Geoscience Reference
In-Depth Information
13 Neurocomputing for
GeoComputation
Manfred M. Fischer and Robert J. Abrahart
CONTENTS
Abstract .......................................................................................................................................... 307
13.1 Introduction .......................................................................................................................... 308
13.2 What Is a Computational Neural Network? ......................................................................... 308
13.2.1 Pattern Classification ................................................................................................ 309
13.2.2 Clustering/Categorisation ......................................................................................... 309
13.2.3 Function Approximation ........................................................................................... 310
13.2.4 Prediction/Forecasting .............................................................................................. 310
13.2.5 Optimisation ............................................................................................................. 310
13.3 How Do Computational Neural Networks Work? ................................................................ 311
13.4 Characteristics of the Processing Elements .......................................................................... 312
13.5 Network Topologies .............................................................................................................. 314
13.6 Learning in a Computational Neural Network ..................................................................... 315
13.7 Classification of Computational Neural Networks ............................................................... 317
13.7.1 Backpropagation CNN.............................................................................................. 317
13.7.2 Radial Basis Function CNN ..................................................................................... 318
13.7.3 ART Network ........................................................................................................... 319
13.7.4 Self-Organising Feature Map ................................................................................... 320
13.8 Advantages, Application Domains and Examples................................................................ 321
13.8.1 Advantages ................................................................................................................ 321
13.8.2 Application Domains ................................................................................................ 322
13.8.3 Examples ................................................................................................................... 322
13.9 Conclusions and Outlook ...................................................................................................... 324
References ...................................................................................................................................... 324
ABSTRACT
This chapter provides an introduction to computational neural networks (CNNs), which are parallel
distributed information structures that can be used to carry out pattern classification, clustering,
function approximation and optimisation. An overview is presented of how CNNs function includ-
ing a description of the network processing elements (PEs) and the different network topologies
and how CNNs learn. A classification of CNNs into different types is then provided followed by a
discussion of the advantages of these tools and their application domains. The chapter concludes
with two examples to demonstrate their use in two diverse areas: one on using CNNs to model
interregional telecommunication traffic flows in Austria and the other on comparing three neural
classifiers of Landsat imagery for Vienna.
307
 
Search WWH ::




Custom Search