science Department University of California Los Angeles, CA 90024 This system requires color to operate the graphics interface, but may be operated without graphics. Support for Sun, Ardent Titan, HP 300, and IBM PC RT machines is specifically mentioned --but other Unix platforms should function as well. Specific graphics support is provided for Matrox VIP 1024, Imagraph AGC-1010P, HP Starbase and X Windows. A version providing support for monochrome graphics is expected to be released in Fall, 1990. ++++++++++++++++++++++++ Mactivation A specialized simulator for investigating associative memory using the delta rule and Hebbian Learning. Version 3.2 A public domain version is available for anonymous ftp from the University of Colorado at Boulder (boulder.colorado.edu, 128.138.240.1) or possibly by contacting the author. Mike Kranzdorf University of Colorado Optoelectronic Computing Systems Center Campus Box 525 Boulder, Colorado 80309-0525 mikek@boulder.colorado.edu Future versions will probably not be public domain, but will be availble from Oblio, Inc., 5942 Sugarloaf Road, Boulder, Colorado 80309. Provided as executables for the Apple MacIntosh. ++++++++++++++++++++++++ Several special purpose simulators are provided with the following book: McClelland, J. L., and David.E. Rumelhart. _Explorations in Parallel Distributed Processing_. Cambrige: MIT Press, 1988. Versions exist which contain C code for the IBM PC, and a version has recently been released for the Apple MacIntosh. ++++++++++++++++++++++++ Hopfield-style Network Simulator A Special Purpose simulator for experimentation with the Hopfield-style network developed by the author. Software is available by e-mail upon request from the author, Arun Jagota (jagota@cs.buffalo.edu). It is written in C and should be useful on 32-bit Unix machines, and a MSDOS version is also supplied. ++++++++++++++++++++++++ Several demonstration-type simulators have been published as source code in various journals. These are cited below: Brown, Robert Jay. "An Artificial Neural Network Experiment." _Dr. Dobb's Journal_ (April, 1987) 16ff. Colvin, Gregory. "Synapsys: A Neural Network." _The C Users Journal_ (April, 1989) 59ff. King, Todd. "Using Neural Networks for Pattern Recognition." _Dr. Dobb's Journal_ (January, 1989) 14ff. Klimasauskas, Casey. "Neural Nets and Noise Filtering." _Dr. Dobb's Journal_ (January, 1989) 32ff. Jones, Wiliam P., and Josiah Hoskins. "Back-Propagation." _Byte_ (October, 1987) 155ff. ------------------------------ End of Neuron Digest [Volume 6 Issue 50] ****************************************