neuron-request@HPLMS2.HPL.HP.COM ("Neuron-Digest Moderator Peter Marvit") (10/31/90)
Neuron Digest Tuesday, 30 Oct 1990 Volume 6 : Issue 63 Today's Topics: Music Info request: Neural Net Application Tools Re: Neuron Digest V6 #62 A Short Course in Neural Networks and Learning Theory Neural Network Simulation Service Neuron Digest V6 #62 PRE-PRINT availability New Book on Neural Networks (PC Tools) info on a workshop 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: MUSIC From: Niall Griffith <ngr@cs.exeter.ac.uk> Date: Thu, 18 Oct 90 17:26:59 +0100 I am working at the Connection Science lab at Exeter, and I am writing a review of connectionist research on music. It would be really useful if you could send me as many references as you have on the subject. I will of course make these publicly available. Niall Griffith Centre for Connection Science JANET: ngr@uk.ac.exeter.cs Dept. Computer Science University of Exeter UUCP: ngr@expya.uucp Exeter EX4 4PT DEVON BITNET: ngr@cs.exeter.ac.uk@UKACRL UK ------------------------------ Subject: Info request: Neural Net Application Tools From: lambert@cod.nosc.mil (David R. Lambert) Date: Fri, 26 Oct 90 09:54:59 -0700 I would like recommendations for neural net application tools which are reliable and easy for students to learn and use. I need to detect and recognize patterns in data consisting of a dozen or so variables, each with a fairly small number of discrete values. My platforms are IBM-PC, VAX, and Macintosh. I am currently beginning to look at Brainmaker, MacBrain, and NeuralWare Explorer. David R. Lambert, PhD Email: lambert@nosc.mil ------------------------------ Subject: Re: Neuron Digest V6 #62 From: howard@aic.hrl.hac.com Date: Fri, 26 Oct 90 18:42:09 -0700 > I'm interested in any references dealing with the use of > neural nets for the real time control or simulation of movement, > especially locomotion. Ultimately, the group I'm working > with wants to understand gait (locomotion ) disorders in > humans - to me this means simulation. Look for Nigel Goddard's work (student of Jerry Feldman). Goddard is finishing his PhD at U. Rochester on a NN for recognizing gait. He uses "moving light displays" created by placing a light on each joint and filming movements. ----------Mike Howard, Hughes Research Labs ------------------------------ Subject: A Short Course in Neural Networks and Learning Theory From: john@cs.rhbnc.ac.uk Date: Sun, 28 Oct 90 12:45:21 +0000 ------------------------------------- A SHORT COURSE IN NEURAL NETWORKS AND LEARNING THEORY 7th and 8th January, 1991 ------------------------------------- Dr John Shawe-Taylor, Department of Computer Science, Royal Holloway and Bedford New College, University of London, Egham, Surrey TW20 0EX UK The two day course will give an introduction to Neural Networks and Learning Theory. It will be an opportunity to hear of recent results which place the subject of Connectionism on a firm theoretical foundation, by linking it with Computational Learning Theory. John Shawe-Taylor has himself contributed to recent developments in applying results of Computational Learning Theory to Neural Networks. A key feature of the course will be its hands-on practical flavour. Both days will include sessions where participants will have an opportunity to test out ideas in practical working examples. This will highlight the real problems in network design and training. However, by applying the theoretical results of Computational Learning Theory many difficult problems will become better understood, and in some cases tractable solutions will suggest themselves. There follows a summary of the two days. Day 1: Connectionism and Neural Networks ---------------------------------------- The day starts with an overview of connectionism stressing the main strengths and weaknesses of the approach. Particular emphasis will be given to areas where the techniques will find industrial applications in the near future. At the same time the areas where major problems remain to be solved will be outlined and an indication of current trends in research will be given, as well as implementation techniques. Particular emphasis will be placed on Feedforward Neural Networks. Such networks will be discussed in more detail. This will be followed by an opportunity to gain first-hand experience of the problems involved in designing and training networks. In a concluding session a summary will put the practical experiences in perspective with particular reference to current research. Day 2: Learning Theory for Feedforward Networks ----------------------------------------------- During the second day the focus will be on Computational Learning Theory and its application to the problems of training feedforward Neural Networks. The day begins with an overview of the field of Computational Learning Theory. This is followed by a discussion of contributions that the theory has made in understanding connectionist architectures. The results range from "negative", such as the fact that certain training problems will be difficult or infeasible, to "positive", such as the existence of methods for estimating the size of training sample needed to give good generalisation with high confidence. The practical sessions of the day will involve applying these insights to the problems of designing and training feedforward Neural Networks. It is possible to register for just one of the two days. For more details and registration information, please write to: Dr Penelope Smith, Industrial Liaison Officer, RHBNC, Egham, Surrey TW20 0EX or email your postal address to: john@cs.rhbnc.ac.uk ------------------------------ Subject: Neural Network Simulation Service From: David Kanecki <kanecki@vacs.uwp.wisc.edu> Date: Sun, 28 Oct 90 17:49:05 -0600 CONNECT/LINK ------------ NEURAL NETWORK/ HYPOTHESIS CORRELATION SERVICES To introduce people to neural networks and hypothesis correlation services, a limited time offer is available where one can have at no charge neural network/hypothesis correlation analysis of information. For example, given a series of results one can use the service to aid in prediction based upon new or old circumstances. To use the service, send the information via e-mail or regular mail with a Stamped Self Addressed Enveloped included and use the information format sheet below. Based upon that no fee will be charged, 5 request per organization or individual will be the limit accepted. My e-mail address is: kanecki@vacs.uwp.wisc.edu. And, my regular mail address is: David H. Kanecki, Bio. Sci., A.C.S. P.O. Box 93 Kenosha, WI 53141 United States, USA or (414)-654-8710 After 7 PM CST 1. DATA FROM EXPERIMENT OR NOTEBOOK: Example: A system has two states of events and one state of action. Based upon test the following information was obtained: Observation 1 Observation 2 Action 1 ------------- ------------- -------------- No Stimulus No Stimulus No Action occurred No Stimulus Stimulus Action 1 occurred Stimulus No Stimulus Action 1 occurred Stimulus Stimulus No Action occurred >From this data a request sheet was prepared as in the example below: --- Or verbal description so I can do coding, if need be --- 2. DATA FROM NOTEBOOK TRANSCRIBED FROM EXPERIMENT TO REQUEST FORM: REQUEST for Neural Network/Hypothesis Correlation Service ---------------------------------------------------------- Name: __________________________________________________ Organization:___________________________________________ Address: _______________________________________________ E-Mail Address:_________________________________________ Phone No: ______________________________________________ FAX No: ________________________________________________ Data Coding Sheet: Setup: Number of Stimulus States: 2 ( Observation 1 and 2) Number of Action States: 1 ( Action 1) Stimulus Active Action Active --------------- --------------- Case 1: None None Case 2: Stimulus 2 Action 1 Case 3: Stimulus 1 Action 1 Case 4: Stimulus 1 and 2 None Hypothesis Query: (What action will occur if a stimulus is active) Stimulus --------- 1). Stimulus 1 2). Stimulus 2 3). Stimulus 1 & 2 --------------------------------- Based upon the data sent in and queries asked I will use the neural network/ hypothesis correlation program to generate a response to hypothesis queries based upon the data specified in the setup section of the form. 3. RESULTS OF ANALYSIS Results of Analysis Form ------------------------- Percent of Information learned from sample data : 100% Special Coding used by Operator to increase retention: YES Active Stimulus Given Action Predicted --------------------- ----------------- 1. None No action, Action 1 inactive 2. Stimulus 2 Action 1 active 3. Stimulus 1 Action 1 active 4. Stimulus 1 and 2 No action, Action 1 inactive 4. Allow 6 to 14 days for Return/Reply May this service help you and your associates understand simulation better for enrichment and creativity. Your comments and feedback are welcome. ------------------------------ Subject: Neuron Digest V6 #62 From: JJ Merelo <jmerelo@ugr.es> Date: 29 Oct 90 17:07:00 +0200 Here are some references for Kohonen network [Koh82] T. Kohonen, Self organized formation of topologically correct feature maps, Biological Cybernetics 43: 59- 69. [KOH82] KOHONEN,T.: "Clustering, taxonomy, and topological maps of patterns". In: Proceedings of the 6th Int. Conf. on Pattern Recognition. IEEE Computer Society Press. 1982. [KOH84] KOHONEN,T; MKISARA,K.; SARAMKI,T.: "Phonotopic maps -insightfull representation of phonological features for speech recognition"; Proceedings of IEEE 6th Int. Conf. on Pattern Recognition. Montreal (Canada). pp.182-185. 1984 [KOH88a] KOHONEN, T.: "Self-Organization and Associative Memory"; Springer-Verlag; 1st Edt. 1984; 2nd Edt. 1988 [KOH88b] KOHONEN T. "The 'Neural' Phonetic Typewriter" IEEE Computer; Vol.21, No3, pp.11-22; 1988 You can get them from Kohonen himself, writing to him. I have not found many other references. And speech recognition people seem to think not very well about Kohonen's self-organizing map JJ Merelo Granada University ( Spain ) Electronics and Computer Tech. Dept. ------------------------------ Subject: PRE-PRINT availability From: P.Refenes@cs.ucl.ac.uk Date: Mon, 29 Oct 90 17:42:25 +0000 The following pre-print (SPIE-90, Boston, Nov. 5-9 1990) is available. (write or e-mail to A. N. Refenes at UCL) AN INTEGRATED NEURAL NETWORK SYSTEM for HISTOLOGICAL IMAGE UNDERSTANDING A. N. REFENES, N. JAIN & M. M. ALSULAIMAN Department of Computer Science, University College London, Gower Street, WC1, 6BT, London, UK. ABSTRACT This paper describes a neural network system whose architecture was designed so that it enables the integration of heterogeneous sub-networks for performing specialised tasks. Two types of networks are integrated: a) a low-level feature extraction network for sub-symbolic computation, and b) a high-level network for decision support. The paper describes a non trivial application from histopathology, and its implementation using the Integrated Neural Network System. We show that with careful network design, the backpropagation learning procedure is an effective way of training neural networks for histological image understanding. We evaluate the use of symmetric and asymmetric squashing functions in the learning procedure and show that symmetric functions yield faster convergence and 100% generalisation performance. ------------------------------ Subject: New Book on Neural Networks (PC Tools) From: Russ Eberhart <RCE1%APLVM.BITNET@CORNELLC.cit.cornell.edu> Date: Tue, 30 Oct 90 08:19:51 -0500 Announcing a new book on neural networks: NEURAL NETWORK PC TOOLS: A PRACTICAL GUIDE Edited by Russell Eberhart and Roy Dobbins The Johns Hopkins University Applied Physics Laboratory Published by Academic Press: ISBN 0-12-228640-5 TABLE OF CONTENTS Foreword (by Bernard Widrow) Introduction (by Eberhart & Dobbins) a. Myths versus realities b. Purpose of book c. Organization of book d. Main neural network use categories Chapter 1 - Background and History (by Eberhart & Dobbins) a. Introduction 1. What is a neural network? 2. What is a neural network tool? b. Biological basis for neural network tools 1. Introduction 2. Neurons 3. Differences between biological structures and NNT's 4. Where did neural networks get their name? c. Neural network development history 1. Introduction 2. The Age of Camelot 3. The Dark Age 4. The Renaissance 5. The Neoconnectionist Age Chapter 2 - Implementations (by Eberhart & Dobbins) a. Introduction b. Supervised training: The back-propagation model 1. Introduction 2. Topology and notation 3. Network input 4. Feedforward calculations 5. Training by error back-propagation 6. Running the back-propagation NNT c. Unsupervised training: Self-organization and associative memory 1. Introduction 2. Topology and notation 3. Network initialization and input 4. Training calculations 5. Running the self-organization NNT Chapter 3 - Systems (by Eberhart & Dobbins) a. Specification of the task b. How to optimize the use of the neural network tool c. How to choose the proper neural network tool d. The importance of preprocessing 1. Use NN's wisely...don't try to do everything with them 2. Design for overall optimal system performance e. Relationship to other areas including expert systems f. Problem categories appropriate for neural networks 1. How to out-expert expert systems 2. Don't invent a Cadillac when a VW will do 3. Pattern recognition 4. Biopotential waveform analysis and classification Chapter 4 - Software Tools (by Dobbins & Eberhart) a. Introduction b. Implementing neural networks on the PC 1. Using C and assembly language 2. Back-propagation networks 3. Vector and matrix operations c. Running neural networks 1. Getting data into and out of the network 2. Setting attributes 3. What's it doing? d. Implementation issues Chapter 5 - Development Environments (by Dobbins & Eberhart) a. Introduction b. What is a neural network development environment? 1. Desirable characteristics of development environments 2. Why a development environment? 3. A brief survey of neural network development systems c. Introduction to network modeling languages d. Specifying neural network models 1. Specifying network architecture 2. Activation functions 3. Learning rules 4. Specifying the environment 5. Update rules 6. Neural network paradigms e. CASENET: A neural network development environment Chapter 6 - Hardware Implementations (by D. Gilbert Lee, Jr.) a. When do you really need hardware assistance? b. What's the deal about accelerator boards? c. Transputers: when transputing is a cost-effective approach d. What's possible to implement on computers smaller than a Cray e. Mini-Case Study: Ship Pattern Recognition Chapter 7 - Performance Metrics (by Eberhart, Dobbins, & Hutton) a. Introduction b. Percent correct c. Average sum-squared error d. Normalized error e. Receiver operating characteristic curves f. Recall and precision g. Sensitivity, specificity, positive predictive value and false alarm rate h. Chi-square test Chapter 8 - Network Analysis (by Vincent Sigillito & Russ Eberhart) a. Introduction b. Network analysis 1. Introduction 2. The "divide by three" Problem 3. Other considerations 4. The "square within a square" problem 5. Distributions of hidden neurode activity levels 6. Analyzing weights in trained networks Chapter 9 - Expert Networks (by Maureen Caudill) a. Introduction b. Rule-based expert systems c. Expert networks 1. Fuzzy logic 2. Fuzzy cognitive maps 3. An expert bond-rating network 4. A hierarchical expert network 5. Knowledge in an expert network d. Expert network characteristics e. Hybrid expert networks 1. Explanation by confabulation 2. Rule extraction 3. True hybrid expert Chapter 10 - Case Study I - EEG Waveform Classification (by Eberhart and Dobbins) a. System specifications b. Background c. Data preprocessing and categorization d. Test results Chapter 11 - Case Study II - Radar Signal Processing (by Vincent Sigillito and Larrie Hutton) a. The radar system b. Methods c. Implementation d. Conclusion Chapter 12 - Case Study III - Technology in Search of a Buck (by Tom Zaremba) a. Introduction b. Markets to watch and markets to trade c. Futures market forecasts d. Statistical futures market data e. Sources and value of character-of-market data f. Model description g. Are neural nets suited to implementing technical analysis models? h. What was tried with the multilayer perceptron model? i. How and why was the multilayer perceptron implemented in EXCEL? f. What was learned, what remains to be done and has any money been made? Chapter 13 - Case Study IV - Optical Character Recognition (by Gary Entsminger) a. Summary of the problem b. System configuration c. Scanner interfacing d. Objects in Pascal e. Notes and conclusions Chapter 14 - Case Study V - Making Music (by Eberhart & Dobbins) a. Introduction b. Representing music for neural network tools c. Network configurations d. Stochasticity, variability and surprise d. Playing your music with MIDI e. Now what? Glossary References Appendix A - Batchnet BP NNT code, with pattern, weight, run and demo files Appendix B - Self-Organizing NNT code, with pattern run and demo files Appendix C - Turbo Pascal code for optical character recognition shell Appendix D - Source code for music composition files Appendix E - Additional NNT resources a. Organizations/societies b. Conferences/symposia c. Journals/magazines/newsletters d. Bulletin boards e. Computer data bases Appendix F - Matrix multiplication code for transputers Index ------------------------------ Subject: info on a workshop From: salam@frith.egr.msu.edu Date: Wed, 24 Oct 90 12:17:30 -0400 This is a submission regarding information on a workshop that is organized and managed by the LifeLong Eduation Program at MSU. I hope that it would be of interest to some people. A Tutorial Workshop on Neural Nets Theory, Design, and (Electronic) Implementation November 12-13, 1990 Michigan State University East Lansing, Michigan 48824 A Tutorial Workshop on Neural Nets Theory Design, and (Electron- ic) Implementation November 12-13, 1990 Overview: It is recognized that no matter how fast conventional digital computers would become, it is unlikely that they could outdo the human performance in tasks such as pattern recognition or associative memory. It is only logical therefore that for such tasks the architecture of the microprocessor or the computer should emulate that of the brain. Many workers have proposed various architectures that model some aspects of the highly interconnected nerves system and the brain. These architectures, often referred to as n_e_u_r_a_l n_e_t_s, basically consist of a large number of simple processors (or neurons) which are highly interconnected and work asynchro- nously and in parallel. Some design procedures have also been proposed by many researchers; some procedures have been based on intuitive arguments and physical reasoning alone, however. Conse- quently, although the proposed neural devices have worked well in some case studies, they have been found to fail in numerous other cases as well. It is essential, therefore, to lay a foundation for the proper design of neural processing devices and develop effective learning algorithms. It is equally essential that the designed architectures and the learning algorithms lend them- selves naturally to the chosen medium of implementation. In essence, one has to accommodate the prevalent technologies and pursue a methodology that would ultimately balance the hypothesis of theoretical models with the constraints of the media of imple- mentation. A Tutorial Workshop on Neural Nets Theory Design, and (Electron- ic) Implementation June 4-5, 1990 Objectives: This workshop provides an in-depth introduction to re- cent formulation of neural networks spanning modeling, theory, applications, and (electronic) silicon implementation. It intro- duces the basic principles and mechanisms behind the present designs of neural nets; it identifies the advantages and the lim- itations of the existing design methodologies for specific appli- cations. The course presents novel learning schemes and explains what makes them work and (if and) when they might fail. >From a practical view point, the course will also focus on implementa- tions utilizing CMOS VLSI technologies. Recent design implementa- tions on VLSI chips, resulting from the research activities of the instructors, will also be described. Who Should Attend: This workshop is designed for those who wish to learn about the recent development in neural nets, their current use, their method of implementations, and their potential impact on science and technology. Prerequisite: At least a Bachelor's degree in engineering, phy- sics, mathematics, science, or equivalent. Background in circuits and systems is helpful. Faculty: Anthony N. Michel: received the Ph.D. degree in electri- cal engineering from Marquette University and the D.Sc. degree in applied mathematics from the Technical University of Graz, Aus- tria. He has seven years of industrial experience. From 1968 to 1984 he was at Iowa State University. At present he is Frank M. Freimann Professor of Engineering and Dean of the College of Electrical Engineering, University of Notre Dame, Notre Dame, IN. He is author and coauthor of three texts and several other publi- cations. Dr. Michel received the 1978 Best Transactions Paper Award of the IEEE Control Systems Society (with R. D. Rasmussen), the 1984 Gullemin-Cauer Prize Paper Award for the IEEE Circuits and Systems Society (with R. K. Miller and B. H. Nam), an IEEE Centennial Medal. He is a former Associate Editor and a former Editor of the IEEE Transactions on Circuits and Systems and a former Associate Editor of the IEEE Transactions on Automatic Control. He was the Program Chairman of the 1985 IEEE Conference on Decision and Control. He has been the present general chairman of the 1990 International Symposium on Circuits and Systems and he is presently an associate editor of the IEEE Transactions on Neural Networks. Fathi M. A. Salam (program chairman): is an associate professor of electrical engineering at MSU. He received his B.S. and Ph.D. in electrical science from the University of California-Berkeley, and holds master's degree in both mathematics and electrical en- gineering. The author or coauther of more than 70 technical pa- pers, he was associate editor of the IEEE transactions on Cir- cuits and Systems (CAS) for nonlinear circuits and systems from 1985-87. He was cochair of the Engineering Foundation Conference on Qualitative Methods for Nonlinear Dynamics in June 1986. He is the co-editor of the book, Dynamical Systems Approaches to Non- linear Problems in Circuits and Systems, SIAM, January 1988. He presently an associate editor of both the IEEE Transactions on Neural Networks as well as the Journal of Circuits, Systems, and Computers. Hisresearch interests include nonlinear phenomena in circuits and systems, analysis and design of neural networks, adaptive systems, and robotics. Timothy Grotjohn: received his B.S. and M.S. degrees from the University of Minnesota in 1982 and 1984 respectively. He then continued his studies at Purdue University with an American Elec- tronics Association-Hewlett Packard Faculty Development Fellow- ship completing his Ph.D. degree in 1986. He joined MSU in the Department of Electrical Engineering in 1987. His research area is the simulation, modeling and characterization of semiconductor devices and processes. He has done consulting at AT&T Bell La- boratories and he has worked two summers at AT&T Bell Labora- tories. He has also been a Visiting Researcher at the Institute of Microelectronics in Stuttgart, West Germany. Summary Date: November 12-13, 1990 Days: Monday-Tuesday Registration: 8:30 a.m. - 9:00 a.m. Time: Monday, November 12 Session Time: 8:30 a.m. - 5:00 p.m. daily Place: The Kellogg Center for Continuing Education Fee: $395.00 per person Credit: 1.5 CEU A Tutorial Workshop on Neural Nets Theory Design, and (Electronic) Implementation June 4-5, 1990 Daily Schedule: Sessions meet from 8:30 a.m. to 5:00 p.m. each day. Monday, November 12 8:30 a.m. Session I, Room 104 Artificial Neural Nets - an introduction - -neural nets or processing networks of the brain: a different architecture for engineering technology - -biological neuronal networks and their architectures the synaptic weight and its models - -advantages of neural network processors: fault-tolerance, parallel processing, asynchronous processing - -mathematical models: the feedforward and feedback models, mul- tilayered models, the Hopfield model, the Grossberg model, the Hoppensteadt model, and newly introduced models. - -the discrete models vs. the analog models A mathematical formulation - gra- dient systems Lunch, Centennial Room 1:15 p.m. Session II , Room 104 Engineering Design & Applications - -Basic analysis-results - -Conditions for proper design - -Speed of convergence Four design schemes; discrete/ continuous - -Lower block triangular form - -Versatile function generator design - -A/D converters - -Resistor sorters 7:30 - 9:30 p.m. Optional tour of neural net research facility at Michigan State. Performance of PC-interfaced Artificial Neur- al Net chips will be demonstrated. Dr. Salam will be available for informal questions and review. Tuesday November 13 8:30 a.m. Session III, Room 104 Programming the network, learning, or how to store memories - -Learning: supervised and unsupervised - -The back propagation algorithm: when it works and why it might fail; continuous-time (analog) form; extensions and improvements - -The outer product (the Hebb) rule: theoretical justifications as well as limitations; the discrete sum vs. the integral form; practical experience with the rule; modifications - -New (1990) learning rule(s) that stores data for feedback neural nets Noon Lunch, Centennial Room 1:15 p.m. Session IV , Room 104 Implementation via simulation, electronics, and electro-optics - -Implementation media: software vs. hardware - -Hardware: electro-optics vs. electronics - -Electronics: digital vs. analog - -Advantages of analog silicon VLSI for (artificial) neural nets - -Basic elements of analog VLSI - -Designed/implemented analog MOS neural network VLSI chips 5:00 Adjournment General Information Program Fee: The program fee of $395.00 includes tuition, program materials, refreshment, and lunch. Group Discounts: Group discounts of 20% are available for three or more participants registered together in advance from the same company. Registration: Don't delay. Register today. In order to maintain reasonable class sizes, registrations are accepted on a first come, first served basis until an optimum number is achieved. The final deadline is May 31, 1990. Mail the registration form with payment today. Allow one week for your return confirmation. For immediate confirmation of registration, use your VISA/MasterCard by telephone (800) 447-3549 [in Michigan (800) 462-0846] or FAX (517) 353-3900. Changes and Cancellations: Michigan State University reserves the right to make changes in program speakers or presenters if un- foreseen circumstances so dictate. Michigan State University also reserves the right to cancel programs when enrollment criteria are not met, or where conditions beyond its control prevail. Every effort will be made to contact each enrollee when a program is cancelled. All program fees will be refunded when a program is cancelled by Michigan State University. Any additional costs in- cluded by the enrollee of cancelled programs are the responsibil- ity of the enrollee. Written cancellations by preregistrants postmarked ten or more days prior to the seminar will be fully refunded, except for a $25.00 processing fee. No refund will be allowed for withdrawal postmarked less than ten days prior to the seminar. If you fail to attend the program and do not notify En- gineering Lifelong Education, you are liable for the entire fee. Continuing Education Units (CEU): The Continuing Education Unit is defined as "Ten contact hours of participation in an organized continuing education experience under responsible sponsorship, capable direction, and qualified instruction." Michigan State University maintains a permanent record of all CEU's issued. In- dividuals may use transcripts as evidence of participation in continuing education programs. This program carries 1.5 CEU's. Housing: Housing is the responsibility of each participant. Hous- ing will be available at The Kellogg Center for Continuing Educa- tion. Room rates are in the range of $50 to $80. Room rates are subject to change. To make your reservations, please call 517- 355-5090 or complete the form attached. How to Reach Kellogg Center: Kellogg Center is located on campus, on Harrison Avenue. For motorists, exit from US-127, or I-496 at Trowbridge Road. When Trowbridge ends, turn left on Harrison to the center (east side of Harrison). Lansing's Capital City Air- port has limousine and taxi service to the center. The center is approximately one mile from the East Lansing train station served by Amtrak. For further information, contact Dr. Anthony Rigas, Director, Engineering Lifelong Education, A394 Engineering Build- ing, Michigan State University, East Lansing, MI 48824-1226. Telephone (800) 447-3549; or, in Michigan, (800) 462-0846. **************************************************************** **************************************************************** Please return this preregistration form if you plan to attend. Neural Nets Engineering Lifelong Education A-394 Engineering Building Michigan State University East Lansing, MI 48824-1226 Name______________________________________________ Title_______________________________________________ Institution/Company__________________________________ Address____________________________________________ City____________________State________ZIP____________ Daytime phone ( )______________________ Yes, I wish to receive CEU's. My S.S.# is_______-_______-_______ Conference Registration Fee: $395.00 pre-registeded %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~%Enclosed total :__________ Make check for fee and meals payable to Michigan State Universi- ty. To pay by VISA/MC card, please complete the following: Exp. date___________________ __Visa __ MasterCard Number__________________________________________ Signature________________________________________ (Engineering Lifelong Education cannot accept other credit cards.) November 12-13, 1990 Overnight Housing Reservation Arrival date___________ Departure date___________ Estimated arrival time___________________________ Single occupan- cy Shared occupancy (half twin). Person you wish to share with_____________________________ Regular room; if none avail- able, please book a room in another nearby facility at comparable rate. Regular room; if none available, please book a deluxe room. Deluxe room. Late Arrival Guarantee: __ Visa __ MasterCard __ AMEX Exp. date__________________________ charge card no.________________________________ Signature_____________________________________ Workshop on Neural Nets Engineering Lifelong Education A-394 Engineering Building Michigan State University East Lansing, MI 48824-1226 ------------------------------ End of Neuron Digest [Volume 6 Issue 63] ****************************************