GeoComputational Modelling — Techniques and Applications: Prologue

Abstract

1 GeoComputational Modelling - Techniques and Applications: Prologue Man/red M Fischer and Yee Leung PART A: Concepts, Modelling Tools and Key Issues 2 Computational Neural Networks - Tools for Spatial Data Analysis Manfred M Fischer 2.1 Introduction 2.2 Why Computational Neural Networks? 2.3 Definition of a Computational Neural Network 2.4 Properties ofthe Processing Elements 2.5 Network Topologies 2.6 Learning in a Computational Neural Network 2.7 A Taxonomy of Computational Neural Networks 2.8 Outlook - How Do Neurocomputing Techniques Differ? 3 Evolving Computational Neural Networks Through Evolutionary Computation Xin Yao 3.1 Introduction 3.2 Evolving Computational Neural Network Architectures 3.3 EPNet 3.4 Experimental Studies 3.5 Evolutionary Leaming and Optimization 3.6 A Population ofECNNs as an Ensemble 3.7 Conclusions 15 17 20 21 24 27 29 34 35 37 40 48 60 61 69 x Contents 4 Neural and Evolutionary Computation Methods ror Spatial Classifieation and Knowledge Aequisition Yee Leung 4.1 Introduction 4.2 Spatial Classification by Multilayer Feedforward Neural Networks 4.3 Spatial Classification by Other Unidirectional Neural Networks 4.4 Spatial Classification by Recurrent Neural Networks 4.5 Clustering by Scale-Space Algorithms 4.6 Rule Learning by a Radial Basis Function Neural Network 4.7 Rule Learning by a Hybrid Fuzzy Neural Network 4.8 Rule Acquisition by Genetic Algorithms - The SCION System 4.9 Fuzzy Rule Acquisition by Genetic Algorithms - The GANGO System 4.10 Conclusions 5 Cellular Dynamies: Modelling Urban Growth as a Spatial Epidemie Michael Batly 5.1 Defining Urban Growth as Sprawl 5.2 Growth as an Epidemie: Spatially Aggregate Models 5.3 Simplifications and Extensions to the Aggregate Model 5.4 Growth as Spatial Diffusion: Spatially Disaggregate Models 5.5 A Computable Structure Based on Cellular Automata 5.6 The Dynamics ofUrban Regeneration 5.7 Classifying Urban Growth through Morphology 5.8 Conclusions: Applieations and Policy PART B: Spatial Application Domains 6 Spatial Pattern Reeognition in Remote Sensing by Neural Networks Graeme Wilkinson 6.1 Introduction 6.2 Artificial and Biological Neural Networks 6.3 Recent Developments in Remote Sensing 6.4 Uses ofNeural Networks in Remote Sensing 6.5 Creation ofNeural Network Input Veetors 71 73 78 80 80 83 89 94 100 107 109 112 116 122 125 130 134 139 145 146 147 148 150 7 8 Contents xi 6.6 6.7 6.8 6.9 6.10 6.11 Neural Networks in Unsupervised Classification ofRemote Sensing Data Neural Networks in Supervised Classification ofRemote Sensing Data 'Soft Computing' Approaches Using Neural Networks Managing Complexity Hybrid Analysis Methodologies Conclusions Fuzzy ARTMAP - A Neural Classifier for Multispectral Image Classification Sucharita Gopal and Man/red M Fischer 7.1 Introduction 7.2 Adaptive Resonance Theory and ART 1 7.3 The ARTMAP Neural Network Architecture 7.4 Generalization to Fuzzy ARTMAP 7.5 The Spectral Pattern Recognition Problem 7.6 Fuzzy ARTMAP Simulations and Classification Results 7.7 Summary and Conclusions Neural Spatial Interaction Models Man/red M Fischer 8.1 Introduction 8.2 The Model Class under Consideration 8.3 Training Neural Spatial Interaction Models: Classical Techniques 8.4 A New Global Search Approach for Network Training: The Differential Evolution Model 8.5 Selecting Neural Spatial Interaction Models: The Model Choice Issue 8.6 Evaluating the Generalization Performance of a Neural Spatial Interaction Model 8.7 Conclusion and Outlook 150 154 157 159 162 164 165 166 173 177 180 181 188 195 196 200 205 208 214 218 xii Contents 9 A Neural Network Approach for Mobility Panel Analysis Günter Haag 9.1 Introduction 220 9.2 The Gennan Mobility Panel 221 9.3 Classical Panel Analysis 223 9.4 Application ofComputational Neural Networks to the Gennan 223 Mobility Panel 9.5 Analysis ofthe Variable LOG[DAU_SUM] 228 9.6 Analysis ofthe Variable NUTZPKW 232 9.7 Conclusions and Outlook 234