ICANN, the International Conference on Artificial Neural Networks, is the official conference series of the European Neural Network Society which started in Helsinki in 1991. Since then ICANN has taken place in Brighton, Amsterdam, Sorrento, Paris, Bochum and Lausanne, and has become Europe's major meeting in the field of neural networks. This book contains the proceedings of ICANN 98, held 2-4 September 1998 in Skovde, Sweden. Of 340 submissions to ICANN 98, 180 were accepted for publication and presentation at the conference. In addition, this book contains seven invited papers presented at the conference. A conference of this size is obviously not organized by three individuals alone. We therefore would like to thank the following people and organizations for supporting ICANN 98 in one way or another: . the European Neural Network Society and the Swedish Neural Network Society for their active support in the organization of this conference, . the Programme Committee and all reviewers for the hard and timely work that was required to produce more than 900 reviews during April 1998, . the Steering Committee which met in Skovde in May 1998 for the final selection of papers and the preparation of the conference program, . the other Module Chairs: Bengt Asker (Industry and Research), Harald Brandt (Applications), Anders Lansner (Computational Neuroscience and Brain Theory), Thorsteinn Rognvaldsson (Theory), Noel Sharkey (co­ chair Autonomous Robotics and Adaptive Behavior), Bertil Svensson (Hardware and Implementations), . the conference secretary, Leila Khammari, and the rest of the

Inhalt

Invited Presentations.- (Back) Towards Diagrammatic Representation and Reasoning in a Connectionist Framework.- Variational Learning in Graphical Models and Neural Networks.- Synchronization: The Computational Currency of Cognition.- Applications of Vapnik's Theory for Prediction.- Brains, Gases and Robots.- Self-Organization of Very Large Document Collections: State of the Art.- Silicon Artificial Neural Networks.- Oral Presentations: Theory: Theory I: Algorithms.- Efficient Top-Down Jacobian Evaluation of Tree-Structured Neural Networks.- Tractable Undirected Approximations for Graphical Models.- Convex Cost Functions for Support Vector Regression.- Asymptotically Optimal Choice of ?-Loss for Support Vector Machines.- Support Vector Regression with Automatic Accuracy Control.- Theory II: Dynamical systems, Time Series.- Queuing Theory for Spike Driven Synaptic Dynamics.- The Deformable Feature Map - Adaptive Plasticity for Function Approximation.- Asymptotic Behavior of a Neural Network with Dynamic External Input.- Theory III: Signal Decomposition Methods.- Independent Component Analysis for Time-Dependent Stochastic Processes.- Denoising of Sensory Data by Maximum Likelihood Estimation of Sparse Components.- Kernel PCA Pattern Reconstruction via Approximate Pre-Images.- The Pearson Mixture Model for Cluster Analysis and Data Visualisation.- Theory IV: Learning Theory.- Why Feed-Forward Networks are in a Bad Shape.- Transients and Asymptotics of Natural Gradient Learning.- An Analysis of Convergence in Generalized LVQ.- Ultrametric Structure in Autoencoder Error Surfaces.- The Dynamics of Matrix Momentum.- Dynamics of Batch Learning in Multilayer Neural Networks.- An Asymptotic Analysis of AdaBoost in the Binary Classification Case.- Least Absolute Shrinkage is Equivalent to Quadratic Penalization.- Oral Presentations: Applications: Applications I: Process Control, Diagnosis.- Neural Network Modelling of Ore Grade Spatial Variability.- Neural Virtual Sensors - Estimation of Combustion Quality in SI Engines Using the Spark Plug.- Real-Time Ion Temperature Profiles in the JET Nuclear Fusion Experiment.- Detection of Spatio-Temporal Ultrasonic Transient Families Using Filter Banks and Neural Nets.- Applications II: Image Processing.- Object Recognition with Multiple Feature Types.- Learning Gestures with Time-Delay RBF Networks.- Comparison of Adaptive Strategies for On-Line Character Recognition.- A Binary Correlation Matrix Memory k-NN Classifier.- Applications III: Speech, Telecom, Finance.- Multinet: A New Connectionist Architecture for Speech Recognition.- Simulation Support and ATM Performance Prediction.- TDNN Approach to Measuring Raindrop Sizes and Velocities.- Optimizing the Evidence.- Applications IV: Medicine, Sequence Analysis.- Adapting an Ensemble Approach for the Diagnosis of Breast Cancer.- Independent Component Analysis in Wave Decomposition of Auditory Evoked Fields.- Experiments in Gait Pattern Classification with Neural Networks of Adaptive Architecture.- Using Feature Selection to Find Inputs that Work Better as Extra Outputs.- Applications V: Hybrid Systems, General Techniques.- Analysis of Multi-Choice Questionnaires Through Self-Organizing Maps.- A Niching Genetic Algorithm for Selecting Features for Neural Network Classifiers.- The Wavelet Transform for Time Series Prediction.- Empirical Evaluation of Bayesian Sampling for Neural Classifiers.- Oral Presentations: Computational Neuroscience and Brain Theory.- State-Dependent Spatio-Temporal Restructuring of Receptive Fields in the Primary Visual Pathway.- Frequency Spectrum of Coupled Stochastic Neurons with Refractoriness.- Analysing the Context Dependence of Receptive Fields in Visual Cortex.- A Neural Model of Stereopsis.- Continuous Dynamics of Neuronal Delay Adaptation.- The Critical Synaptic Number for Rhythmogenesis and Synchronization in a Network Model of the Cerebellar Granular Layer.- A Model of Cortical Plasticity: Integration and Segregation Based on Temporal Input Patterns.- Self-Organisation and Cortical Dynamics.- Oral Presentations: Connectionist Cognitive Science and Artificial Intelligence.- Continuous Time Recurrent Neural Networks for Grammatical Induction.- Encoding Structure in Boolean Space.- Non-Compositional Representation in Connectionist Networks.- A Connectionist Model for Categorical Perception and Symbol Grounding.- Enriched Lexical Representations, Large Corpora, and the Performance of SRNs.- Oral Presentations: Autonomous Robotics and Adaptive Behavior.- Using Focus to Direct Environmental Mapping.- A Biologically Inspired Adaptive Control Architecture Based on Neural Networks for a Four-Legged Walking Machine.- 2-D Pole Balancing with Recurrent Evolutionary Networks.- GripSee: A Robot for Visually-Guided Grasping.- Dynamics of a Classical Conditioning Model.- CMAC Models Learn to Play Soccer.- Pseudo-Parametric Q-Learning Using Feedforward Neural Networks.- Continuous Q-Learning Resource Allocation Network.- Oral Presentations: Hardware and Implementations.- Sparse Distributed Memory Mapping on Partial Tree Shape Neurocomputer.- PRESENCE, a Hardware Implementation of Binary Neural Networks.- Analog VLSI Implementation of a Spike Driven Stochastic Dynamical Synapse.- Neural Dynamics in Real-Time for Large Scale Biomorphic Neural Networks.- Contents, Volume 2: Poster Presentations Poster Presentations: Theory Spotlight Presentations: Theory I: Algorithms.- A Study on Functional-Link Neural Units with Maximum Entropy Response.- Mean Field Theory Based on Belief Networks for Approximate Inference.- On the Use of Local RBF Networks to Approximate Multivalued Functions and Relations.- Learning Higher Order Boltzmann Machinces Using Linear Response.- Order Parameters for Self-Organizing Maps.- Slope Centering: Making Shortcut Weights Effective.- Spotlight Presentations: Theory II: Dynamical Systems, Time Series.- Constrained Second-Order Recurrent Networks for Finite-State Automata Induction.- Exact Learning Curves for EKF Training.- Spotlight Presentation: Theory IH: Signal Decomposition Methods.- Sparse Regression: Utilizing the Higher-Order Structure of Data for Prediction.- Poster Session I: Theory.- A Linear Programming Neural Circuit Model.- Learning Invariance Manifolds.- Developmental Evolution of an Edge Detecting Retina.- Activity Driven…