neuron-request@hplabs.hpl.hp.com ("Neuron-Digest Moderator Peter Marvit") (02/27/90)
Neuron Digest Monday, 26 Feb 1990 Volume 6 : Issue 17 Today's Topics: FYI: super neurocomputer Translation-Invariant Pattern Recognizer NN's Re: Translation-Invariant Pattern Recognizer NN's Re: Translation-Invariant Pattern Recognizer NN's Re: Translation-Invariant Pattern Recognizer NN's Edelman 3rd Intn'l Symposium on AI (Mexico) Neocognitron (revised) Code NEW JOURNAL: MINDS & MACHINES 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: FYI: super neurocomputer From: fletcher@hplabs.hp.com Date: Thu, 11 Jan 90 09:48:06 -0800 [[ Editor's Note: A friend of mine sent this to me and I haven't looked up the reference. Does anyone have any additional information about Fujitsu? -PM ]] this is from a newswire abstract service I get: *-*3 Fujitsu unveils fast neurocomputer Fujitsu has developed a `super neurocomputer' with processing speedsthat are several hundred times faster than those of models currently on the market. The company added that it hoped to commercialise the new machine in 1991. International Herald Tribune. English. 1990-01-09. Size: 3 c. Wall Street Journal/Europe. English. 1990-01-09. Size: 1 c The 8 Jan 90 US WSJ is the issue most likely to have this story. ------------------------------ Subject: Translation-Invariant Pattern Recognizer NN's From: linco@eng.umd.edu (Sam Lin) Organization: Merriversity of Uniland, College Purgatory Date: 08 Feb 90 00:03:19 +0000 In their paper, "Neural Nets for Adaptive Filtering and Adaptive Pattern Recognition (IEEE Computer, March 1988)," Bernard Widrow and Rodney Winter describe a feed-forward network for pattern recognition, whose responses are invariant to translation, rotation, and scale change of the input pattern. This network is based on multiple ADALINE neurodes, and the Madaline rule II (MRII) training method. Does anyone know where I can find further information on this network? Is there a follow-up article? Sources of information on the translation/rotation/scale invariance of backpropagation (or other networks) in pattern recognition would also be appreciated. Thanks in advance. Sam Lin Internet: linco@eng.umd.edu UUCP: ..!uunet!eng.umd.edu!linco ------------------------------ Subject: Re: Translation-Invariant Pattern Recognizer NN's From: cs.utexas.edu!usc!elroy.jpl.nasa.gov!aero!robert@tut.cis.ohio-state.edu (Bob Statsinger) Organization: Ahh...wouldntya like to know! Date: 15 Feb 90 20:14:45 +0000 A different approach to Invariant Pattern Recognition may be found in Christoph von der Malsburg's Dynamic Link Architecture. Malsburg also uses a layered architecture, but his approach is self-organization. Patterns are represented as labeled topological graphs composed of the *correlation* behavior of neighboring cells in a layer. Pattern recognition is then implemented through graph matching. The correlational graphs express neighborhood relationships between features in the input and yield a robust invariance to translation, scaling, or distortion. See "Pattern Recognition by Labeled Graph Matching" Neural Networks, vol. 1, pp 141-148, 1988 Bob Statsinger Robert@aerospace.aero.org The employers expressed herein are strictly mine and are not necessarily those of my opinion's....uh..er...whatever... ------------------------------ Subject: Re: Translation-Invariant Pattern Recognizer NN's From: snorkelwacker!usc!elroy.jpl.nasa.gov!aero!aerospace.aero.org!plonski@tut.cis.ohio-state.edu (Mike Plonski) Organization: The Aerospace Corporation Date: 15 Feb 90 23:55:29 +0000 You can also look at the following refernces which I sliced out of my reference file in tib (sort of bib) style format. The following abreviations may help. The %X are my comments and are not intended as a complete summary. - ---- ABREV. ---- ba:~ (52) tibabb neunet D NEUNET Neural Networks\ ba:~ (53) tibabb icnn88 D ICNN88 IEEE International Conference on Neural Networks\ (55) tibabb icnn87 D ICNN87 IEEE First International Conference on Neural Networks\ ba:~ (57) tibabb aip151 D AIP151 AIP Conf. Proc. 151 - Neural Networks for Computing\ - ------REF.-------------------- ba:~ (56) tiblook invariant %T Invariant Pattern Recognition by Means of Fast Synaptic Plasticity %A Jochim Buhmann %A Klaus Schulten %J |ICNN88| %V I %P 125-13 %A H. H. Chen %A Y. C. Lee %A G. Z. Sun %A H. Y. Lee %A T. Maxwell %A C. L. Giles %T High Order Correlation Model for Associative Memory %J |AIP151| %P 86-99 %K symmetry invariant pattern recognition %X Gives Lyapunov function for high order nets, shows convergence of high order for synchronous and asynchronous schems, uses character recognition with shift invariance, shows higher order models can store more patterns than second order (normal) correlation models for assoc. memory 2nd order storage for hopfield net is limited to N/(4*log(N)) %T Neocognitron: A Hierarchical Neural Network Capable of Visual Pattern Recognition %A Kunihiko Fukushima %J |NEUNET| %V 1 %N 2 %P 119-130 %D 1988 %K scale, position invariant %A C. L. Giles %A T. Maxwell %T Learning, invariance, and generalization in high order neural networks %J Applied Optics %V 26 %N 23 %D 1 |DEC| 1987 %P 4972-4978 %A C. L. Giles %A R. D. Griffin %A T. Maxwell %T Encoding Geometric Invariances in Higher Order Neural Networks %J Neural Information Processing Systems %I AIP Conf. Proc. %E D. Z. Anderson %D 1988 %P 301-309 %X good paper, derives constraints for 3rd order network with translational, rotational, and positional invariance; Gives constraints for simultaneous translation and scale invariance; Used 8x8 input size and and 1 output node to distinguish the letters C and T, Claims good results for scale factors upto 4x; Higher Order connections are equivalent to Sigma-Pi units; Uses back propagation to train scale invariant weights in multilayers; invariance is independent of training rules and sigmoid function; %T Distortion Invariant Character Recognition by a Multi-Layer Perceptron and Back Propogation Learning %A A. Khotanzad %A J. H. Lu %J |ICNN88| %V I %P 625-632 %A G. Krishnan %A D. Walters %T Psychologically plausible features for shape recognition in a neural network %J |ICNN88| %V II %P 127-134 %K image invariance %A T. Maxwell %A C. L. Giles %A Y. C. Lee %A H. H. Chen %T Transformation Invariance Using Higher Order Correlations in Neural Net Architectures %J Proc. IEEE Intl. Conf. on Systems, Man, and Cybernetics %C Atlanta, GA %D 14-17 |OCT| 1986 %P 627-632 %K HOLU, group theory, tolerance, temporal invariance %X Good Paper for introduction to positional invariance using HOLU (Higher order learning units); use group invariant approach; invariant autoassociative memory; good simple example of sign-change invariance for XOR problem; transformation invariance is imposed by averaging the input of each unit over transformation group using group theory; The higher the order the more storage capacity. position invariance - we have one output neuron that responds the same invariant of the position; position tolerance - we have an output neuron for each possible location that responds to the input; Discusses storage limitations for higher order nets; %A T. Pawlicki %T Recognizing Image Invariants in a Neural Network Architecture %J |ICNN88| %V II %P 135-142 %A W. L. Reber %A J. Lyman %T An Artificial Neural System Design for Rotation and Scale Invariant Pattern Recognition %V IV %P 277-283 %J |ICNN87| %K image processing %X Good Paper, has good intro with references for various transforms to achieve invariance, strictly NN aproach with 7 slabs: 1 input, Polar Transform, Log/Polar Transform, 1D FT rows, 1D FT columns, classifier, output display; total of 5419 PE and 238,266 IC %T Performance of Backpropagation for Rotation Invariant Pattern Recognition %A Hedong Yang %A Clark C. Guest %J |ICNN87| %V IV %P 365-370 %\def\Zzz{Special Issue on Neural Networks} %J Applied Optics %A \Zzz %V 26 %N 23 %D 1 |DEC| 1987 %K special Issue %X contains papers on Adaptive, associative, and self-orgainizing functions in neural computing; ART 2: self-organization of stable category recognition codes for analog input patterns; Probing cognitive processes through the structure of event-related potentials during learning: an experimental analysis; Adaptive bidirectional associative memories; Dynamic control of an artifical neural system: the property inheritance network; Learning, invariance, and generalization in high-order neural networks; Counterpropagation networks; Neural network model for selective attention in visual pattern recognition and associative recall; Retinal model for adaptive contrast sensitivity and resolution; Associative learning of scene parameters from images; Optimization of the computational load of a hypercube supercomputer onboard a mobile robot; Neural dynamics of attentionaly modulated Pavlovian conditioning: blocking interstimulus interval, and secondary reinforcement; Dynamic optical interconnects: volume holograms as optical two-port operators; Optical implementations of associative networks with versatile adaptive learning capabilities; Designs and devices for optical bidirectional associative memories; Multilayer optical learning networks; Electronic neural network chips; Building a hierarchy with neural networks: an example - image vector quantization; Electronic hardware implementations of neural networks; Optoelectronic analogs of self-programming neural nets: architecture and methodologies for implementing fast stochastic learning by simulated annealing; Optoelectronic resonator neural networks %T Invariant Pattern Recognition and Recall by an Attentive Self Organizing Art Architecture in a Nonstationary World %A Gail A. Carpenter %A Stephen Grossberg %J |ICNN87| %V II %P 737-745 %T Coordinate Transformation From an Image Plane Directly to an Invariant Feature Space %A Richard A. Messner %A Harold Szu %J IEEE Computer Society Conference on Computer Vision and Pattern Recognition %C Washington, D.C. %P 522-530 %D 1983 %K image %T Invariant Pattern Recognition Using Multiple Filter Image Representations %A Christoph Zetzsche %A Terry Caelli %J Computer Vision, Graphics, and Image Processing %V 45 %P 251-262 %D 1989 %K image %T An Image Processing Architecture for Real Time Generation of Scale and Rotation Invariant Patterns %A Richard A. Messner %A Harold H. Szu %J Computer Vision, Graphics, and Image Processing %V 31 %P 50-66 %D 1985 %K image %T Rotation Invariant Pattern Recognition Using Zernike Moments %A Alireza Khotanzad %A Yaw Hua Hong %J 9th Int. Conf. on Pattern Recognition %I IEEE %V I %P 326-328 %K image processing, othogonal functions %D 1988 ---------------------------------------------------------------------------- . . .__. The opinions expressed herein are soley |\./| !__! Michael Plonski those of the author and do not represent | | | "plonski@aero.org" those of The Aerospace Corporation. _______________________________________________________________________________ ------------------------------ Subject: Re: Translation-Invariant Pattern Recognizer NN's From: debbiec@hprmokg.HP.COM (Debbie Clingingsmith) Organization: HP Networked Computer Mfg. Operation Date: 20 Feb 90 21:40:27 +0000 Dr. Widrow gave a presentation at the 1989 WESCON in S.F. He presented his original 1960 paper where ADALINE and Madaline were described (first time in WESCON history where a paper was presented twice). Anyway, the paper was reprinted in the conference proceedings. Hope this helps. Debbie Clingingsmith Telnet: 1-785-5226 NCMO - North Outside: (916) 785-5226 Hewlett-Packard Internet: debbiec%hprmokg@hplabs.hp.com 8000 Foothills Blvd. Roseville, Ca. 95678 ------------------------------ Subject: Edelman From: Stephen Smoliar <smoliar@vaxa.isi.edu> Date: Mon, 26 Feb 90 13:06:17 -0800 Patrick Thomas' suggestion to look at Gerald Edelman's new book, THE REMEMBERED PRESENT, may very well be a good one. I am about half way through the book and find it fascinating reading. [Unfortunately, Peter, I have accepted a commitment to review it for ARTIFICIAL INTELLIGENCE; so I'm not sure if I shall be able to provide you with a review as well.] However, my personal feeling is that anyone who tries to use this as a first exposure to Edelman's work will find it rather rough going. The book is the third volume of a trilogy, having been preceded by NEURAL DARWINISM and TOPOBIOLOGY. I think that anyone who is interested in "real" brains would do well to start with NEURAL DARWINISM. I shall now allow myself a shameless plug and recommend the review of wrote of this book for ARTIFICIAL INTELLIGENCE which appeared in Volume 39, Number 1 (May 1989). I like to think that I have written an essay which will "prime" any reader interested in artificial intelligence for an approach to Edelman's way of thinking. Also, I organized a panel at the 1989 IJCAI for members of the AI community to respond to Edelman (who had given an invited talk). The position papers for this panel were published in the conference proceedings (pages 1669-1671). As far as the middle volume of the trilogy is concerned, I must confess that I skipped over TOPOBIOLOGY, not for lack of interest but for lack of time. [[ Editor's Note: Thanks, Stephen. Perhaps someone else could give a short synopsis, without any copyright infringement. -PM ]] ------------------------------ Subject: 3rd Intn'l Symposium on AI (Mexico) From: "Centro de Inteligencia Artificial(ITESM)" <ISAI@TECMTYVM.MTY.ITESM.MX> Organization: Instituto Tecnologico y de Estudios Superiores de Monterrey Date: Mon, 19 Feb 90 17:18:12 -0400 To whom it may concern: We are sending you here the information concerning the "Third International Symposium on Artificial Intelligence". Please display it in your bulletin board. THANK YOU VERY MUCH IN ADVANCE. Sincerely, The Symposium Publicity Committee. THIRD INTERNATIONAL SYMPOSIUM ON ARTIFICIAL INTELLIGENCE: APPLICATIONS OF ENGINEERING DESIGN & MANUFACTURING IN INDUSTRIALIZED AND DEVELOPING COUNTRIES OCTOBER 22-26, 1990 ITESM, MEXICO The Third International Symposium on Artificial Intelligence will be held in Monterrey, N.L. Mexico on October 22-26, 1990. The Symposium is sponsored by the ITESM (Instituto Tecnologico y de Estudios Superiores De Monterrey) in cooperation with the International Joint Conferences on Artificial Intelligence Inc., the American Association for Artificial Intelligence, the Sociedad Mexicana de Inteligencia Artificial and IBM of Mexico. OBJETIVES: * Promote the development and use of AI technology in the solution of real world problems in the application areas of engineering design & manufacturing. * Analize the state-of-the-art of AI technology in different countries and its potencial use and transfer from universities and research institutions to industries for the solution of problems. * Evaluate efforts made in the use of AI technology in all coutries. Provide general guidelines for the use and spread of AI technology in all countries, and assess them in the next annual ISAI meeting. ISAI FORMAT: ISAI consists of a tutorial and a conference. The tutorial is a set of seminars on relevant AI topics given in two days (Oct. 22-23). The conference is a set of lectures given during three days (Oct. 24-26). It consists of invited papers and selected papers from the "Call for Papers" invitation. Areas of application include: * Production planning * resource management, * quality management, * automated assembly, * machine loads, * inventory control, * computer aided product design, *computer aided product manufacturing * human resources management, * forecasting, *client/customer support, * process control and ES, * automatic process inspection, * use of industrial robots, * market and competition analysis, * strategic planning of manufacturing, * technology management and social impact of AI technology in industrial environments. AI techniques include: * Knowledge acquisition and representation, * natural language processing, * robotics , * speech recognition, * computer vision, * neural networks and genetic algorithms, * parallel architectures, * automated learning, * automated reasoning, * search and problem solving, * knowledge engineering tools and methodologies, * uncertainty management and AI programming languages. IMPORTANT: In order to encourage an atmosphere of friendship and exchange among participants, some social events are being organized. For your convinience we have arranged a free shuttle bus service between the hotel zone and the ITESM during the three day conference. FEES: (VALID BEFORE AUGUST 31) Tutorial.- Professionals $ 250 USD Students $ 125 USD Conference.- Professionals $ 180 USD Students $ 90 USD Simultaneous Translation FORMAL DINNER-INCLUDES DINNER, OPEN BAR, MUSIC-(OCT 26) TUTORIAL FEE INCLUDES: Tutorial material. Welcoming cocktail party (Oct.22) CONFERENCE FEE INCLUDES: Proceedings. WELCOMING COCKTAIL PARTY (OCT.24) COCKTAIL PARTY. (OCT.25) REGISTRATION PROCEDURE: Send personal check payable to "I.T.E.S.M." to: "Centro de Inteligencia Artificial, Attention: Leticia Rodriguez, Sucursal de Correos "J", C.P. 64849, Monterrey, N.L. Mexico" INFORMATION: CENTRO DE INTELIGENCIA ARTIFICIAL, ITESM. SUC. DE CORREOS "J", C.P. 64849 MONTERREY, N.L. MEXICO. TEL (52-83) 58-20-00 EXT.5132 or 5143. TELEFAX (52-83) 58-07-71, (52-83) 58-89-31, NET ADDRESS: ISAI AT TECMTYVM.BITNET ISAI AT TECMTYVM.MTY.ITESM.MX GENERAL CHAIR: FRANCISCO J. CANTU-ORTIZ, ITESM, MEXICO ADVISORY BOARD: SAUL AMAREL, RUTGERS UNIVERSITY, USA WOODROW BLEDSOE, U. OF TEXAS AT AUSTIN, USA RANDOLPH GOEBEL, UNIVERSITY OF ALBERTA, CANADA ADOLFO GUZMAN, INTERNATIONAL SOFTWARE SYSTEMS, USA RAJ REDDY, CARNEGIE MELLON UNIVERSITY, USA DONALD WALKER, BELLCORE, USA PROGRAM CHAIRS: JOSE LUIS AGUIRRE, ITESM, MEXICO ROCIO GUILLEN, ITESM, MEXICO HUGO TERASHIMA, ITESM, MEXICO PROGRAM COMMITTEE: RENE BANARES, UNAM, MEXICO RAMON BRENA, ITESM, MEXICO OFELIA CERVANTES, U.AMERICAS, MEXICO FRANCISCO CERVANTES, UNAM, MEXICO JOHN DEBENHAM, U.TECHNOLOGY, SIDNEY AUSTRALIA GERHARD FISCHER, UNIVERSITY OF COLORADO, USA MARK FOX, CARNEGIE MELLON UNIVERSITY, USA DAVID FULLER, U.CATOLICA DE CHILE EUGENIO GARCIA, ITESM, MEXICO JOSE LUIS GONZALEZ, ITESM, MEXICO IGNACIO GROSSMAN, CARNEGIE MELLON UNIVERSITY STEFFEN HOELDOBLER, TECHNISCHE HOCHSCHULE DARMSTADT, FRG HEINZ ULRICH HOPPE, GMD-IPSI, GERMANY S.LAKSHMIVARAHAN, UNIVERSITY OF OKLAHOMA, USA FERNANDO LARA, ITESO, MEXICO CHRISTIAN LEMAITRE, UNAM, MEXICO LORI LEVIN, CARNEGIE MELLON UNIVERSITY JAY LIEBOWITZ, GEORGE WASHINGTON UNIVERSITY, USA RAMON LOPEZ-MANTARAS, BLANES, SPAIN GUILLERMO MORALES, I.POLITECNICO NACIONAL, MEXICO JUDEA PEARL, UCLA, USA JORGE PHILLIPS, STANFORD UNIVERSITY, USA S.RAMANI, NATIONAL CENTRE FOR SOFTWARE TECH, INDIA FRANCOIS RECHENMANN, INRIA, GRENOBLE, FRANCE DANIEL REHAK, CARNEGIE MELLON UNIVERSITY, USA RICHARD STERN, CARNEGIE MELLON UNIVERSITY, USA SAROSH TALUKDAR, CARNEGIE MELLON UNIVERSITY, USA MASARU TOMITA, CARNEGIE MELLON UNIVERSITY, USA JOHN TRUJILLO, EAFIT, COLOMBIA MANUEL VALENZUELA, ITESM, MEXICO CARLOS ZOZAYA, CONDUMEX, MEXICO Publicity and Tutorial Chair: MORAIMA CAMPBELL, ITESM, MEXICO Local Arrangement Chair: LETICIA RODRIGUEZ, ITESM, MEXICO Exhibits Chair: LUCILA PENA, ITESM, MEXICO ------------------------------ Subject: Neocognitron (revised) Code From: pasteur!helios.ee.lbl.gov!ncis.tis.llnl.gov!blackbird!efrethei (Erik J. Fretheim) Organization: Air Force Institute of Technology; WPAFB, OH Date: 16 Feb 90 14:42:45 +0000 I am interested in doing some experiments using a neocognitron in the revised (1988 or so) version. Does anyone have a PD version up and running which they would be willing to share, or pointers which they would be willing to provide. I hate recoding the wheel (net). Thank you. Erik Fretheim efrethei@afit.af.mil AFIT/ENA Box 4151 WPAFB, OH 45431 USA (513)255-5276 ------------------------------ Subject: NEW JOURNAL: MINDS & MACHINES From: rapaport@cs.Buffalo.EDU Organization: SUNY/Buffalo Computer Science Date: 02 Feb 90 18:42:17 +0000 ======================================================================== CALL FOR PAPERS---CALL FOR PAPERS---CALL FOR PAPERS---CALL FOR PAPERS ======================================================================== Kluwer Academic Publishers announces MINDS AND MACHINES Journal for Artificial Intelligence, Philosophy, and Cognitive Science ISSN 0924-6495 (previously announced as `Philosophy and Artificial Intelligence') EDITORIAL FOCUS: Machines and Mentality Knowledge and its Representation Epistemic Aspects of Computer Programming Connectionist Conceptions Artificial Intelligence and Epistemology Computer Methodology Computational Approaches to Philosophical Issues Philosophy of Computer Science Simulation and Modeling Ethical Aspects of Artificial Intelligence EDITOR: James H. Fetzer, University of Minnesota, Duluth, MN, USA BOOK REVIEW EDITOR: William J. Rapaport, SUNY Buffalo, Buffalo, NY, USA EDITORIAL BOARD (as of January 1990) Jon Barwise Stanford University, USA Andy Clark University of Sussex, UK Robert Cummins University of Arizona, USA Fred Dretske University of Wisconsin at Madison, USA Jerry Fodor Rutgers University, USA Clark Glymour Carnegie-Mellon University, USA John Haugeland University of Pittsburgh, USA Jaakko Hintikka Florida State University, USA David Israel SRI International, USA Frank Keil Cornell University, USA Henry Kyburg University of Rochester, USA John McCarthy Stanford University, USA Donald Nute University of Georgia, USA Zenon Pylyshyn University of Western Ontario, Canada Barry Richards Imperial College, London, UK Roger C. Schank Northwestern University, USA John Searle University of California at Berkeley, USA Stephen Stich Rutgers University, USA Terry Winograd Stanford University, USA MINDS AND MACHINES affords an international forum for discussion and debate of important and controversial issues concerning significant developments within its areas of editorial focus. Well-reasoned contributions from diverse theoretical perspectives are welcome, and every effort will be made to insure their prompt publication. Among the features that are intended to make this journal distinctive within the field are these: o Strong stands on controversial issues are specifically encouraged; o Important articles exceeding normal journal length may appear; o Special issues devoted to specific topics will be a regular feature; o Review essays discussing current problem situations will appear; o Critical responses to previously published pieces are also invited. This journal is intended to foster a tradition of criticism within the AI and philosophical communities on problems and issues of common concern. Its scope explicitly encompasses philosophical aspects of computer science. All submissions will be subject to review. Publication will begin with a single volume of four issues per year. The first issue will appear in January 1991. Contributors should send 4 copies of their manuscript to: James H. Fetzer, Editor MINDS AND MACHINES Department of Philosophy University of Minnesota Duluth, MN 55812 USA phil@ub.d.umn.edu AI_and_PHIL@ub.d.umn.edu Correspondence concerning books for review should be sent to: William J. Rapaport, Book Review Editor MINDS AND MACHINES Center for Cognitive Science Department of Computer Science SUNY Buffalo Buffalo, NY 14260 USA rapaport@cs.buffalo.edu rapaport@sunybcs.bitnet Subscription information and sample copies will be available from: Kluwer Academic Publishers Group P.O. Box 322 3300 AH Dordrecht The Netherlands or Kluwer Academic Publishers 101 Philip Drive Norwell, MA 02061 USA ======================================================================== CALL FOR PAPERS---CALL FOR PAPERS---CALL FOR PAPERS---CALL FOR PAPERS ======================================================================== ------------------------------ End of Neuron Digest [Volume 6 Issue 17] ****************************************