Artificial neural nets and genetic algorithms proceedings of the international conference in norwich, u k , 1997

I I I I I I I I I I I I I I I I I I George D Smith Nigel C Steele Rudolf F Albrecht Artificial Neural Nets and Genetic Algorithms Proceedings of the International Conference in Norwich, U.K., 1997 Springer-Verlag Wien GmbH Dr George D Smith School of Information Systems University of East Anglia, Norwieh, U.K Dr Nigel C Steele Division of Mathematics School of Mathematical and Information Sciences Coventry University, Coventry, u.K Dr Rudolf F Albrecht Institut für Informatik Universität Innsbruck, Innsbruck, Austria This work is subject to copyright All rights are reserved, whether the whole or part of the material is concerned, specifieally those of translation, reprinting, re-use of illustrations, broadcasting, reproduction by photocopying machines or similar means, and storage in data banks © 1998 Springer-Verlag Wien Originally published by Springer-Verlag Wien 1998 Camera-ready copies provided by authors and editors Graphie design: Ecke Bonk Printed on acid-free and chlorine-free bleached paper SPIN 10635776 With 384 Figures ISBN 978-3-211-83087-1 ISBN 978-3-7091-6492-1 (eBook) DOI 10.1007/978-3-7091-6492-1 Preface This is the third in a series of conferences devoted primarily to the theory and applications of artificial neural networks and genetic algorithms The first such event was held in Innsbruck, Austria, in April 1993, the second in Ales, France, in April 1995 We are pleased to host the 1997 event in the mediaeval city of Norwich, England, and to carryon the fine tradition set by its predecessors of providing a relaxed and stimulating environment for both established and emerging researchers working in these and other, related fields This series of conferences is unique in recognising the relation between the two main themes of artificial neural networks and genetic algorithms, each having its origin in a natural process fundamental to life on earth, and each now well established as a paradigm fundamental to continuing technological development through the solution of complex, industrial, commercial and financial problems This is well illustrated in this volume by the numerous applications of both paradigms to new and challenging problems The third key theme of the series, therefore, is the integration of both technologies, either through the use of the genetic algorithm to construct the most effective network architecture for the problem in hand, or, more recently, the use of neural networks as approximate fitness functions for a genetic algorithm searching for good solutions in an 'incomplete' solution space, i.e one for which the fitness is not easily established for every possible solution instance Turning to the contributions, of particular interest is the number of contributions devoted to the development of 'modular' neural networks, where a divide and conquer approach is adopted and each module is trained to solve a part of the problem Contributions also abound in the field of robotics and, in particular, evolutionary robotics, in which the controllers are adapted through the use of some evolutionary process This latter field also provided a forum for contributions using other related technologies, such as fuzzy logic and reinforcement learning Furthermore, we note the relatively large number of contributions in telecommunications related research, confirming the rapid growth in this industry and the associated emergence of difficult optimisation problems The increasing complexity of problems in this and other areas has prompted researchers to harness the power of other heuristic techniques, such as simulated annealing and tabu search, either in their 'pure' form or as hybrids The contributions in this volume reflect this trend Finally, we are also pleased to continue to provide a forum for contributions in the burgeoning and exciting field of evolutionary hardware We would like to take this opportunity to express our gratitude to everyone who contributed in any way to the completion of this volume In particular, we thank the members of the Programme Committee for reviewing the submissions and making the final decisions on the acceptance of papers, Romek Szczesniak (University of East Anglia) for his unenvious task of preparing the LaTeX source file, Silvia Shilgerius (Springer-Verlag) for the final stages of the publication process and, not least, to all researchers for their submissions to ICANNGA97 We hope that you enjoy and are inspired by the papers contained in this volume George D Smith Norwich Nigel C Steele Coventry Rudolf F Albrecht Innsbruck Contents Advisory and Programme Committees xvi Robotics and Sensors Obstacle Identification by an Ultrasound Sensor Using Neural Networks D Diep, A Johannet, P Bonnefoy and F Harroy A Modular Reinforcement Learning Architecture for Mobile Robot Control R M Rylatt, C A Czarnecki and T W Routen Timing without Time - An Experiment in Evolutionary Robotics H H Lund Incremental Acquisition of Complex Behaviour by Structured Evolution S Perkins and G Hayes Evolving Neural Controllers for Robot Manipulators R Salama and R Owens Using Genetic Algorithms with Variable-length Individuals for Planning Two-Manipulators Motion J Riquelme, M Ridao, E F Camacho and M Toro 11 16 21 26 ANN Architectures Ensembles of Neural Networks for Digital Problems D Philpot and T Hendtlass A Modular Neural Network Architecture with Additional Generalization Abilities for Large Input Vectors A Schmidt and Z Bandar Principal Components Identify MLP Hidden Layer Size for Optimal Generalisation Performance M Girolami Bernoulli Mixture Model of Experts for Supervised Pattern Classification N Elhor, R Bertrand and D Hamad 31 35 40 44 Power Systems Electric Load Forecasting with Genetic Neural Networks F J Marin and F Sandoval Multiobjective Pressurised Water Reactor Reload Core Design Using a Genetic Algorithm G T Parks Using Artificial Neural Networks to Model Non-Linearity in a Complex System P Weller, A Thompson and R Summers 49 53 58 viii 'Iransit Time Estimation by Artificial Neural Networks T Tambouratzis, M Antonopoulos-Domis, M Marseguerra and E Padovani 62 Evolware Evolving Asynchronous and Scalable Non-uniform Cellular Automata M Sipper, M Tomassini and M S Capcarrere One-Chip Evolvable Hardware: 1C-EHW H de Garis 66 71 Vision Evolving Low-Level Vision Capabilities with the GENCODER Genetic Programming Environment 78 P Ziemeck and H Ritter NLRFLA: A Supervised Learning Algorithm for the Development of Non-Linear Receptive Fields S L Funk, Kumazawa and J M Kennedy Fuzzy-tuned Stochastic Scanpaths for AGV Vision J Griffiths, Q H Mehdi and N E Gough On VLSI Implementation of Multiple Output Sequential Learning Networks A Bermak and H Poulard 83 88 93 Speech/Hearing Automated Parameter Selection for a Computer Simulation of Auditory Nerve Fibre Activity using Genetic Algorithms C P Wong and M J Pont Automatic Extraction of Phase and Frequency Information from Raw Voice Data S McGlinchey and C Fyfe 98 103 A Speech Recognition System using an Auditory Model and TOM Neural Network 107 Fahlman-Type Activation Functions Applied to Nonlinear PCA Networks Provide a Generalised Independent Component Analysis 112 E Hartwich and F Alexandre M Girolami and C Fyfe Blind Source Separation via Unsupervised Learning B Freisleben, C Hagen and M Borschbach 116 Signal/Image Processing and Recognition Neural Networks for Higher-Order Spectral Estimation F.-L Luo and R Unbehauen Estimation of Fractal Signals by Wavelets and GAs H Cai and Y Li Classification of 3-D Dendritic Spines using Self-Organizing Maps G Sommerkorn, U Seiffert, D Surmeli, A Herzog, B Michaelis and K Braun Neural Network Analysis of Hue Spectra from Natural Images C Robertson and G M Megson 121 126 129 133 ix Detecting Small Features in SAR Images by an ANN Finch, D F Yates and L M Delves Optimising Handwritten-Character Recognition with Logic Neural Networks G Tambouratzis 138 143 Medical Applications Combined Neural Network Models for Epidemiological Data: Modelling Heterogeneity and Reduction of Input Correlations M H Lamers, J N Kok and E Lebret A Hybrid Expert System Architecture for Medical Diagnosis L M Brasil, F M de Azevedo and J M Barreto Enhancing Connectionist Expert Systems by lAC Models through Real Cases N A Sigaki, F M de Azevedo and J M Barreto 147 152 157 G A Theory and Operators A Schema Theorem-Type Result for Multidimensional Crossover M.-E Balazs Mobius Crossover and Excursion Set Mediated Genetic Algorithms S Baskaran and D Noever The Single Chromosome's Guide to Dating M Ratford, A Tuson and H Thompson A Fuzzy Taguchi Controller to Improve Genetic Algorithm Parameter Selection C.-F Tsai, C G D Bowerman, J l Tait and C Bradford Walsh Functions and Predicting Problem Complexity R B Heckendorn Migration through Mutation Space: A Means of Accelerating Convergence in Evolutionary Algorithms H Copland and T Hendtlass 161 166 171 175 179 183 GA Models/Representation Dual Genetic Algorithms and Pareto Optimization M Clergue and P Collard Multi-layered Niche Formation C Fyfe Using Hierarchical Genetic Populations to Improve Solution Quality J R Podlena and T Hendtlass A Redundant Representation for Use by Genetic Algorithms on Parameter Optimisation Problems A J Soper and P F Robbins 188 193 198 202 GA Applications A Genetic Algorithm for Learning Weights in a Similarity Function Y Wang and N Ishii 206 x Learning SCFGs from Corpora by a Genetic Algorithm B Keller and R Lutz Adaptive Product Optimization and Simultaneous Customer Segmentation: A Hospitality Product Design Study with Genetic Algorithms E Schifferl Genetic Algorithm Utilising Neural Network Fitness Evaluation for Musical Composition A R Burton and T Vladimirova 210 215 219 Parallel GAs Analyses of Simple Genetic Algorithms and Island Model Parallel Genetic Algorithms T Niwa and M Tanaka Supervised Parallel Genetic Algorithms in Aerodynamic Optimisation D J Doorly and J Peiro 224 229 Combinatorial Optimisation A Genetic Clustering Method for the Multi-Depot Vehicle Routing Problem S Salhi, S R Thangiah and F Rahman A Hybrid Genetic / Branch and Bound Algorithm for Integer Programming A P French, A C Robinson and J M Wilson Breeding Perturbed City Coordinates and Fooling Travelling Salesman Heuristic Algorithms R Bradwell, L P Williams and C L Valenzuela Improvements on the Ant-System: Introducing the MAX-MIN Ant System T Stiitzle and H Hoos A Hybrid Genetic Algorithm for the 0-1 Multiple Knapsack Problem C Cotta and J M Troya Genetic Algorithms in the Elevator Allocation Problem J T Alander, J Herajiirvi, G Moghadampour, T Tyni and J Ylinen 234 238 241 245 250 255 Scheduling/Timetabling Generational and Steady-State Genetic Algorithms for Generator Maintenance Scheduling Problems K P Dahal and J R McDonald Four Methods for Maintenance Scheduling E K Burke, J A Clarke and A J Smith A Genetic Algorithm for the Generic Crew Scheduling Problem N Ono and T Tsugawa Genetic Algorithms and the Timetabling Problem B C H Turton Evolutionary Approaches to the Partition/Timetabling Problem D Corne 259 264 270 275 281 xi Telecommunications - General Discovering Simple Fault-Tolerant Routing Rules by Genetic Programming I M A Kirkwood, S H Shami and M C Sinclair The Ring-Loading and Ring-Sizing Problem J W Mann and G D Smith Evolutionary Computation Techniques for Telephone Networks Traffic Supervision Based on a Qualitative Stream Propagation Model I Servet, L Trave-Massuyes and D Stern NOMaD: Applying a Genetic Algorithm/Heuristic Hybrid Approach to Optical Network Topology Design M C Sinclair Application of a Genetic Algorithm to the Availability-Cost Optimization of a Transmission Network Topology B Mikac and R Inkret Telecommunications - 289 294 299 304 FAP Breeding Permutations for Minimum Span Frequency Assignment C L Valenzuela, A Jones and S Hurley A Practical Frequency Planning Technique for Cellular Radio T Clark and G D Smith Chaotic Neurodynamics in the Frequency Assignment Problem K Dorkofikis and N M Stephens A Divide-and-Conquer Technique to Solve the Frequency Assignment Problem A T Potter and N M Stephens Applications - 285 308 312 317 321 General Heuristics Genetic Algorithm Based Software Testing J T Alander, T Mantere and P Turunen An Evolutionary /Meta-Heuristic Approach to Emergency Resource Redistribution in the Developing World A Tuson, R Wheeler and P Ross Automated Design of Combinational Logic Circuits by Genetic Algorithms C A Coello Coello, A D Christiansen and A Hernandez Aguirre Forecasting of the Nile River Inflows by Genetic Algorithms M E EI-Telbany, A H Abdel- Wahab and S I Shaheen Evolutionary ANN s I - 325 329 333 337 RBFs A Comparative Study of Neural Network Optimization Techniques T Ragg, H Braun and H Landsberg GA-RBF: A Self-Optimising RBF Network B Burdsall and C Giraud-Carrier Canonical Genetic Learning of RBF Networks Is Faster R Neruda 341 346 350 xii Evolutionary ANNs II The Baldwin Effect on the Evolution of Associative Memory A Imada and K Araki Using Embryology as an Alternative to Genetic Algorithms for Designing Artificial Neural Network Topologies C MacLeod and G Maxwell 354 359 Evolutionary ANNs III Empirical Study of the Influences of Genetic Parameters in the Training of a Neural Network P Gomes, F Pereira and A Silva Evolutionary Optimization of the Structure of Neural Networks by a Recursive Mapping as Encoding B SendhoJJ and M Kreutz Using Genetic Engineering To Find Modular Structures for Architectures of Artificial Neural Networks C M Friedrich Evolutionary Learning of Recurrent Networks by Successive Orthogonal Inverse Approximations C Gegout 364 368 373 378 Reinforcement Learning Evolutionary Optimization of Neural Networks for Reinforcement Learning Algorithms H Braun and T Ragg Generalising Experience in Reinforcement Learning: Performance in Partially Observable Processes C H C Ribeiro 384 389 Genetic Programming Optimal Control of an Inverted Pendulum by Genetic Programming: Practical Aspects F Gordillo and A Bernal Evolutionary Artificial Neural Networks and Genetic Programming: A Comparative Study Based on Financial Data S.-H Chen and C.-C Ni A Canonical Genetic Algorithm Based Approach to Genetic Programming F Oppacher and M Wineberg Is Genetic Programming Dependent on High-level Primitives? D Heiss-Czedik DGP: How To Improve Genetic Programming with Duals J.-L Segapeli, C Escazut and P Collard Fitness Landscapes and Inductive Genetic Programming V Slavov and N I Nikolaev 393 397 401 405 409 414 625 neurons, with evolved cellular automata (CA) based neural circuits, by the year 2001 We already have 10 million neurons, and expect to achieve our target on time By evolving neural net modules with roughly 100 neurons each, at electronic speeds (e.g in less than a second) in special FPGA (XC6264 chip) based evolvable hardware (called a CAM-Brain Machine (CBM)), we will be able to download these CA based neural circuit modules (each with its own evolved user specified function) into user specified brain architectures embedded in a RAM based space of trillions of CA cells The same CBM (programmable) hardware then updates the whole RAM CA space frequently enough (e.g 30 times a second, i.e at over 100 billion CA cells a second) for real-time operation By the end of 1997, our Brain Builder Group (BBG) expects to see the completion of parallel tasks, namely the design and fabrication of the CBM (which was started in January 1997), the construction of a robot kitten called 'ROBOKONEKO' (in Japanese), and the creation of a 10,000 module artificial brain architecture to control the robot kitten's many behaviors The modules for the artificial brain will be evolved in 1998 with the CBM, and put into the (life sized) kitten robot After 1998, the BBG hopes to work on more ambitious projects, such as household cleaner robots, and with substantially more brain builder researchers on the team (One of my goals for Japan is to see the country create a 'J-Brain Project', which would aim to build a 10,000,000 module artificial brain with 2000 human 'EEs' (evolutionary engineers) over the time period 2000-2005) 20 years from now, brain-like computers should generate a trillion dollar industry Workshop Summary UEA, Norwich, England, April 1, 1997 The Conference was preceded by a one-day Workshop in which the morning was devoted to the theory and applications of artificial neural networks and the afternoon to an introduction to heuristic search, including genetic algorithms, followed by a detailed look at some case studies The workshop sessions and associated abstracts are as follows: An Introduction to Artificial Neural Networks Professor Nigel Steele, Coventry University, UK This session, as the opening session of the workshop, is aimed at those with limited knowledge of the field of artificial neural networks and is not primarily intended for experienced practitioners An introduction to the field is given, covering the basic concepts of network structure and network learning Attention is focussed initially on the multi-layer percept ron network, with learning by error backpropagation Ideas on improving learning performance are discussed and some practical hands-on experience incorporated Subsequently, the radial basis function network is introduced, in both its standard and adaptive forms The close relationship of this type of network and fuzzy inference systems will also be discussed Again, some practical, hands-on experience will be available The session will close with a discussion of an application in the field of robot navigation Hardware Neural Networks Professor Kevin Warwick, University of Reading, UK Design and Application Certain types of artificial neural networks are well suited for implementation in hardware This session introduces the basic principles of such networks and describes their method of operation Although a range of networks is considered, Kohonen and Hopfield type networks are looked at in detail and digital networks such as the n-tuple type are also presented Applications of the networks described are quite widespread, hence during this session, a number of actual applications will be described Emphasis will be placed on the principles of the application, reasons for employing a neural network in each case and the positive and negative features which result Finally, we take a look at future implementation possibilities An Introduction to GAs and Other Search Paradigms Dr Colin Reeves, Coventry University, UK Genetic algorithms (GAs) have become popular tools for solving difficult optimisation problems over the last decade This part of the workshop will provide an introduction to the concept of GAs, using numerical examples It will also explain some of our current understanding of how they work, and deal with some of the important issues involved in their implementation The workshop will also compare and contrast GAs with other recent general search paradigms such as simulated annealing, tabu search and perturbation methods It will thus lead naturally into the final workshop session 627 Case Studies in Genetic Algorithms and Other Search Paradigms Mr Jason Mann, Nortel, Harlow, UK In this session, we look at key issues in the application of GAs and related heuristics, such as simulated annealing, tabu search and others This is done through a number of case studies drawn from real-world applications that the UEA research group MAG has carried out Of particular interest are representation issues and the associated choice of appropriate (genetic) operators These case studies, drawn particularly from applications in the telecommunications sector, are supported by laboratory sessions in which participants will have the opportunity to use some of the latest software toolkits supporting the respective search paradigms Subject Index accelerating convergence, 183 adaptive resonance theory (ART), 219,220, 222 airfoil design, 229 Akaike Information Criterion (AIC), 338, 500-503 Ant system (AS), 245-248 associative memory, 354-357,450 auditory processing, 98, 99 autonomous guided vehicle (AGV), 88, 91, 450, 452 autoregressive model (AR), 51,468,470,472 non-linear with exogenous variables (NARX), 597, 600 self-exciting threshold (SETAR), 473 with exogenous variables (ARX), 50 autoregressive moving average (ARMA), 121-124, 126, 337-339, 468 periodic (PARMA), 337-340 backpropagation (BP), 2, 4, 8, 31, 34-36, 51, 59, 62-65,83-87,93,135,136,140,149,152, 155, 184, 364, 368, 371, 376, 437, 438, 442, 443, 468, 488, 514, 518-521, 523, 527, 537, 540-542, 568, 589, 593-595, 602, 603 complementary reinforcement backpropagation learning (CBRP), 6-9 resilient (RPROP), 343, 375, 376, 505-507 Baldwin effect, 354-357 Barycentric correction procedure (BCP), 527, 528, 530, 531 Bernoulli mixture model, 44-47 bias-variance dilemma, 41 blast furnace walls, 532 blind source separation (BSS), 112, 113, 116, 117, 119 BLX-0.5 operator, 202, 204, 205 branch and bound (B&B), 238-240,250,252,253, 259,264,317 B-splines, 229 CAM-Brain, 75 cellular automata (CA), 66-69, 71, 75 non-uniform, 66-69 cellular programming, 66, 68, 69 centre point selection, 514 chaotic neurodynamics, 317 character recognition, 143, 144, 146, 198, 210, 362 CHC adaptive search algorithm, 204 classification, 4, 32, 40, 42, 43, 64, 82, 93, 94, 105, 106, 108, 109, 124, 129-133, 135, 136, 155, 185, 186, 200, 219, 220, 222, 242, 266, 309, 344, 346, 347, 349, 374376, 378, 422, 435, 442-444, 485, 502, 514-516, 518, 523, 525, 527-529, 531535, 537-543, 552, 553, 579 digital, 31 error, 341, 344 pattern, 40, 44, 45, 47, 149 supervised, 532 unsupervised, 532 vowel, 103-105, 527, 530 classifier systems (CSs), 19, 193, 449, 539, 540, 555, 611, 612, 615-618 learning (LCS), 611-614 clustering, 131, 149, 220, 234-237, 346-349, 454457,547,551,555,556,566,568,570-573 adaptive, 234, 235 k-means, 346-349, 515 non-stationary, 555, 556 sequence, 454 stochastic, 551 unsupervised, 551 codevectors, 461 co-evolution, 575, 576, 579, 581-583, 586, 587 combinational logic circuit design, 333, 335 combinatorial optimisation, 378 competing conventions problem, 509 complex system, 58 conditional probability, 45, 415, 416, 479, 542, 565 densities, 477-480 connection engineering, 157, 158 connectionism, 449, 568, 570-573 629 constraint satisfaction (C-SAT), 275 constructive algorithms, 93, 96, 341, 527, 528, 531 context-free grammar (CFG), 210, 211, 288 stochastic (SCFG), 210-213 controllers, 7, 12-14, 16-19, 21, 23, 24, 175-177, 255-257, 382, 389, 393-396, 405, 445448, 575-578, 580, 582, 588, 592-595, 606, 608, 609 animat,510 fuzzy, 607-609 Taguchi, 175-177 microcontrollers, 88, 91 model-based, 601 monolithic, 17, 19 neural network, 11-14,21,23,24 PID, 575-578 robot, 7, 11, 12, 16, 17, 21, 26, 445, 448 co-operation, 17, 19, 50, 110, 543, 544, 615-617, 619-623 crossover, 21-23, 27-29, 50, 54, 55, 68, 74, 80, 100, 127, 128, 154, 155, 161-164, 166168, 171, 173, 174, 176, 183, 184, 193, 198, 202-205, 207, 212, 215-217, 222, 227, 232, 235, 236, 239, 251, 261, 262, 265, 268, 271, 272, 275, 277, 278, 292, 301, 302, 306, 331-333, 338, 347, 351, 355, 365, 369, 371-373, 393, 401-404, 406, 410-412, 419-421, 511, 548, 549, 561, 577, 613 blend, 204 column, 22 combinatorial, 275 cycle (CX), 310, 311 heuristic tie-breaking (HTBX), 55 link-block, 302 link-group, 301, 302 masked, 222, 333 multidimensional, 161, 163, 164 multi-point, 222, 616 116bius, 166-168, 170 n-dimensional, 161, 251 one-point, 161, 163, 164, 167, 168, 212, 220, 243,265,300-302,355,402,403,620 order (OX), 310, 311 partially matched (P11X), 310, 311, 330 modified, 330 permutation, 310, 311 position-based, 275, 278 single-point, 338, 613 Sub-graph Active-Active Node (SAAN), 420 Sub-graph Inactive-Active Node (SIAN), 420, 421 two-dimensional, 162-164 two-point, 173, 222, 243, 261, 272, 273, 306, 355, 581 uniform, 184, 225, 243, 282, 355 customer segmentation, 215-217 data collection, 215, 216 data mining, 547,548 undirected, 547 decision trees, 405, 414, 415, 417, 497, 543, 544 dendritic spines, 129, 131 developmental artificial neural networks (DANNs), 509, 510 diffusion models, 224 distal learning, 592-594 divide and conquer, 321-323 evolutionary (EDAC), 241, 242 dominance, 54, 608 double convex region benchmark, 40, 42, 43 dynamical systems, 526, 579, 592 identification, 588, 590, 591 economic order quantity (EOQ), 584-586 elevator allocation problem, 255, 256 embryological algorithm, 359 encoding, 8, 11, 54, 161-164, 167, 182, 189,202, 211, 212, 216, 219, 229, 251, 253, 262, 265, 275, 277, 278, 300, 329, 333, 335, 347, 349, 368, 369, 371, 373, 389, 399, 430,431,449,455,612,620 autoencoding, 150 cellular, 373, 374, 376, 419 direct surface, 229, 230 8-bit,51 JPEG,133 linear, 161, 163, 301 multidimensional, 161, 163 n-dimensional, 161 non-linear autoencoding networks, 148, 150 one-dimensional, 163, 164 recursive, 368, 369, 372 temporal, 110 two-dimensional, 162-164 630 ensembles, 31, 33, 34, 477, 479, 480 epidemiological dataset, 147, 148 equity valuation, 440, 441, 444 evolutionary algorithms (EA), 12, 14, 16, 17, 2631, 68, 183, 185, 186, 281, 282, 284, 286, 330-332, 341, 342, 344, 368, 371-375, 380, 381, 384, 406, 414, 473, 547, 549, 550, 577, 583-586 evolutionary neural networks (EANNs), 49, 384, 397-399,473,475 evolutionary program induction (EPI), 401-404 evolutionary topology optimization (ENZO), 341344, 384, 386-388 evolution strategies (ES), 202, 555-558,560-563 evolvable hardware (EHW), 71-73, 75, 76 one-chip (IC-EHW), 71-76 evolware, 66, 67, 69 excursion sets, 166, 169, 560, 561, 564 expectation maximisation (EM) algorithm, 478, 480 expert networks, 7, 8, 44-47, 500 expert systems (ESs), 152,155,157-159,176,433, 440-443 hybrid (HES), 152 neural network (NNES), 152-155 rule-based (RBES), 152, 155 Fahlman type activation functions, 112-115 feedforward neural networks (FFNN), 11-13, 23, 25,41,44,49-51,133,147-152,157,158, 346, 350, 365, 371, 374, 378-382, 397, 436, 438, 439, 442, 448, 473, 477, 478, 491, 495, 518, 520, 521, 523, 526, 532535,537,601,602,604 static (SFANN), 588 finite impulse response (FIR), 465 finite state automata (FSA), 460, 616, 617, 619, 620 fitness landscapes, 166, 169, 173, 251, 364, 408, 414-416, 560, 564, 581 forecasting, 49, 337-340, 441, 442, 444, 467, 468, 475 electric load, 49, 51 formal languages, 210, 212 fractal signals, 126, 128 frequency assignment problem (FAP) , 310, 312, 313,317-319,321,323 minimum span (MSFAP), 308 Frey and Slates' letter recognition problem, 34, 199, 200 fuzzy logic, 88, 89, 133, 153, 155, 175, 256, 606609 fuzzy neural networks, 539-542 fuzzy vector bundles, 523 gamma test, 425 gate arrays dynamically programmable (DPGAs), 73 field programmable (FPGAs), 71-73, 75, 76 gating network, 6-9, 44-47, 500-504 Gaussian mixture conditional density model (GMCD), 44, 46, 47, 477-480 GENCODER, 78-80 gene pooling, 229, 347, 348 generalisation, 5, 31, 32, 35, 37-41, 43, 133, 136, 389, 526, 573 generic growth rules, 509 GENESIS, 260, 511 Genetic Algorithms (GAs) co-evolutionary (CGA), 579-582 cooperative coevolutionary (CCGA), 583-587 distributed (DGA), 230-233, 253 dual (DGA), 188-192,409,410,412 excursion set mediated (ESMGA), 166-170 generational, 251, 259-262 Goldberg's simple (SGA), 220 hierarchical (HGA), 198-200 hybrid GA, 238, 239, 250-253, 264, 299, 300 interactive, 215 multiobjective (MOGA), 55, 56 multi-population, 198 niched Pareto (NPGA), 188 parallel (PGA), 224, 226, 227, 229-231, 233, 236 steady state, 164, 173, 259-262, 270, 282 structured (SGA), 193, 194, 196 genetic engineering, 373-376 genetic programming (GP), 18, 78, 79, 285-288, 333, 334, 373-375, 393, 396-399, 401410,412,414,419-422,445,510 dual-based (DGP), 409, 410, 412 multi-population, 285, 286, 288 neural networks (GPNN), 397-399 parallel distributed (PDGP), 419-423 631 uni-population (UGP), 287, 288 GENITOR, 260, 261, 581 grammatical inference, 210, 211, 213 graph planarisation problem, 428-431 graph rewriting, 373, 375, 376 harmony theory (HT), 428-431 Heaviside function, 496 Heaviside perceptron networks, 495, 497, 498 Hebbian learning, 103, 104, 354-357, 465, 570 Hebbian matrix, 356, 357 Hebbian model, 464 Hessian matrix, 40, 41, 342 high-level primitives, 405, 406, 408 hill climbing, 98, 188, 250, 450, 510, 547, 548, 550, 599, 600, 612 first-ascent (FAHC), 330, 331 steepest-ascent (SAHC), 330, 331 stochastic (SHC), 329-332, 548-550 Hopfield neural network model, 318, 319, 354, 357 hybrid learning system (HLS), 449-452 image processing, 22, 78, 79, 81, 129, 130, 138 synthetic aperture radar (SAR), 138-142 incest prevention, 171-173 independent component analysis (ICA), 112-116 information content, 342, 565 integer programming, 238, 239, 259, 264, 268 intelligent data analysis, 543 interactive activation and competition (lAC), 6265,157-159,457 internal model control (IMC), 601, 603, 604 introns, 420 inventory control, 583, 584 inverse kinematics, 21, 25 inverse modelling, 592 inverted pendulum, 393-396 Iris data set, 349, 500, 502 classification problem, 199, 200, 502 Kalman neural gas network (KNGN), 468, 470, 471 Kohonen layer, 131 size-adaptive, 131 Kohonen network, 103, 424-426, 532-534 Kullback-Leibler divergence, 461 A-calculus, 405, 406, 408 learning vector quantization (LVQ), 148-151, 196, 436 learning weights, 206, 209 Levenberg-Marquardt algorithm, 40, 41, 51, 601603 linear ANN, 50 linear quadratic regulator (LQR) design, 395, 396 LISP, 79, 285, 286, 397, 398,409,612 MAX-MIN Ant System (MMAS), 245,246,248 Meddis computational model, 98,99, 101 meta neural networks (MNN), 505-508 meta-heuristic, 308, 329, 341, 344, 384 mexican-hat function, 376 minimum deceptive problem (MDP), 410, 412 minimum description length (MDL) principle, 211, 398 MITER, 565, 568 modular neural networks (MNN), 7, 35, 38, 44, 47, 376, 447, 448, 468, 469, 471, 500, 501, 503, 504 multilayer perceptron (MLP), 35-43, 45,202,341, 346, 348, 419, 424-426, 446, 447, 468, 470-472, 535, 537, 538, 580 multiobjective optimisation (MOO), 53, 54, 56, 188, 189, 193 multiobjective problem (MOP), 53, 188 minimal (MMOP), 189, 191 musical rhythm composition, 219 mutual information (MI), 113, 114,537,538,565568 nearest neighbour heuristic algorithm (NNHA), 31, 33, 34, 241-244, 248 Nine Men's Morris, 384, 385, 387, 388 NOMaD, 299-302 Non-Linear Dimensionality Reduction (NLDR), 150 non-linear receptive field (NLRF), 83-87 learning algorithm (NLRFLA), 83, 85-87 nuclear reactor, 58-61 boiling water (BWRs), 62, 65 obstacle recognition, optical network topology design, 299 optical pan-European network (OPEN), 299 optimal brain damage (OBD), 40, 41, 43, 342 optimal brain surgeon (OBS), 40, 41, 43, 341-344 632 Pareto optimisation, 54, 188, 189 partially observable processes, 389, 390 pattern recognition, 130, 210, 334, 361, 442, 532, 539, 542 peak expiratory flow (PEF), 147-150 polling systems, 436-439, 505, 507 population-based incremental learning (PBIL) , 193, 194, 196 structured (SPBIL), 194-196 possibility distributions, 539, 540 predictive maintenance (PM), 432,433 pressurised water reactor (PWR), 53, 55 reload core design problem, 53, 55 principal component analysis (PCA), 40, 42, 43, 104, 113, 116, 130 non-linear, 112-115 prisoner's dilemma, 615, 616, 619 iterated (IPD), 615 repeated (RPD), 619 probabilistic neural networks (PNN), 443, 468, 514-517 reduced (RPNN), 514-517 process control, 175, 177, 178, 579, 582, 601, 606 process identification, 588, 590, 602 pruning, 40-42, 341, 342, 344, 354-356, 373, 516, 517, 551, 552, 565 destructive, 41 magnitude based (MbP), 341-344 network, 41, 361 principal components (PCP), 40 stochastic, 551 quantum computation, 482 quantum neuron, 482-485 quickpropagation, 135 quickprop algorithm, 113, 114, 135 Q-Iearning, 389-391, 616 radial basis function (RBF) networks, 50, 346-348, 350-353,446,514 random vector functional link (RVFL), 477-480 recurrent neural networks (RNN) , 84, 85, 318, 372, 378-382, 397, 436-439, 442, 459462,468,488,491,493,588 Elman recurrent networks, 8, 50, 51 recursive mapping, 368, 369 redundant representation, 202, 205 reinforcement learning (RL), 6, 8, 18, 384, 385, 388, 389, 391, 582, 615 associative, 6, relational schemata, 189 relearning, 31-34 resource management, 329 resource redistribution, 329 ring loading problem, 289, 290, 292 ring sizing problem, 289 robotics, 11 evolutionary, 11, 17, 19 robots, 1-3,6, 11-16, 19, 21-24, 26-28, 364-366, 445, 447, 448 adaptive, 11 calibration, 25 hexapod, 21 Khepera, 11-14 manipulators, 21, 25, 26 mobile, 1, 2, 6, multirobots systems, 29 robot arm, 21, 23, 24,44,448 trajectory, 26 rote learning, 157, 158 routing, 72, 73, 234-237, 285-289, 299, 302, 304, 337, 544 fault-tolerant, 285, 286, 288 multi-depot vehicle routing problem (MDVRP), 234-237 network, 285, 288 single-row, 428, 429 vehicle routing problem (VRP), 234, 235 rules extraction, 439, 543, 544, 565-568 generation, 565 induction, 547-550 insertion, 543, 544 saccades, 88-91 saliency, 40, 41, 43, 537 satisfiability problem, 238, 239 scheduling, 67, 260, 264, 268, 269, 275, 327 crew scheduling problem (CSP), 270-272 generator maintenance scheduling (GMS), 259, 260, 262, 264 permutation-encoded, 330 seduction function, 171, 172 self organising maps (SOM), 35, 86, 87, 129, 131, 551 633 Kohonen feature maps (SOFM), 103, 107, 424, 425, 449-452, 459, 461, 532 sequential assignment algorithms, 308-310 sequential learning, 93, 94, 527, 528, 530, 531 SETAR(2;kl ,k2 )-model, 473 signal processing, 107, 121, 126 similarity function, 206-209, 390 linear, 208 nonlinear, 208 similarity metrics, 171, 172 simulated annealing (SA), 21, 188, 264-269, 275, 289, 312-317, 321-323, 330, 331, 346, 428, 430, 431, 436, 437, 547-551 stochastic, 319 simulation, 23, 24, 98-100, 203, 409-412, 438, 580 simulator, 11, 12, 17,56,58-60,255,256,448,473, 475 training, 17 softmax non linearity, 44 software testing, 325, 327 spectra, 133 bispectral estimation, 121, 123, 124 entropy, 459, 462 higher order, 121 estimation, 121, 124 hue, 133, 134, 136 speech, 109, 110, 112, 114, 210 enhancement, 110 natural, 112, 114 part-of-speech tagging, 210 perception, 98 processing, 107, 110 processing system, 107 recognition, 107-110, 210 recognition system, 107 voice data, 103, 104 stability-plasticity dilemma, 220 stochastic competitive evolutionary neural tree (SCENT), 551-554 stochastic machines (SMs), 459-462 stochastic transition matrices, 88, 89 stress testing, 326 automated dynamic, 325 structure identification, 597, 602 structured evolution, 16-18 subpopulations, 183-186, 217, 226, 227, 229, 230, 583-587 SUGAL, 178 symbolic, 570-573 symbolic sequences, 459-462 symbolism, 570 synchronous digital hierarchy (SDH), 289 synchronous optical network (SONET), 289 systolic architecture, 93, 95, 96 T-colouring problem, 321 tabu search (TS), 31, 234, 264, 265, 268, 269, 275, 317,321,330,331,547-550 Taguchi method, 175-177 telecommunications, 285, 288, 289, 299, 304, 436 temporal difference learning (TD(A)), 384-388 temporal organization map (TOM), 107-109 thermal profiles, 532 3D artificial neural structures, 509 threshold accepting (TA), 330, 331 time delay networks, 454 time series, 126, 337, 338, 368, 370-372,441, 442, 455, 457, 462, 464, 465, 467-475, 477, 479,480 Box-Jenkins, 337, 338 prediction, 337, 340, 370, 464, 468, 471 timetabling problem, 275, 277-279, 281, 282, 284 partition (PTP), 281-284 Tit-For-Tat (TFT), 616,617,620-622 transit time estimation, 62-65 decimated, 64, 65 integer, 64 transmission network topology, 304 travelling salesman problem (TSP), 234, 241, 245, 248, 278, 310, 318 asymmetric (ATSP), 245, 248 geometric (GTSP), 241,242 symmetric, 245, 248 two spirals problem, 46 ultrasonic sensor, 1, variable selection, 535 vision, 21, 78, 79, 81-83, 88, 91 early, 78, 82 what and where tasks, 500, 502, 503 VLSI implementation, 93, 94, 96, 428, 527, 531 Walsh functions, 179 Walsh sums, 181, 182 634 wavelength-agile optical transport and access network (WOTAN), 299 wavelet transform, 126, 127, 346 wave propagation, 486 weighted sum, 41, 95, 155,260,282,318,355,465, 468, 486, 487, 545, 549 Wright-Fisher model (W-F model), 224 XOR problem, 46, 185, 351, 352, 419, 421, 422, 484, 485 two-input, 484 Young's Information Criterion (YIC), 598 0-1 multiple knapsack problem (0-1 MKP), 250, 251, 253 SpringerE urographies Wilfrid Lefer, Michel Grave (eds.) extensions and styles of specification Techniques are being developed that facili- Daniel Thalmann, Michiel van de Panne (cds.) tate the expression of user-oriented re- Visualization in Scientific Computing '97 quirements and the refinement and checking of specifications of interactive Computer Animation and Simulation '97 systems This book reflects the state of the Proceedings of the Eurographics Workshop art in this important area and also contains Proceedings of the Eurographics Workshop in a summary of working group discussions in Budapest, Hungary, about how the various techniques repre- September 2-3, 1997 Boulogne~sur-Mer, France, April 28-30, 1997 sented might be applied to a common case 1997.92 partly coloured figures VII, 187 pages study 1997 121 partly coloured figures VIIl, 203 pages Soft cover DM 85,-, 55 595,-, US $ 59.00 Soft cover DM 89,-, 55 625,-, US $ 59.00 ISBN 3-211-83049-9 ISBN 3-211-83048-0 Visualization is now recognized as a powerful approach to get insight in large datasets Julie Dorsey, Philipp Slusallek (eds.) Rendering Techniques '97 III various domains such as finance, physics, ters, simulated animals, fluids, and other dynamic phenomena The animation tech- book cover technical aspects as well as concreteapplications of visualization problem of synthesizing the realistic movement and behaviour of human-like charac- produced by scientific experimentations and simulations The contributions to this The contributions to this book address the Proceedings of the Eurographics Workshop niques are driven by the goals of efficiency, in SI Etienne, France, June 16-18, 1997 as required by real-time interactive animations, and quality, as demanded byanima- astronomy and medicine, providing researchers and engineers with valuable 1997 172 partly coloured figures IX, 342 pages tions used in feature films This series of information for setting up new powerful Soft cover DM 118,-,55826,-, US $ 74.95 workshops provides a high-quality interna- environments ISBN 3-211-83001-4 tional forum for the exchange of new ideas The papers in this volume present new on, simulation of dynamic natural pheno- related to the themes of character animati- Michael Douglas Harrison, Juan Carlos Torres (eds.) research results in the areas of finite-ele- mena, motion capture and analysis, physi- ment and Monte-Carlo illumination algo- cally-based modeling, behavioral anima- rithms, image-based rendering, ray tracing, tion, and visualization Design, Specification and Verification ofInteractive Systems '97 clustering techniques, texture generation and sampling, and efficient hardware rendering While some contributions report results from more efficient or elegant algo- Proceedings of the Eurographics Workshop in Granada, Spain, June 4-6, 1997 rithms, others pursue new and experimental approaches to find better solutions to the open problems in rendering 1997 129 figures VIII, 320 pages Soft cover DM 118,-, 55 826,-, US $ 79.95 ISBN 3-211-83055-3 An increasing recognition of the role of the human-system interface is leading to new SpringerWienNewYork Sachsenplatz 4-:6 P.O.Box 89, A-120l Wien, Fax +43-1-330 24 26,.e-mail: order@springer.at, Internet: http~/www.!\pringer.at New York, NY 10010, 175 Fifth Avenue D-14197 Berlin, Heidelberger Platz Tokyo 113, 3-13, Hongo 3-chom~, Bunkyo-ku SpringerAdvances in Con1puting Science Franc Solina, Walter G Kropatsch, Reinhard KIette, Ruzena Bajesy (cds.) Series Advances in Computing Science Chris Brink, Wolfram KahL Gunther Schmidt Advances in Computer Vision Relational Methods in Computer Science Springer-Verlag Wien New York presents 1997 % figures VIII, 266 pages 1997 30 figures Xv, 272 pages the new book series "Advances in Com- Soft cover DM 69,-, tiS 485,-, US $ 44.95 Soft cover DM 69,-, tiS 485,-, US $ 49.95 puting Science" Its title has been chosen ISBN 3-211-83022-7 ISBN 3-211-82971-7 to emphasize its scope: "computing science" comprises all aspects of science and mathematics relating to the computing pro- Computer vision solutions used to be very The calculus of relations turned into an cess, and has thus a much broader meaning specific and difficult to adapt to different or important conceptual and methodological than implied by the term "computet sci- even unforeseen situations The current tool in computer science The methods ence"- The series is expected to include development is calling for simple to use yet presented in this bOOk include questions of contributions from a wide range of disci- robust applications that could be employed relational including, for example, numerical in various situations This trend requires program specification, resource-conscious analysis, discrete mathematics and system the reassessment of some theoretical issues linear logic, semantic and refinement theory, natural sciences, engineering, infor- in computer vision A better general under- consideration, mation science, electronics, and naturally, standing of vision processes, new insights reasoning about programs, tabular methods computer science itself and better theories are needed The papers in software construction, algorithm de- Contributions in the form of monographs selected from the conference staged in velopment, linguistic problems, followed or collections of articles dealing with Dagstuhl in 1996 to gather scientists from by a comprehensive bibliography The advances in any aspect of the computing the West and the former eastern-block reader gets an overview of the wide-ranging process, or its applications are welcome countries address these goals and cover applicability of relational methods They must be concise, theoretically sound such fields as 2D images (scale space, computer science p
ines and written in English; they may have morphology, resulted, works, Hough transform, texture, pyra- e.g., from research projects, segmentation, neural net- advanced workshops and conferences The mids), recovery of 3-D structure (shape publications (of the series) should address from shading, optical flow, 3-D object re- not only the specialist in a particular field cognition) and bow vision is integrated into but also a more general scientific audience a larger task-driven framework (hand-eye An International Advisory Board which calibration, navigation, perception-action will review papers before publication will cycle) guarantee the high standard of volumes published within this series SpringerWienNewYork Sachsenplatz 4-6, EO.Box 89, A-120l Wien, Fax +43-1-330 24 26, e-mail: order@springer.at,lnternet:http://www.springer.at New York, NY IDOlO, 175 Fifth Avenue· D-14197 Berlin, Heidelberger Platz 3· Tokyo 113, 3-13, Hongo 3-chome, Bunkyo-ku databases, applications nonclassical logics to for in SpringerTexts and Monographs in Symbolic COlnputation Bob E Caviness, Jeremy R Johnson (eds.) Alfon,,;!) Miola, Marco Temperini (eds.) Norbert Kajler (cd.) Quantifier Elimination and Cylindrical Algebraic Decomposition Advances in the Design of Symbolic Computation Systems Human-Computer Interaction in Symbolic Computation 1998 20 figures XIX, 431 pages 1997.39 figures X, 259 pages 1998.68 figures Approx 230 pages Soft cover DM 118,-,05826,-, US $ 79.95 Soft cover DM 98,-, 05682,-, US $ 79.95 Soft cover DM 89,-, 05 625,-, US $ 59.95 ISBN 3-211-82794-3 ISBN 3-211-82844-3 ISBN 3-211-82843-5 George Collins' discovery of Cylindrical New methodological aspects related to There are many problems which current Algebraic a design and implementation of symbolic user interfaces either not handle well or method for Quantifier Elimination (QE) for computation systems are considered in this not address at all The contributions to the elementary theory of real closed fields volume aiming at integrating such aspects this volume concentrate on three main areas: interactive books, 'computer-aided Decomposition (CAD) as brought a major breakthrough in automat- into a homogeneous software environment ing mathematics with recent important for scientific computation The proposed instruction, and visualization They range applications in high-tech areas (e:g robot methodology is based on a combination of from a description of a framework for motion), also stimulating fundamental re- different techniques: 'algebraic specificao authoring search it; computer algebra over the past tion through modular approach and com- books and of a tool for the direct manipula- and browsing mathematical three decades pletion algorithms, approximated and exact tion of equations and graphs to the presen- This volume is a state-of-the-art collection algebraic object- tation "fnew techniques, such as the use of chains of recurrences for expediting the computing methods, of important papers on CAD and QE and on oriented programming paradigm, auto- the related area of algorithmic aspects of mated theorem proving through methods visualization of mathematical functions real geometry It contains papers from a it la Hilbert and methods of natural deduc- Students, symposium held in Linzin 1993, reprints of tion In particular the proposed treatment involved in the design and implementation of scientific software will be able to draw researchers, and developers seminal papers from the area including of mathematical objects, via techniques for Tarski's landmark paper as well as a survey' method abstraction, structures classifica- upon the presented research material here outlining the developments in CAD based tion, and exact representation, the pro- to create ever-more powerful and user- QE that have taken place in the last twenty gramming methodology which supports the friendly applications years design and implementation issues, and reasoning capabilities supported by the whole framework are described SpringerWienN ewYork Sachsenplatz 4-6, r.O.Box 89, A-120l Wien, Fax +43-1-330 24 26, e-mail: order@springer.at Internet: http://www.springer.at New York, NY 100lD, 175 Fifth Avenue· D-14197 Berlin, Heid~lberger Platz • Tokyo 113, 3-13, Hongo 3-chQrne, Bunkyo-ku SpringerComputerScience Computing Jean-Michel Jolion, Walter G Kropatsch (eds.) Walter G Kropatsch, Reinhard Klette, Franc Solina in cooperation with R Albrecht (eds.) Graph Based Representations in Pattern Recognition Theoretical Foundations of Computer Vision Archives for Informatics and Numerical Computation Editorial Board: 1998 76 figures Approx 170 pages 1996 87 figures VII, 256 pages R Albrecht, Innsbruck; H Brunner, SI John's; Soft cover DM 1l0,-, tiS 770,- Soft cover DM 165,-, oS 1155,- R E Burkard, Graz; W Hackbusch, Kiel; Reduced price for subscribers to "Computing": Reduced price for subscribers to "Computing"; G R Johnson, Fort Collins; Soft cover DM 99,-, oS 693,- Soft cover DM 148,50, tiS 1039,50 W Knodel, Stuttgart; W G Kropatsch, Wien; ISBN 3-211-83121-5 ISBN 3-211-82730-7 H J Stetter, Wien Computing, Supplement 12 Computing, Supplement 11 Computing publishes original papers and Graph-based representation of images is Computer Vision is a rapidly growing field short communications from all fields of sci- becoming a popular tool since it represents of research investigating computational entific computing in English Contributions in a compact way the structure of a scene to and algorithmic issues associated with im- may be of theoretical or applied nature, the be analyzed and allows for an easy manipu- age acquisition, processing, and under- essential criterion is computational rele- lation of sub-parts or of relationships be- standing It serves tasks like manipulation, vance Subject areas include discrete tween parts Therefore, it is widely used to recognition, mobility, and communication mathematics, symbolic computation, par- control the different levels from segmenta- in diverse application areas such as manu- allel computation, computer arithmetic, ar- tion to interpretation facturing, robotics, medicine, security and chitectural concepts for computers and net- The 14 papers in this volume are grouped virtual reality This volume contains a se- works, operating systems, programming in the following subject areas: hypergraphs, lection of papers devoted to theoretical languages, software engineering, perfor- recognition and detection, matching, seg- foundations of computer vision covering a mance and complexity evaluation, data mentation, broad range of fields, e.g motion analysis, bases, image processing, computer graph- representation implementation problems,