NEURON-Request@ti-csl.csc.ti.COM (NEURON-Digest moderator Michael Gately) (11/05/87)
NEURON Digest Thu Nov 5 08:19:28 CST 1987 - Volume 2 / Issue 27 Today's Topics: need info Books WANTED. Six Generation Computing Product Review Time-averaging of neural events/firings Shift-Invariant Neural Nets for Speech Recognition Names for Neuromorphic Systems Contest to name etc... (several) Workshop on Neural Computers Neural Networks Applied Optics Issue ---------------------------------------------------------------------- Date: 28 Oct 87 14:50:39 GMT From: Duc Tran <dgis!duc@lll-tis.arpa> Subject: need info HELP !! HELP !!! I need to get a hold of the following conference proceedings: Conference on Neural Networking for Computing, Snowbird, Utah (1986) If anybody can provide me pointers to where I can get it, I would be appreciated very much !! Duc Tran duc@dgis uunet!dgis!duc tel: 703-998-4647 ------------------------------ Date: 28 Oct 87 20:38:28 GMT From: John Eckrich <astroatc!johne@speedy.wisc.edu> Subject: Books WANTED. I am relatively new to neural networks/architectures and am interested in learning more. I would greatly appreciate any assistance you could provide in helping me in this endeavor. If you know of some good books, articles, or journals, etc. please send me some E-mail. 10Q in advance. ------------------------------------------------------------------------- Jonathan Eckrich | (rutgers, ames)!uwvax!astroatc!johne Astronautics Technology Center | ihnp4!nicmad!astroatc!johne Madison, WI | (608) 221-9001 ------------------------------ Date: 3 Nov 87 15:13:25 PST (Tuesday) Subject: Six Generation Computing From: Michael_R._Emran.OsbuSouth@xerox.com Can anybody out there direct me to find update information about Six- Generation Computing and the Japanese progress after the 1984's Proposals? I have a presentation next week for my Expert System course. All comments, info(s), or leads to any article is appreciated in advance. Mike ------------------------------ Date: Wed, 28 Oct 87 12:53:26 CST From: mnorton@rca.com Subject: product review I recently saw a presentation by Nestor on their Neural Network-based recognition systems. Fellow reader might be interested to known that they claim to have fielded 12 of their Handwriting Recognition Systems and that to the best of their knowledge (and mine), this is the first commercial application of a neural network. The presentation include a demonstration of handwriting recognition on a Toshiba Labtop. Future products will include object recognition from photographs (target recognition from aerial photography) and 3-D solid recognition. I suspect they are far ahead of any competition in terms of producting net-based products. The network model they use is a feedforward, three-layer, perceptron-like network which they call RCE (Reduced Coulomb Energy). It was mentioned that a paper might be included in the March 1988 issue of IEEE Computer (Special Issue on Neural Networks) which describes their model formally. Mark J. Norton RCA Advanced Technology Laboratories, AI Lab mnorton%henry@RCA.COM ------------------------------ Date: Thu, 29 Oct 87 15:53:23 est From: Michael Cohen <mike@bucasb.bu.edu> Subject: Time-averaging of neural events/firings We never did here at the Center for Adaptive Systems. Our architectures are far more general. You should look at a general bibliography. Michael Cohen ---- Center for Adaptive Systems Boston University (617-353-7857) Email: mike@bucasb.bu.edu Smail: Michael Cohen Center for Adaptive System Department of Mathematics, Boston University 111 Cummington Street Boston, Mass 02215 ------------------------------ Date: Fri, 30 Oct 87 20:31:32+0900 From: kddlab!atr-la.atr.junet!waibel@uunet.UU.NET (Alex Waibel) Subject: Shift-Invariant Neural Nets for Speech Recognition A few weeks ago, there was a discussion on AI-list, about connectionist (neural) networks being afflicted by an inability to handle shifted patterns. Indeed, shift-invariance is of critical importance to applications such as speech recognition. Without it a speech recognition system has to rely on precise segmentation and in practice reliable errorfree segmentation cannot be achieved. For this reason, methods such as dynamic time warping and now Hidden Markov Models have been very successful and achieved high recognition performace. Standard neural nets have done well in speech so far, but due to this lack of shift-invariance (as discussed on AI-list a number of these nets have been limping along in comparison to these other techniques. Recently, we have implemented a time-delay neural network (TDNN) here at ATR, Japan, and demonstrate that it is shift invariant. We have applied it to speech and compared it to the best of our Hidden Markov Models. The results show, that its error rate is four times better than the best of our Hidden Markov Models. The abstract of our report follows: Phoneme Recognition Using Time-Delay Neural Networks A. Waibel, T. Hanazawa, G. Hinton^, K. Shikano, K.Lang* ATR Interpreting Telephony Research Laboratories Abstract In this paper we present a Time Delay Neural Network (TDNN) approach to phoneme recognition which is characterized by two important properties: 1.) Using a 3 layer arrangement of simple computing units, a hierarchy can be constructed that allows for the formation of arbitrary nonlinear decision surfaces. The TDNN learns these decision surfaces automatically using error backpropagation. 2.) The time-delay arrangement enables the network to discover acoustic-phonetic features and the temporal relationships between them independent of position in time and hence not blurred by temporal shifts in the input. As a recognition task, the speaker-dependent recognition of the phonemes "B", "D", and "G" in varying phonetic contexts was chosen. For comparison, several discrete Hidden Markov Models (HMM) were trained to perform the same task. Performance evaluation over 1946 testing tokens from three speakers showed that the TDNN achieves a recognition rate of 98.5 % correct while the rate obtained by the best of our HMMs was only 93.7 %. Closer inspection reveals that the network "invented" well-known acoustic-phonetic features (e.g., F2-rise, F2-fall, vowel-onset) as useful abstractions. It also developed alternate internal representations to link different acoustic realizations to the same concept. ^ University of Toronto * Carnegie-Mellon University For copies please write or contact: Dr. Alex Waibel ATR Interpreting Telephony Research Laboratories Twin 21 MID Tower, 2-1-61 Shiromi, Higashi-ku Osaka, 540, Japan phone: +81-6-949-1830 Please send Email to my net-address at Carnegie-Mellon University: ahw@CAD.CS.CMU.EDU ------------------------------ Date: Fri 30 Oct 87 09:17:25-PST From: Ken Laws <LAWS@iu.ai.sri.com> Subject: Names for Neuromorphic Systems I'll vote for adaptive networks. I'm not sure that fits constraint relaxation via Hopfield networks or hill climbing with stochastic annealing, but it fits better than any of the other suggested terms. (I'm in the camp that sees no relation to simulation of neurons, other than the coincidence that biological neural networks have some capabilities that we would like to understand and then surpass.) -- Ken ------------------------------ Date: Fri, 30 Oct 87 11:10:09 CST From: im4u!rutgers!m.cs.uiuc.edu!matheus (Chris J. Matheus) Subject: Re: NEURON Digest - V2 / #26 This summer I picked up the following name for computer simulated neural networks: "Artificial Neural Systems" Unfortunately, I cannot identify the originator of the term. I simply recall hearing it used in a few presentations and reading it occasionally in papers. Other than being a bit long to say (it can be shortened to ANS's: "anzes"), the name seems appropriate in the way it captures the general flavor of this field of research. But this matter is not going to be decided by a simple vote. Rather, it will depend upon what name(s) end up being adopted in the literature and accepted by the scientific community at large. ------------------------------------------------------------------------------ Christopher J. Matheus usenet: {ihnp4, convex, philabs}!uiucdcs!matheus Inductive Learning Group arpa: matheus@a.cs.uiuc.edu University of Illinois csnet: matheus@uiuc.csnet Urbana, IL 61801 phone: (217) 333-3965 ------------------------------------------------------------------------------ ------------------------------ Date: Fri, 30 Oct 87 11:33:39 PST From: Dr Josef Skrzypek <skrzypek@cs.ucla.edu> Subject: Contest How about NEURONIA -- field of euphoric neuro-builders ------------------------------ Date: Fri, 30 Oct 87 21:41:59 PST From: Dr Jacques J Vidal <vidal@cs.ucla.edu> Subject: Contest to name etc... I have used - Neuromimetic Systems - Networks to designate artificial neural nets, plus "Neuromimetics" and, (in french), "Neuroinformatique" (Neuroinfomatics??) to refer to the whole field. However "Artificial Neural Networks" (ANNs) saeems OK and should appease the neuron modeling purists. PDP should be avoided. It apply just as well to models of computation that have almost no neuronal flavor. ------------------------------ Date: 4 November 87, 11:05 CET From: ECKMILLE%DD0RUD81.BITNET@wiscvm.wisc.edu Subject: Workshop on Neural Computers Dear Fellow Scientist, The proceedings of the NATO-Workshop (ARW) on NEURAL COMPUTERS in Neuss/Duesseldorf - 28.September - 2. October 1987 - will be published as: NEURAL COMPUTERS R. Eckmiller and C. v.d. Malsburg (eds.) at Springer-Verlag, Heidelberg The Book will be distributed in January 1988. During the Pre-Publication Sale you have the opportunity to order one or more copies for only $ 25 (25 US Dollars) if you send me the exact mailing address and a check before 10 December 1987. The official price as of January 1988 will be about $100. Please note that this book includes "Author Index", "Subject Index", and "Collection of References from all Contributions". The List of Contributers and the Table of Contents are enclosed for your information. Sincerely yours, Rolf Eckmiller, Ph.D. Department of Biophysics Universitaetsstr.1 D-4000 Duesseldorf, FRG Tel.(211)311-4540 cut here-------------------and mail this slip-------------------------------- I order _____ copies of the book NEURAL COMPUTERS for $ 25 each during the pre-publication sale (payment before 10 December 87). Please send the copies upon delivery (Jan.1988) to the following address:________________________________________________________ ________________________________________________________________ I enclose a check for ___US $ or transfer ___DM (corresponding to ___US $) to your account: No. 626 171 at SPARDA Bank Wuppertal, (Account Holder: Rolf Eckmiller, Ph.D.), Bankleitzahl 330 605 92. Signature_______________________ Date_______________1987 NEURAL COMPUTERS R. Eckmiller and C. v.d. Malsburg, eds. Springer-Verlag, Heidelberg (January 1988) LIST OF CONTRIBUTORS Akers, Lex A. (USA) Aleksander, Igor (UK) de Almeida, Luis B. (PORTUGAL) Anderson, Dana Z. (USA) Anninos, Photios (GREECE) Arbib, Michael A. (USA) Atlan, H. (ISRAEL) Barhen, J. (USA) Beroule, Dominique (FRANCE) Berthoz, Alain (FRANCE) Bienenstock, Elie (FRANCE) Bilbro, G.L. (USA) Buhmann, J (W.GERMANY) Caianiello, Eduardo R. (ITALY) Carnevali, P. (ITALY) Cotterill, Rodney M. J. (DENMARK) Daunicht, Wolfgang (W.GERMANY) Dress, William (USA) Dreyfus, Gerard (FRANCE) Droulez, J. (FRANCE) Eckmiller, Rolf (W.GERMANY) Feldman, Jerome A. (USA) Ferry, D.K. (USA) Fukushima, Kunihiko (JAPAN) Gardner, E. (UK) Garth, Simon (UK) Ginosar, R. (ISRAEL) Graf, H.P. (USA) Grondin, R.O. (USA) Gulyas, B. (BELGIUM) Guyon, I. (FRANCE) Hancock, P.J.B. (UK) Hartmann, Georg (W.GERMANY) Hecht-Nielsen, Robert (USA) Hertz, John (DENMARK) Hoffmann, Klaus-Peter (W.GERMANY) Huberman, Bernardo A. (USA) Iverson, L. (CANADA) Jorgensen, C.C. (USA) Koch, Christof (USA) Koenderink, Jan J. (NETHERLAND) Kohonen, Teuvo (FINLAND) Korn, Axel (W.GERMANY) Mackie, Stuart (USA) Mallot, Hanspeter (W.GERMANY) v. d. Malsburg, Christoph (W.GERMANY) Marinaro, M. (ITALY) May, David (UK) Moller P. (DENMARK) Moore, Will R. (UK) Negrini, R. (ITALY) Nylen, M. (DENMARK) Orban, Guy (BELGIUM) Palm, Guenther (W.GERMANY) Patarnello, Stefano (ITALY) Pellionisz, Andras J. (USA) Personnaz, L. (FRANCE) Phillips, William A. (UK) Reece, M. (UK) Ritter, Helge (W.GERMANY) Sami, M.G. (ITALY) Scarabottolo, N. (ITALY) Schulten, Klaus (W.GERMANY) Schwartz, D.B. (USA) v. Seelen, Werner (W.GERMANY) Sejnowski, Terrence J. (USA) Shepherd, Roger (UK) Singer, Wolf (W.GERMANY) Smith, L.S. (UK) Snyder, Wesley (USA) Stefanelli Renato (ITALY) Stroud, N. (UK) Tagliaferri, R. (ITALY) Torras, Carme (SPAIN) Treleaven, Philip (UK) Walker, M.R. (USA) Wallace, David J. (UK) Weisbuch, Gerard (FRANCE) White, Mark (USA) Willson, N.J. (UK) Zeevi, Joshua Y. (ISRAEL) Zucker, Steven (CANADA) Zuse, Konrad (W.GERMANY) ------------------------------ Date: Wed, 4 Nov 87 18:10 EDT From: MIKE%BUCASA.BITNET@wiscvm.wisc.edu Subject: Neural Networks Applied Optics Issue NEURAL NETWORKS: A special issue of Applied Optics December 1, 1987 (vol. 26, no. 23) Guest editors: Gail A. Carpenter and Stephen Grossberg The Applied Optics special issue on neural networks brings together a selection of research articles concerning both biological models of brain and behavior and technological models for implementation in government and industrial applications. Many of the articles analyze problems about pattern recognition and image processing, notably those classes of problems for which adaptive, massively parallel, fault-tolerant solutions are needed, and for which neural networks provide solutions in the form of architectures that will run in real-time when realized in hardware. The articles are grouped into several topics: adaptive pattern recognition models, image processing models, robotics models, optical implementations, electronic implementations, and opto-electronic implementations. Each type of neural network model is typically specialized to solve a variety of problems. Models of back propagation, simulated annealing, competitive learning, adaptive resonance, and associative map formation are found in a number of articles. Each of the articles may thus be appreciated on several levels, from the development of general modeling ideas, through the mathematical and computational analysis of specialized model types, to the detailed explanation of biological data or the fabrication of hardware. The table of contents follows. Single copies of this special issue are available from the Optical Society of America, at $18/copy. Orders may be placed by returning the form below, or by calling (202) 223-8130 (ask for Jeana Macleod). ------------------------------------------------------------------------------- Please send ____ copies of the Applied Optics special issue on neural networks (vol. 26, no. 23) to: NAME: __________________________________________________ ADDRESS: _______________________________________________ _______________________________________________ _______________________________________________ TELEPHONE(S):___________________________________________ TOTAL COST: $ ____________ $18/copy, including domestic or foreign surface postage (+ $10/copy for air mail outside U.S.) PAYMENT: _____ Check enclosed (payable to Optical Society of America, or OSA) or _____ Credit card: American Express ____ VISA ____ MasterCard ____ Account number __________________________________ Expiration date _________________________________ Signature (required) ____________________________ SEND TO: Optical Society of America Publications Department 1816 Jefferson Place NW Or call: (202) 223-8130 (Jeana Macleod) Washington, DC 20036 USA (credit cards) _______________________________________________________________________________ NEURAL NETWORKS: A special issue of Applied Optics December 1, 1987 (vol. 26, no. 23) Guest editors: Gail A. Carpenter and Stephen Grossberg TABLE OF CONTENTS ADAPTIVE PATTERN RECOGNITION MODELS Teuvo Kohonen. Adaptive, associative, and self-organizing functions in neural computing Gail A. Carpenter and Stephen Grossberg. ART 2: Self-organization of stable category recognition codes for analog input patterns Jean-Paul Banquet and Stephen Grossberg. Probing cognitive processes through the structure of event-related potentials during learning: An experimental and theoretical analysis Bart Kosko. Adaptive bidirectional associative memories T.W. Ryan, C.L. Winter, and C.J. Turner. Dynamic control of an artificial neural system: The Property Inheritance Network C. Lee Giles and Tom Maxwell. Learning and generalization in high order neural networks: An overview Robert Hecht-Nielsen. Counterpropagation networks Kunihiko Fukushima. A neural network model for selective attention in visual pattern recognition and associative recall IMAGE PROCESSING MODELS Michael H. Brill, Doreen W. Bergeron, and William W. Stoner. Retinal model with adaptive contrast sensitivity and resolution Daniel Kersten, Alice J. O'Toole, Margaret E. Sereno, David C. Knill, and James A. Anderson. Associative learning of scene parameters from images ROBOTICS MODELS Jacob Barhen, N. Toomarian, and V. Protopopescu. Optimization of the computational load of a hypercube supercomputer onboard a mobile robot Stephen Grossberg and Daniel S. Levine. Neural dynamics of attentionally modulated Pavlovian conditioning: Blocking, inter-stimulus interval, and secondary reinforcement OPTICAL IMPLEMENTATIONS Dana Z. Anderson and Diana M. Lininger. Dynamic optical interconnects: Volume holograms and optical two-port operators Arthur D. Fisher, W.L. Lippincott, and John N. Lee. Optical implementations of associative networks with versatile adaptive learning capabilities Clark C. Guest and Robert Te Kolste. Designs and devices for optical bidirectional associative memories Kelvin Wagner and Demetri Psaltis. Multilayer optical learning networks ELECTRONIC IMPLEMENTATIONS Larry D. Jackel, Hans P. Graf, and R.E. Howard. Electronic neural-network chips Larry D. Jackel, R.E. Howard, John S. Denker, W. Hubbard, and S.A. Solla. Building a hierarchy with neural networks: An example - image vector quantization A.P. Thakoor, A. Moopenn, John Lambe, and Satish K. Khanna. Electronic hardware implementations of neural networks OPTO-ELECTRONIC IMPLEMENTATIONS Nabil H. Farhat. Opto-electronic analogs of self-programming neural nets: Architectures and methodologies for implementing fast stochastic learning by simulated annealing Yuri Owechko. Opto-electronic resonator neural networks (Please Post this to Your Mailing List) ------------------------------ End of NEURON-Digest ********************