Tiwari / Knowles / Avineri Applications of Soft Computing

Recent Trends
1. Auflage 2010
ISBN: 978-3-540-36266-1
Verlag: Springer Berlin Heidelberg
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, Band 36, 420 Seiten

Reihe: Advances in Intelligent and Soft Computing

This book provides a comprehensive overview of recent advances in the industrial applications of soft computing. It covers a wide range of application areas, including optimisation, data analysis and data mining, computer graphics and vision, prediction and diagnosis, design, intelligent control, and traffic and transportation systems. The book is aimed at researchers and professional engineers engaged in developing and applying intelligent systems.

Autoren/Hrsg.

Tiwari, Ashutosh

Knowles, Joshua

Avineri, Erel

Dahal, Keshav

Roy, Rajkumar

Inhaltsverzeichnis

1;Title Page
;4
2;Copyright Page
;5
3;Preface;6
4;Message from WFSC Chair;7
5;Message from WSCIO General Chair and Technical Chair
;8
6;WSCIO Organization and Program Committee;10
7;Table of contents
;13
8;Part I Computer Graphics, Imaging and
;18
8.1;Efficient Genetic Algorithms for Arabic Handwritten Characters Recognition
;19
8.1.1;1 Introduction;19
8.1.2;2 Genetic Algorithms;21
8.1.3;3 Recognition System;22
8.1.3.1;3.1. Feature Extracted Using Genetic Algorithm (GA);23
8.1.3.2;3.2. Selection;24
8.1.3.3;3.3. Feature Selection Using Genetic Algorithm (GA);25
8.1.3.3.1;3.3.1 Introducing Variable Weights;26
8.1.4;4 Experimental Results;27
8.1.4.1;4.1 Feature Selection Using GA;28
8.1.5;5. Compare Genetic Algorithm Approach with Previous Works
;28
8.1.6;6 Conclusion;29
8.1.7;References;29
8.2;Generic Black Box Optimisation Algorithms for Colour Quantisation
;31
8.2.1;1 Introduction;31
8.2.2;2 Simulated Annealing;32
8.2.3;3 Simulated Annealing for Colour Quantisation;33
8.2.4;4 Experimental Results;34
8.2.5;5 Conclusions;37
8.2.6;References;37
8.3;Neural Network Combined with Fuzzy Logic to Remove Salt and Pepper Noise in Digital Images
;38
8.3.1;1 Introduction
;38
8.3.2;2 Neural Network Approach for Noisy Pixel Identification;39
8.3.3;3 Fuzzy Thresholding;41
8.3.4;4 Results of the Image Pre-Processing;45
8.3.5;5 Concluding Remarks;48
8.3.6;References;48
8.4;Computing Optimized NURBS Curves using Simulated Evolution on Control Parameters
;49
8.4.1;1 Introduction;49
8.4.2;2 Image Contour Extraction;50
8.4.3;3 Detection of Significant Points;50
8.4.4;4 NURBS;50
8.4.5;5 Outline of Simulated Evolution (Sim E)
;51
8.4.6;6 Proposed Approach;51
8.4.6.1;6.1 Problem Mapping;52
8.4.6.1.1;6.1.1 Initialization;52
8.4.6.1.2;6.1.2 Evaluation;52
8.4.6.1.3;6.1.3 Selection;53
8.4.6.1.4;6.1.4 Allocation and Weight Optimization;53
8.4.6.2;6.2 Algorithm Outline;54
8.4.7;8 Conclusion;57
8.4.8;Acknowledgments;57
8.4.9;References;58
9;Part II Control and Robotics
;59
9.1;Design of A Takagi-Sugeno Fuzzy Compensator for Inverted Pendulum Control Using Bode Plots
;60
9.1.1;1. Introduction;60
9.1.2;2. Mathematical Models of Inverted Pendulum;61
9.1.3;3. Local Compensator Design Using Bode Plots;62
9.1.4;4. Design of T-8 Fuzzy Compensator;64
9.1.5;5. Simulations;66
9.1.6;6. Conclusions;68
9.1.7;Acknowledgments;69
9.1.8;References;69
9.2;Soft Computing in Accuracy Enhancement of Machine Tools
;70
9.2.1;1 Introduction;70
9.2.2;2 Fuzzy Logic for Friction Compensation;71
9.2.3;3 Neural Networks-Based Error Modeling;73
9.2.4;4 Genetic Algorithms for Parameter Optimization;76
9.2.5;5 Conclusions;78
9.2.6;References;79
9.3;Mobile Robot Navigation: Potential Field Approach Vs. Genetic-Fuzzy System
;80
9.3.1;1 Introduction;80
9.3.2;2 Dynamic Motion Planning of a Car-Like Robot;82
9.3.2.1;2.1 Statement of the Problem;82
9.3.2.2;2.2 Motion Planning Scheme;82
9.3.3;3 Developed Motion Planning Algorithms;83
9.3.3.1;3.1 Approach 1: Potential Field Method;83
9.3.3.2;3.2 Approach 2: Genetic-Fuzzy System;84
9.3.4;4 Results and Discussion;85
9.3.5;5 Concluding Remarks;88
9.3.6;Acknowledgment;89
9.3.7;References;89
9.4;Intelligent Tuning and Application of a PID Controller Using Universal Model
;90
9.4.1;1 Introduction;90
9.4.2;2 Control Design;91
9.4.3;3 Tuning and Analysis;93
9.4.3.1;3.1 Fuzzy Approach;93
9.4.3.2;3.2 Neural Approach;94
9.4.3.3;3.3 Stability Analysis;95
9.4.4;4 Simulation Results;96
9.4.4.1;4.1 CSTR Application;96
9.4.5;5 Conclusion;98
9.4.6;References;99
10;Part III Design
;100
10.1;Dynamic Reconfiguration Algorithm for Field Programmable Analog Scalable Device Array (FPADA) with Fixed Topology
;101
10.1.1;1 Introduction;101
10.1.2;2 Reconfigurable Operational Amplifier Baseline;102
10.1.3;3 The Reconfiguration Algorithm;104
10.1.4;4 Experimental Results;107
10.1.5;5 Conclusion;109
10.1.6;References;110
10.2;Advances in the Application of Machine Learning Techniques in Drug Discovery, Design and Development
;111
10.2.1;1 Introduction;111
10.2.2;2 SVM Applications in Pharmaceuticals Research;112
10.2.2.1;2.1 SVM in Cheminformatics and Quantitative Structure-Activity Relationship (QSAR) Modelling.
;112
10.2.2.2;2.2 SVM in Bioinformatics;114
10.2.2.3;2.3 SVM in Clinical Diagnosis and Epidemiology;115
10.2.3;3 Drug Research Applications of Genetic Programming;115
10.2.3.1;3.1 GP in Cheminformatics and QSAR.;115
10.2.3.2;3.2 GP in Bioinformatics.;115
10.2.3.3;3.3 GP in Clinical Diagnosis and Epidemiology Research.;116
10.2.4;4 Biological Applications of Particle Swarm Optimisation;116
10.2.4.1;4.1 PSO in Cheminformatics and QSAR.;116
10.2.4.2;4.2 PSO in Bioinformatics.;117
10.2.4.3;4.3 PSO in Clinical Diagnosis and Epidemiology Research.;117
10.2.5;5 Discussion;117
10.2.6;Acknowledgements;118
10.2.7;References;118
10.3;A Review on Design Optimisation and Exploration with Interactive Evolutionary Computation
;123
10.3.1;1. Introduction;123
10.3.2;2. Single Objective IEC for Conceptual Design;124
10.3.2.1;2.1 Machine Design;125
10.3.2.2;2.2 Image Evolution;125
10.3.2.3;2.3 Virtual Modelling System;125
10.3.2.4;2.4 Aesthetic Design;126
10.3.3;3. Combining Multi-objective Optimisation with IEC for Industrial Design
;126
10.3.3.1;3.1 Micro-electrical Mechanical System Design;127
10.3.3.2;3.2 Interactive Evolutionary Design System;127
10.3.3.3;3.3 Interactive Multi-objective Optimisation Design Strategy;127
10.3.3.4;3.4 Interactive Multi-objective Design Optimisation;128
10.3.3.5;3.5 Multi-criteria Decision Making Strategy;128
10.3.3.6;3.6 Interactive Multi-objective Animation Design;129
10.3.4;4. lEe as the Generator of Artistic Design;129
10.3.5;5. lEe as a Design Tool: Trends and Issues;130
10.3.6;References;131
11;Part IV Pattern Recognition
;133
11.1;Mapping of Natural Patterns by Liquid Architectures Implementing Neural Cliques
;134
11.1.1;1 Introduction;134
11.1.2;2 Neural Microcircuit as a Generic Computational Unit;136
11.1.3;3 Computing with Neural Cliques;137
11.1.4;4 Word Recognition Task;139
11.1.5;5 Discussion;140
11.1.6;References;141
11.2;Pattern Recognition Using Modular Neural Networks and Fuzzy Integral as Method for Response Integration
;143
11.2.1;1. Introduction;143
11.2.2;2. Modular Neural Networks;144
11.2.2.1;2.1 Multiple Neural Networks;145
11.2.2.2;2.2 Main Architectures with Multiple Networks;146
11.2.3;3. Methods for Response Integration;146
11.2.4;4. Proposed Architecture and Results;148
11.2.4.1;4.1 Proposed Architecture;149
11.2.4.2;4.2 Description of the Integration Module;150
11.2.4.3;4.3 Summary of Results;150
11.2.5;5. Conclusions;152
11.2.6;References;152
11.3;Genetic Algorithm-Evolved Bayesian Network Classifier for Medical Applications
;153
11.3.1;1 Introduction;153
11.3.2;2 Signal Processing;154
11.3.3;3 Bayesian Networks;156
11.3.4;4 Method;158
11.3.5;5 Results;159
11.3.6;6 Discussion;160
11.3.7;7 Conclusion;161
11.3.8;Acknowledgments;162
11.3.9;References;162
11.4;A Real-Time Hand Gesture Interface for Medical Visualization Applications
;163
11.4.1;1 Introduction;163
11.4.2;2 System Overview;164
11.4.3;3 Segmentation;165
11.4.4;4 Feature Extraction and Pose Recognition;167
11.4.4.1;4.1 Hand Tracking and Pose Recognition;167
11.4.4.2;4.2 Maar Features;168
11.4.4.3;4.3 Pose Recognition;168
11.4.4.4;4.4 Optimal Feature and Recognition Parameter Selection;169
11.4.5;5 Test of the Hand Gesture FCM Classifier;170
11.4.6;6 Conclusions;171
11.4.7;Acknowledgments;171
11.4.8;References;172
12;Part V Classification
;173
12.1;A Hybrid Intelligent System and Its Application to Fault Detection and Diagnosis
;174
12.1.1;1 Introduction;174
12.1.2;2 The Hybrid SOM-kMER Model;176
12.1.3;3 The Hybrid SOM-kMER-PNN Model;176
12.1.4;4 Benchmark Data Sets;178
12.1.5;5 Fault Detection and Diagnosis;179
12.1.5.1;5.1 The Experiments;181
12.1.6;5 Conclusion;182
12.1.7;References;182
12.2;Evolutionary Multidimensional Scaling for Data Visualization and Classification
;185
12.2.1;1 Introduction;185
12.2.2;2 Multidimensional Scaling;186
12.2.2.1;2.1 Sammon's Nonlinear Mapping;186
12.2.2.2;2.2 Stochastic Search;187
12.2.3;3 Evolutionary Multidimensional Scaling;188
12.2.4;4 Experiments and Results;189
12.2.5;5 Conclusions;193
12.2.6;References;194
12.3;Extended Genetic Algorithm for Tuning a Multiple Classifier System
;195
12.3.1;1 Introduction;195
12.3.2;2 Basic Genetic Concepts;196
12.3.3;3 Extended Genetic Concepts;197
12.3.3.1;3.1 Coevolutionary Diversity;197
12.3.3.2;3.2 Advisable Behavior and Collective Fitness Evaluation;198
12.3.3.3;3.3 Phylogenetic Evolution v.s, Ontogenetic Evolution;199
12.3.3.4;3.4 Extended Genetic Algorithm;200
12.3.4;4 Results;201
12.3.5;5 Conclusions;202
12.3.6;References;203
12.4;Development of Fuzzy Expert System for Customer and Service Advisor Categorisation within Contact Centre Environment
;205
12.4.1;1. Introduction;205
12.4.2;2. Related Research;206
12.4.3;3. Proposed Methodology;207
12.4.3.1;3.1 Data Collection;207
12.4.3.2;3.2 Clustering Analysis - Identification of Customer and Advisor Categorisation
;208
12.4.3.3;3.3 Development of Fuzzy Expert System;209
12.4.4;4. Experimental Examples and Results;211
12.4.4.1;4.1 Customer Advisor (CSA) Experiments;211
12.4.4.2;4.2 Customer Experimental Results;212
12.4.4.3;4.3 Validation;212
12.4.5;5. Discussion and Future Research;212
12.4.6;6. Conclusions;213
12.4.7;Acknowledgements;213
12.4.8;References;214
12.5;Soft Computing for Intelligent Information Management
;215
12.5.1;Introduction;215
12.5.2;Integrating Query Responses;216
12.5.2.1;Operation scenario;216
12.5.2.2;Response Integrator Module;218
12.5.2.3;Output Results;220
12.5.3;Correspondence Between Meta-Data Hierarchies;221
12.5.3.1;Class Matching;221
12.5.3.2;Application to Film Databases;223
12.5.3.3;Instance and Genre Matching;223
12.5.3.4;Results on Unseen Data;224
12.5.4;Summary;225
12.5.5;References;225
12.6;Soft Computing in Intelligent Data Analysis;226
12.6.1;1 Introduction;226
12.6.2;2 Towards the Automation of Intelligent Data Analysis;227
12.6.2.1;2.1 Main Features of SPIDA;228
12.6.3;3 Data Analytics in Customer Relationship Management;229
12.6.4;4 Data Analytics in Resource Management;232
12.6.5;5 Conclusions;235
12.6.6;References;235
12.7;Neural Network-Based Expert System toPredict the Results of Finite Element Analysis;236
12.7.1;1 Introduction;236
12.7.2;2 Mathematical Formulation of the Problem and Its Solution Using an FE Analysis
;238
12.7.2.1;2.1 Results of ANSYS 7.0 Package;238
12.7.2.2;2.2 Basics of FEM Formulation;239
12.7.2.3;2.3 Results of FE Analysis Using ANSYS 7.0 Package;239
12.7.3;3 Proposed Techniques to Develop NN-based ES;240
12.7.3.1;3.1 Approach 1: NN-based ES Using Back-propagation Learning Algorithm
;240
12.7.3.2;3.2 Approach 2: NN-based ES Using GA-based Learning Algorithm
;240
12.7.4;4 Results and Discussion;241
12.7.5;5 Concluding Remarks;244
12.7.6;6 Scope for Future Work;245
12.7.7;References;245
13;Part VI Identification and Forecasting
;246
13.1;Modular Neural Networks with Fuzzy Integration Applied to Time Series Prediction
;247
13.1.1;1. Introduction;247
13.1.2;2. Monolithic Neural Network Models;248
13.1.3;3. Modular Neural Networks;250
13.1.4;4. Methods for Response Integration;251
13.1.5;5. Simulation and Forecasting Prices in the U.S. Market;253
13.1.6;6. Experimental Results;253
13.1.7;7. Conclusions;255
13.1.8;References;255
13.2;Fuzzy Model Identification for Rapid Nickel-Cadmium Battery Charger through Particle Swarm Optimization Algorithm
;257
13.2.1;1 Introduction;257
13.2.2;2 Fuzzy Model Identification Problem;258
13.2.3;3 Particle Swarm Optimization (PSO) Algorithm;259
13.2.4;4 Rapid Ni-Cd Battery Charger;260
13.2.5;5 Framework for Fuzzy Model Identification with PSO Algorithm
;261
13.2.6;6 Simulation Results;263
13.2.7;7 Conclusions and Further Scope;264
13.2.8;References;265
13.3;Fuzzy Association Rule Mining for Model Structure Identification
;267
13.3.1;1 Introduction;267
13.3.2;2 Fuzzy Association Rule Mining;268
13.3.2.1;2.1 Counting the Fuzzy Support;269
13.3.2.2;2.2 Mining Frequent Item Sets;270
13.3.2.3;2.3 Generation of Fuzzy Association Rules;271
13.3.3;3 MOSSFARM - Model Structure Selection by Fuzzy Association Rule Mining
;272
13.3.3.1;3.1 Generate a Fuzzy Data Set;272
13.3.3.2;3.2 Pruning of the Rule Base;273
13.3.3.3;3.3 Selection of the Relevant Input Variables;273
13.3.4;4 Application Studies;274
13.3.4.1;4.1 Mixed Continuous and Discrete Data;274
13.3.4.2;4.2 Continuous Polymerization Reactor;275
13.3.5;5 Conclusions;275
13.3.6;References;276
13.4;Modeling Public Transport Trips with General Regression Neural Networks; A Case Study for Istanbul Metropolitan Area
;277
13.4.1;1. Introduction;277
13.4.2;2. General Regression Neural Network Method;278
13.4.3;3. Analysis of Data;280
13.4.4;4. ANN Preparation and Prediction Results;281
13.4.4.1;4.1 Preparation of GRNN Configuration;281
13.4.4.2;4.2 GRNN Prediction Results;282
13.4.5;5. Conclusions;284
13.4.6;References;285
14;Part VII Optimization Problems: Assignment, Partitioning and Ordering
;287
14.1;Use of Genetic Algorithm to Optimum Test Frequency Selection
;288
14.1.1;1 Introduction;288
14.1.2;2 Method Description;289
14.1.3;3 Computational Examples;290
14.1.4;4 Conclusions;294
14.1.5;References;294
14.2;Performance Analysis of Parallel Strategies for Bi-objective Network Partitioning
;296
14.2.1;1 Introduction;296
14.2.2;2 Graph Partitioning as Tool to Partitioning Networks;297
14.2.3;3 Sequential Algorithm: Pareto Simulated Annealing;298
14.2.4;4 Parallelization of Pareto Simulated Annealing;298
14.2.4.1;4.1 Master-Slave Parallelization (MS);299
14.2.4.2;4.2 Island Parallelization (I);299
14.2.4.3;4.3 Island with Search Space Division Parallelization (Issd);299
14.2.5;5 Experimental Results;300
14.2.5.1;5.1 Parameter Settings;300
14.2.5.2;5.2 Performance Measures;301
14.2.5.3;5.3 Empirical Results;301
14.2.6;6 Conclusions;303
14.2.7;Acknowledges;304
14.2.8;References;304
14.3;A Guided Rule Reduction System for Prioritization of Failures in Fuzzy FMEA
;306
14.3.1;1 Introduction;306
14.3.2;2 Failure Risk Prioritization Issues of Traditional FMEA;307
14.3.3;3 Modeling of the Fuzzy RPN;307
14.3.3.1;3.1 Fuzzy Membership Functions;307
14.3.4;4. Rule Reduction of the Fuzzy RPN Model;308
14.3.5;5 Algorithm of the Guided Rules Reduction System;309
14.3.6;6 Case Studies and Experiments;311
14.3.6.1;6.1 Background;312
14.3.6.1.1;6.1.1 The Wafer Mounting Process;312
14.3.6.1.2;6.1.2 The Underfill Dispensing process;312
14.3.6.2;6.2 Experiment I: The Traditional and Fuzzy RPN Models;312
14.3.6.3;6.3 Experiment II-The Fuzzy RPN Model with and Without the GRRS.;314
14.3.7;7 Conclusions;314
14.3.8;References;315
14.4;A Hybrid Method of Differential Evolution and SQP for Solving the Economic Dispatch Problem with Valve-Point Effect
;316
14.4.1;1. Introduction;316
14.4.2;2. Description of Economic Dispatch Problem;317
14.4.3;3. Optimization Methods of Economic Dispatch Problem;318
14.4.3.1;3.1 Differential Evolution;319
14.4.3.2;3.2 Sequential Quadratic Programming (SQP);321
14.4.3.3;3.3 Hybrid Approach of DE with SQP;321
14.4.4;4. Case Study of 40 Thermal Units;321
14.4.5;5. Conclusion and Future Research;324
14.4.6;References;325
14.5;Multiobjective Prioritization in the Analytic Hierarchy Process by Evolutionary Computing
;326
14.5.1;1 Introduction;326
14.5.2;2 Multiobjective Prioritization Problem;327
14.5.2.1;2.1 Optimization Criteria;327
14.5.2.2;2.2 Statement of the Prioritization Problem;329
14.5.3;3 Solving the TOP Problem by Evolutionary Computing;330
14.5.4;4 An Illustrative Example;332
14.5.5;5 Conclusions;334
14.5.6;Acknowledgement;335
14.5.7;References;335
14.6;Evolutionary and Heuristic Algorithms for Multiobjective 0-1 Knapsack Problem
;336
14.6.1;1 Introduction;336
14.6.2;2 Basic Definitions and Problem Formulation;337
14.6.3;3 Heuristic Algorithms;339
14.6.3.1;3.1 Profit-Weight Ratio Heuristic;339
14.6.3.2;3.2 Solutions by Dynamic Programming;340
14.6.4;4 Solution by Evolutionary Algorithms;340
14.6.5;5 Discussion & Conclusions;343
14.6.6;Acknowledgement;344
14.6.7;References;344
14.7;The Assignment of Referees to WSC10 Submissions: An Evolutionary Approach
;346
14.7.1;1 Introduction;346
14.7.1.1;1.1 Related Work;347
14.7.2;2 Problem and Method;348
14.7.2.1;2.1 The Referees;348
14.7.2.2;2.2 The Manuscripts;348
14.7.2.3;2.3 Creating a Better Keyword List;348
14.7.2.4;2.4 Early Algorithm Design Choices: Encoding, Operators and General Approach
;349
14.7.2.5;2.5 Dummy Runs and Visualization of Results;350
14.7.2.6;2.6 Rewarding Good Matches;351
14.7.2.7;2.7 Penalizing for 'Redundant' Referees;352
14.7.2.8;2.8 Penalizing for Over-use of a Referee;353
14.7.2.9;2.9 Putting it all Together;353
14.7.3;3 Results;353
14.7.4;4 Summary and Discussion;354
14.7.5;Acknowledgments;355
14.7.6;References;355
15;Part VIII Optimization Methods: Development and Analysis
;356
15.1;Robustness using Multi-Objective Evolutionary Algorithms
;357
15.1.1;1 Introduction;357
15.1.2;2 Measures of Robustness;358
15.1.2.1;2.1 Expectation Measures;359
15.1.2.2;2.2 Variance Measures;360
15.1.3;3 Extending Robustness Measures to Multiple Objectives;360
15.1.4;4 Results and Discussion;361
15.1.4.1;4.1 Test Problems;361
15.1.4.2;4.2 Expectation Results;362
15.1.4.3;4.3 Variance Results;362
15.1.5;5 Conclusions;363
15.1.6;Acknowledgements;365
15.1.7;References;365
15.2;Genetic Programming, Probabilistic Incremental Program Evolution, and Scalability
;367
15.2.1;1 Introduction;367
15.2.2;2 Methods;368
15.2.2.1;2.1 Genetic Programming;368
15.2.2.2;2.2 PIPE;369
15.2.3;3 Test Problems;369
15.2.3.1;3.1 Problem 1: Order;370
15.2.3.2;3.2 Problem 2: Deceptive Trap;370
15.2.3.3;3.3 Other Primitives;371
15.2.4;4 Experiments;372
15.2.4.1;4.1 Description of Experiments;372
15.2.4.2;4.2 Results;372
15.2.5;5 Summary;374
15.2.6;6 Conclusions and Future Work;375
15.2.7;Acknowledgments;375
15.2.8;References;376
15.3;Adaptive Parameter Control of Evolutionary Algorithms Under Time Constraints
;377
15.3.1;1 Introduction;377
15.3.2;2 Problem Formulation;379
15.3.3;3 Static Selection of Control Parameters;380
15.3.4;4 Adaptive Parameter Controlling;380
15.3.5;5 Results;383
15.3.6;6 Conclusions and Future Work;386
15.3.7;References;386
15.4;Role of Chaos in Swarm Intelligence - A Preliminary Analysis
;387
15.4.1;1 Introduction;387
15.4.2;2 Particle Swarm Model;388
15.4.3;3 Iterated Function System and its Sensitivity;388
15.4.4;4 Dynamic Chaotic Characteristics;390
15.4.4.1;4.1 Lyapunov Exponent;390
15.4.4.2;4.2 Correlation Dimension;391
15.4.4.3;4.3 Discussions;392
15.4.5;5 Conclusions and Future Work;395
15.4.6;Acknowledgments;396
15.4.7;References;396
15.5;Multi-parent Recombination Operator with Multiple Probability Distribution for Real Coded Genetic Algorithm
;397
15.5.1;1. Introduction;397
15.5.2;2. Real-Parameter Crossover Operators;399
15.5.2.1;2.1 Exploration and Exploitation;399
15.5.2.2;2.2 Multi-parent Polynomial Distribution Crossover Operator (MPX) and Multi-parent Lognormal Distribution Crossover Operator (MLX)
;399
15.5.3;3. Experimental Setup;400
15.5.4;4. Multi-parent Crossover Operators;401
15.5.4.1;4.1 Multi-parent Multiple Probability Distribution Operator (MMX);401
15.5.4.2;4.2 Multi-parent Polynomial and Lognormal Distribution Based Recombination Operator (MPLX)
;402
15.5.4.3;4.3 Multi-parent Hybrid Recombination Operator (MHX);403
15.5.5;5. Discussion;404
15.5.6;6. Conclusion;405
15.5.7;References;406
16;Part IX Tutorials
;407
16.1;Special Tutorial - State of the Art Face Detection: CascadeBoosting Approaches
;408
16.1.1;1. Introduction;408
16.1.2;2. Problem Definition;408
16.1.3;3. Historical Development and State of the Art;409
16.1.4;4. Boosted Cascades for Face Detection;410
16.1.4.1;4.1 Boosted Classifiers;410
16.1.4.2;4.2 Weak Classifiers;411
16.1.4.3;4.3 Training Boosted Cascades;412
16.1.5;Acknowledgements;412
16.1.6;References;412
16.2;Special Tutorial - Particle Swarms for Fuzzy Models Identification
;413
16.2.1;References;414
16.3;Special Tutorial - Project Management: Issues in Computer Based Monitoring and Control with Soft Computing Approaches
;416
16.3.1;1 Introduction;416
16.3.2;2 Selection of Optimal Pert Chart;417
16.3.3;3 Issues in Comparing PERT Charts;417
16.3.3.1;3.1 Neural Network Based Selection of PERT Charts;418
16.3.3.2;3.2 Fuzzy Based Selection of Pert Charts;419
16.3.4;4. Conclusions;420
16.3.5;References;421

Leseproben

"Evolutionary Multidimensional Scaling for Data Visualization and Classification (p. 177-178)

Summary. Multidimensional Scaling (MDS) is a well established technique for the projection of high-dimensional data in pattern recognition, data visualization and analysis, as well as scientific and industrial applications. In particular, Sammons Nonlinear Mapping (NLM) as a common MDS instance, computes distance preserving mapping based on gradient descent, which depends on initialization and just can reach the nearest local optimum. Improvement of mapping quality or reduction of mapping error is aspired and can be achieved by more powerful optimization techniques, e.g., stochastic search, successfully applied in our prior work. In this paper, evolutionary optimization adapted to the particular problem and the NLM has been investigated for the same aim, showing the feasibility of the approach and delivering competitive and encouraging results.

1 Introduction

Multidimensional Scaling (MDS) is a common dimensional reduction method and widely used in exploratory data analysis and design of pattern recognition and integrated sensor systems for extraction of essential information from multivariate data sets [1]. Extraction of essential information by dimensionality reduction is required for various reasons. For instance, it can avoid the curse of dimensionality and thus improve the ability of classification. The analysis of unknown data by projection and ensuing interactive visualization is another common and important application of dimensionality reduction [2, 3].

MDS represents data in a smaller number of dimensions preserving the similarity structure of the data as much as possible. One common MDS instance is Sammons nonlinear mapping (NLM) that will be emphasized and discussed in this paper. In this method, a criterion denoted stress (error) function is defined and optimization takes place by gradient descent techniques [4]. These optimizations have known drawbacks, which strongly depend on the initialization, can get more easily trapped in a less fortunate local minimum and saturate on an undesirable high error value.

Implications of this behavior can be data visualizations of low reliability, which incorporate twists and misleading neighborhoods. To overcome this problem, there are some methods that have been proposed, e.g., simulated annealing [5] or using neural network [6]. In similar cases and applications, stochastic search optimization in our prior work has been successfully applied to overcome such a problem [7]. In this paper, we investigated another stochastic method, that draws inspiration from the process of natural evolution, i.e., evolutionary computation. In several research activities, MDS has been applied to extract useful information and visualize the progress of evolutionary optimization [8].

Additionally, some recent publications have introduced the application of genetic algorithm for optimizing MDS itself [9, 10]. Similar to those, we have applied evolutionary optimization. However, we adopted from our work on stochastic search [7] the concept of the mutation operator, rather than employing gradient factor, since this approach pursued in [9] takes significant computational effort. Regarding to the goals of mapping improvement, i.e., the reliability (error reduction) and speed computation, we examine the feasibility and assessment of applicability or competitiveness of such approaches adapted to MDS."