Applying Computational Intelligence (eBook)

How to Create Value

Arthur Kordon (Autor)

eBook Download: PDF

2009 | 2010
XXII, 459 Seiten
Springer Berlin (Verlag)
978-3-540-69913-2 (ISBN)

Infos

In theory, there is no difference between theory and practice. But, in practice, there is. Jan L. A. van de Snepscheut The ?ow of academic ideas in the area of computational intelligence has penetrated industry with tremendous speed and persistence. Thousands of applications have proved the practical potential of fuzzy logic, neural networks, evolutionary com- tation, swarm intelligence, and intelligent agents even before their theoretical foundation is completely understood. And the popularity is rising. Some software vendors have pronounced the new machine learning gold rush to 'Transfer Data into Gold'. New buzzwords like 'data mining', 'genetic algorithms', and 'swarm optimization' have enriched the top executives' vocabulary to make them look more 'visionary' for the 21st century. The phrase 'fuzzy math' became political jargon after being used by US President George W. Bush in one of the election debates in the campaign in 2000. Even process operators are discussing the perf- mance of neural networks with the same passion as the performance of the Dallas Cowboys. However, for most of the engineers and scientists introducing computational intelligence technologies into practice, looking at the growing number of new approaches, and understanding their theoretical principles and potential for value creation becomes a more and more dif?cult task.

Arthur K. Kordon is a Data Mining and Modeling Leader in the Data Mining and Modeling Capability of The Dow Chemical Company. He is an internationally recognized expert in applying emerging technologies in industry, and has given talks and chaired panels on the topic at the major computational intelligence conferences such as WCCI and GECCO. He has successfully introduced several novel technologies for improved manufacturing and new product design in the chemical industry, and his research interests include application issues of computational intelligence, robust empirical modeling, intelligent process monitoring and control, and data mining.

Inhalt

FM 2
Outline placeholder 1
Motivation 7
Purpose of the Book 8
Who Is This Book for? 10
How This Book Is Structured 12
What This Book Is NOT About 13
Features of the Book 14
p1 21
Part I: Computational Intelligence in a Nutshell 21
144233_1_En_1_Chapter_OnlinePDF 22
Chapter 1: Artificial vs. Computational Intelligence 22
Artificial Intelligence: The Pioneer 23
Practical Definition of Applied Artificial Intelligence 23
Key Practical Artificial Intelligence Approaches 24
ExpertExpert Systems (RuleRule-Based and Frame-Based) 25
RuleRule-Based ExpertExpert Systems 25
Frame-Based ExpertExpert Systems 26
Inference Mechanisms 26
Backward Chaining 27
Forward Chaining 27
Case-Based ReasoningReasoning 27
Knowledge Management 28
Applied Artificial Intelligence Success Stories 30
Integrated SoftwareSoftware InfrastructureInfrastructure for Applied AIApplied AI 30
Technical Advantages of Applied AIApplied AI 30
Domain Expertise Is Captured 31
Knowledge Is Presented in Natural Language 31
RuleRule Structure Is Uniform 31
Interpretive Capability 31
Separation of Knowledge from Inference Mechanisms 32
Application Areas of AI 32
Advisory SystemsAdvisory systems 32
Decision-MakingDecision-making 33
PlanningPlanning 33
Selection 33
Diagnostics 33
Preserving Knowledge 34
Examples of Successful AI Real-World Applications 34
Applied Artificial Intelligence Issues 36
Technical Issues of Applied AI 36
Knowledge ConsistencyKnowledge consistency 36
Scale-upScale-up 36
Static Nature, No LearningLearning 36
Subjective Nature of Representing Intelligence 37
InfrastructureInfrastructure Issues of Applied AI 37
Limited Computer Capabilities in the Early 1980s 37
High Total Cost of Ownership 37
People Issues of Applied AI 38
Knowledge Extraction 38
Incompetence DistributionIncompetence distribution 38
Legal Issues 39
Artificial Intelligence Application Lessons 39
Application AI Lesson 1: Do not create unrealistic expectationsUnrealistic expectations 39
Application AI Lesson 2: Do not push new technologies by campaigns 39
Application AI Lesson 3: Do not underestimate maintenance and support 40
Application AI Lesson 4: Clarify and demonstrate value as soon as possible 40
Application AI Lesson 5: Develop a strategy for sustainable application growth 40
Application AI Lesson 6: Link the success of the application with incentives to all stakeholders 41
Computational Intelligence: The Successor 41
Practical Definition of Applied Computational Intelligence 42
Key Computational Intelligence Approaches 43
Fuzzy SystemsFuzzy Systems 44
Neural NetworkNeural networks 44
Support Vector Machines 44
Evolutionary ComputationEvolutionary computation 45
Swarm Intelligence 45
Intelligent Agents 45
Key Differences Between AI and CI 46
Key Technical Differences 46
Key Difference #1 - On the main source of representing intelligence 46
Key Difference #2 - On the mechanisms of processing intelligence 46
Key Difference #3 - On the interactions with the environment 47
The Ultimate Difference 47
An Integrated View 48
Summary 48
The Bottom Line 49
Suggested Reading 49
144233_1_En_2_Chapter_OnlinePDF 50
Chapter 2: A Roadmap Through the Computational Intelligence Maze 50
Strengths and Weaknesses of CI Approaches 50
Strengths and Weaknesses of Fuzzy Systems 51
Strengths and Weaknesses of Neural Networks 53
Strengths and Weaknesses of Support Vector Machines 54
Strengths and Weaknesses of Evolutionary Computation 56
Strengths and Weaknesses of Swarm IntelligenceSwarm Intelligence 58
Strengths and Weaknesses of Intelligent Agents 59
Key Scientific Principles of Computational Intelligence 61
Bio-inspired Computing 61
LearningLearning Systems 63
Computer Science 64
Key Application Areas of Computational Intelligence 64
InventionInvention of New Products 64
Systems Design 66
ManufacturingManufacturing 66
Supply Chain 66
Market Analysis 67
Financial Modeling 67
Modeling Social Behavior 68
Health 68
LeisureLeisure 68
Summary 69
The Bottom Line 69
Suggested Reading 69
144233_1_En_3_Chapter_OnlinePDF 70
Chapter 3: Let's Get Fuzzy 70
Fuzzy Systems in a Nutshell 70
Dealing With AmbiguityAmbiguity 71
Fuzzy Sets 73
Fuzzy Systems Created By Experts 74
Fuzzy Systems Created by DataData 77
Benefits of Fuzzy Systems 79
Fuzzy Systems Issues 81
How to Apply Fuzzy Systems 82
When Do We Need Fuzzy Systems? 82
Applying ExpertExpert-Based Fuzzy Systems 82
Applying DataData-Based Fuzzy Systems 83
Typical Applications of Fuzzy Systems 84
Fuzzy Systems MarketingMarketing 87
Available Resources for Fuzzy Systems 89
Key Websites 89
Selected SoftwareSoftware 90
Summary 90
The Bottom Line 90
Suggested Reading 91
144233_1_En_4_Chapter_OnlinePDF 92
Chapter 4: Machine LearningLearning: The Ghost in the Learning Machine 92
Neural Networks in a Nutshell 95
Biological Neurons and Neural Networks 95
Artificial Neurons and Neural Networks 96
Back-propagation 99
Neural Network Structures 101
Support Vector Machines in a Nutshell 103
Statistical LearningLearning Theory 104
Structural Risk Minimization 105
Support Vector Machines for ClassificationClassification 106
Support Vector Machines for RegressionRegression 109
Benefits of Machine LearningLearning 110
Comparison Between Neural Networks and SVM 110
Key Difference #1 - On the method´s basis 111
Key Difference #2 - On the necessary dataData for model developmentModel 111
Key Difference #3 - On the optimization typeOptimization 111
Key Difference #4 - On the generalization capability 112
Key Difference #5 - On the number of modelModel outputs 112
Benefits of Neural Networks 112
Benefits of Support Vector Machines 113
Machine LearningLearning Issues 115
Neural Networks Issues 115
Support Vector Machines Issues 115
How to Apply Machine LearningLearning Systems 116
When Do We Need Machine LearningLearning? 116
Applying Machine LearningLearning Systems 117
Applying Neural Networks: An Example 119
Applying Support Vector Machines: An Example 122
Typical Machine LearningLearning Applications 124
Typical Applications of Neural Networks 124
Typical Applications of Support Vector Machines 127
Machine LearningLearning MarketingMarketing 127
Neural Networks MarketingMarketing 127
Support Vector Machines MarketingMarketing 129
Available Resources for Machine LearningLearning 130
Key Websites 130
Key SoftwareSoftware 130
Summary 131
Suggested Reading 132
144233_1_En_5_Chapter_OnlinePDF 133
144233_1_En_6_Chapter_OnlinePDF 163
144233_1_En_7_Chapter_OnlinePDF 193
Chapter 7: Intelligent Agents: The Computer Intelligence Agency (CIA) 193
Intelligent Agents in a Nutshell 194
Complex Systems 195
Intelligent Agents 197
Agent-Based Integrators 199
Agent-Based Systems 201
Benefits of Intelligent Agents 204
Comparison Between Agents and Objects 204
Comparison Between Agents and ExpertExpert Systems 205
Benefits of Agent-Based Modeling 205
Intelligent Agents Issues 207
How to Apply Intelligent Agents 208
When Do We Need Intelligent Agents? 208
Applying Intelligent Agents 209
Typical Applications of Intelligent Agents 211
Intelligent Agents MarketingMarketing 214
Available Resources for Intelligent Agents 217
Key Websites 217
Selected SoftwareSoftware 217
Summary 217
Suggested Reading 218
p2 219
Part II: Computational Intelligence Creates Value 219
144233_1_En_8_Chapter_OnlinePDF 220
Chapter 8: Why We Need Intelligent Solutions 220
Beat CompetitionCompetition 221
Effective Utilization of Emerging Technologies 221
Fast Response to Changing Environment 222
Effective Operation in the Global Economy 222
Flexible Strategy 223
Low Cost of Operation 224
Accelerate Innovations 224
Business Impact Analysis of InnovationInnovation 225
Automatic Novelty GenerationGeneration 225
Rapid ExplorationExploration of New Ideas 226
Fast CommercializationCommercialization in Practice 226
Produce Efficiently 227
Accurate Production PlanningPlanning 227
Enhanced ObservabilityObservability of Processes 228
Broad Product and Process OptimizationOptimization 228
Advanced Process Control 229
Improved Operating Discipline 229
Distribute Effectively 229
Estimate Demand 230
Handle Global Market ComplexityComplexity 230
Real-Time Operation 231
Optimal SchedulingScheduling 231
Impress Customers 232
Analyze Customers 232
Deliver Simple Solutions 233
Create a Visionary Image 233
Broaden Customer Base 234
Enhance CreativityCreativity 234
Reduce Routine Operations Related to Intelligence 235
Magnify Imagination 235
Add Intellectual Sensors 236
Increase Cognitive Productivity 236
Attract Investors 237
Intelligence and Growth 237
High-Tech Magnetism 238
Technology Credibility 238
Technology SustainabilitySustainability 239
Improve National Defense 239
Intelligent Intelligence 240
RobotRobot Soldiers 240
Smart Weapons 241
Cyber Wars 241
Protect Health 242
Medical Diagnosis 242
Personal Health Modeling 243
Health Monitoring 244
Personal Health Advisor 244
Have Fun 245
Intelligent GamesGames 245
Funny Education 246
Smart Toys 246
Evolutionary Art 247
Virtual HollywoodVirtual Hollywood 247
Summary 248
The Bottom Line 248
Suggested Reading 248
144233_1_En_9_Chapter_OnlinePDF 249
Chapter 9: Competitive Advantages of Computational Intelligence 249
Competitive Advantage of a Research Approach 249
Step 1: Clarify Technical Superiority 250
Step 2: Demonstrate Low Total Cost of Ownership 251
Step 3: Apply in Areas with High Impact 252
Key Competitive Approaches to Computational Intelligence 253
Competitor #1: First-Principles Modeling 254
Competitor #2: Statistical Modeling 257
Competitor #3: HeuristicsHeuristics 259
Competitor #4: Classical OptimizationOptimization 261
How Computational Intelligence Beats the CompetitionCompetition 263
Creating ``Objective Intelligence´´ 263
Dealing with UncertaintyUncertainty 265
Dealing with ComplexityComplexity 267
Generating Novelty 268
Low-Cost Modeling and OptimizationOptimization 270
Summary 271
Suggested Reading 272
144233_1_En_10_Chapter_OnlinePDF 273
Chapter 10: Issues in Applying Computational Intelligence 273
Technology Risks 273
The Change Function 274
Technocentric Culture 275
Increased ComplexityComplexity 276
Technology Hype 276
Modeling Fatigue 277
The Invasion of First-Principles Models 277
Statistical Models Everywhere 278
How to Lie with AI 278
Anything but ModelModel (ABM) Movement 278
Looks Too Academic 279
Difficult to Understand 279
Diverse Approaches 280
Difficult to Track 280
It's Not Yet Ready for IndustryIndustry 281
Perception of High Cost 281
Growing R& D Cost
Expensive InfrastructureInfrastructure 282
Expected Training Cost 282
Anticipated Maintenance Nightmare 282
Missing InfrastructureInfrastructure 283
Specialized Hardware 283
Limited SoftwareSoftware 284
Unclear Organization Structure 284
Work Process Not Defined 285
No MarketingMarketing 285
Product Not Clearly Defined 285
Unclear Competitive Advantages 286
Key Markets Not Identified 286
No Advertisement 287
Wrong Expectations 287
Magic Bullet 287
GIGO 2.0 288
SkepticismSkepticism 288
ResistanceResistance 289
No Application Methodology 290
Method Selection 290
Integration Advantages 291
Application Sequence 291
Few References 291
Summary 291
Suggested Reading 292
p3 293
Part III: Computational Intelligence Application Strategy 293
144233_1_En_11_Chapter_OnlinePDF 294
Chapter 11: Integrate and Conquer 294
11.1The Nasty Reality of Real-World Applications 295
11.2Requirements for Successful Real-World Applications 297
11.3Why Integration Is Critical for Real-World Applications 299
11.3.1Benefits of Integration 300
11.3.2The Price of Integration 301
11.4Integration Opportunities 302
11.4.1Hybrid Intelligent Systems 302
11.4.2Integration with First-Principles Models 306
11.4.3Integration with Statistical Models 308
11.5Integrated Methodology for Robust Empirical Modeling 309
11.5.1Integrated Methodology for Undesigned DataDataThe initial version of the methodology is published in: A. Kordon, G 310
11.5.1.1Variable Selection 311
11.5.1.2DataData RecordRecord Selection 313
11.5.1.3ModelModel GenerationGeneration 313
11.5.1.5ModelModel Linearization 314
11.5.2Integrated Methodology for Designed DataDataThe material in this section was originally published in F. Castillo, 315
11.6Integrated Methodology in Action 316
11.7Summary 324
Suggested Reading 324
144233_1_En_12_Chapter_OnlinePDF 325
Chapter 12: How to Apply Computational Intelligence 325
When Is Computational Intelligence the Right Solution? 325
Obstacles in Applying Computational Intelligence 327
Technical Obstacles in Applying CI 327
Nontechnical Obstacles in Applying CI 328
Checklist ``Are we Ready?´´ 330
Methodology for Applying CI in a Business 330
Steps for Introducing CI in a Business 332
Steps for Applying CI in a Business 333
Steps for Leveraging CI in a Business 335
Computational Intelligence Project Management 336
Define Project Objectives and Scope 337
Define Roles 338
Select Computational Intelligence Methods 339
Prepare DataData 340
Develop ModelModel 342
Deploy ModelModel 343
ModelModel Maintenance and Support 344
CI for Six Sigma and Design for Six Sigma 345
Six SigmaSix Sigma and Design for Six SigmaSigma in IndustryIndustry 346
How CI Fits in Design for Six SigmaSix Sigma 351
Summary 354
The Bottom Line 355
Suggested Reading 355
144233_1_En_13_Chapter_OnlinePDF 356
Chapter 13: Computational Intelligence MarketingMarketing 356
Research MarketingMarketing Principles 356
Key Elements of MarketingMarketing 357
Research MarketingMarketing Strategy 358
TargetTarget Market IdentificationIdentification 358
Product Strategy Definition 359
Promotional Strategy Implementation 360
Techniques - Delivery, Visualization, Humor 361
Message Delivery 361
Effective VisualizationVisualization 363
Combining Mind-mapMind-maps and Clip Art 363
Some VisualizationVisualization Techniques 365
To PP or Not to PP? 365
Humor 367
Dilbert Cartoons 368
Useful QuotationsQuotations 368
Murphy's Laws Related to Computational Intelligence 370
Interactions Between Academia and IndustryIndustry 372
Protecting Intellectual Property 372
Publishing 374
Conference Advertising 375
Technology Development 375
Interaction with Vendors 376
MarketingMarketing CI to a Technical Audience 376
Guidelines for Preparing Technical Presentations for Applied Computational Intelligence 377
Key TargetTarget Audience for Technical Presentations 379
Visionary Guru 380
Open Mind Guru 380
Technical King Guru 381
Political Scientist Guru 381
Retiring Scientist Guru 382
1D Mind Guru 382
MarketingMarketing to a Nontechnical Audience 382
Guidelines for Preparing Nontechnical Presentations for Applied Computational Intelligence 382
Key TargetTarget Audience for Nontechnical Presentations 384
Summary 385
Suggested Reading 386
144233_1_En_14_Chapter_OnlinePDF 387
Chapter 14: Industrial Applications of Computational Intelligence 387
Applications in ManufacturingManufacturing 387
Robust Inferential Sensors 388
Robust Inferential Sensor for Alarm Detection 389
Robust Inferential Sensor for Product Transition Monitoring 390
Robust Inferential Sensor for BiomassBiomass Estimation 391
Automated Operating Discipline 393
Knowledge Acquisition from the Experts 394
Organization of the Knowledge Base 394
Implementation of Prototype for One Process Unit 396
Scaling up to the Full System for All Process Units 396
Operators' Involvement 396
Value Evaluation 396
Empirical EmulatorsEmulators for On-line OptimizationOptimization 397
Motivation for Developing Empirical EmulatorsEmulators 397
Empirical EmulatorsEmulators Structures 397
A Case Study: an Empirical Emulator for OptimizationOptimization of an Industrial Chemical ProcessA. Kordon, A. Kalos, and B. 399
Problem Definition 399
DataData Preparation 399
Empirical Emulator Based on Analytic Neural Networks 400
Empirical EmulatorsEmulators Based on Symbolic RegressionRegression 402
Applications in New Product Development 402
Accelerated Fundamental Model BuildingThe material in this section was originally published in: A. Kordon, H. Pham, C. Bosnyak 403
Potential of GP-Generated Symbolic RegressionRegression in Fundamental Model Building 404
Symbolic Regression in Fundamental Modeling of Structure-Properties 406
Case Study Description 406
Fundamental ModelModel Building Approach 407
Symbolic RegressionRegression Approach 407
Fast Robust Empirical ModelModel Building 408
Symbolic RegressionRegression Models of Blown Film Process Effects 411
Modeling Scope 411
Symbolic RegressionRegression ModelModel for DART Impact 411
Blown Film Process Effects ModelModel Implementation 412
Unsuccessful Computational Intelligence Applications 413
Application with Significant Cultural Change 414
Applications with Low-Quality DataData 414
Acknowledgements 415
Summary 415
The Bottom Line 415
Suggested Reading 415
p4 417
Part IV: The Future of Computational Intelligence 417
144233_1_En_15_Chapter_OnlinePDF 418
Chapter 15: Future Directions of Applied Computational Intelligence 418
Supply-Demand-Driven Applied Research 418
Limitations of Supply-Driven Research 419
What Is Supply-Demand Research? 421
Advantages of Supply-Demand Research 422
Mechanisms of Supply-Demand Research 422
Next-GenerationGeneration Applied Computational Intelligence 424
Computing with Words 424
Basic Principles of Computing with Words 425
Potential Application Areas for Computing with Words 427
Evolving Intelligent Systems 427
Basic Principles of Evolving Intelligent Systems 428
Potential Application Areas for Evolving Intelligent Systems 428
Co-evolving Systems 430
Basic Principles of Co-evolving Systems 431
Potential Application Areas for Co-evolving Systems 432
Artificial Immune Systems 433
Basic Principles of Artificial Immune Systems 434
Potential Application Areas for Artificial Immune Systems 435
Projected Industrial Needs 436
Predictive MarketingMarketing 436
Accelerated New Products Diffusion 437
High-Throughput InnovationInnovation 438
ManufacturingManufacturing at Economic OptimumOptimum 438
Predictive Optimal Supply-Chain 439
Intelligent Security 439
Reduced Virtual Bureaucracy 440
Emerging SimplicitySimplicity 440
Handling the Curse of Decentralization 441
SustainabilitySustainability of Applied Computational Intelligence 442
Potential Roadblocks 443
The Fun of Computational Intelligence 444
Summary 444
Suggested Reading 445
BM 446
Kordon_Index_o 458
: Index 458

Produkt-Details

• Erscheint lt. Verlag: 28.11.2009
• Zusatzinfo: XXII, 459 p.
• Verlagsort: Berlin
• Sprache: englisch
• Themenwelt: Mathematik / Informatik ► Informatik ► Datenbanken
• Themenwelt: Informatik ► Theorie / Studium ► Künstliche Intelligenz / Robotik
• Themenwelt: Technik
• Themenwelt: Wirtschaft ► Betriebswirtschaft / Management ► Logistik / Produktion
• Schlagworte: Artificial Intelligence • Business Intelligence • Computational Intelligence • Data Mining • empirical modeling • Evolutionary Computing • Fuzzy Systems • Industrial Applications • machine learning • Natural Computing • Optimization • process monitoring • Six Sigma • Six Sigma
• ISBN-10: 3-540-69913-9 / 3540699139
• ISBN-13: 978-3-540-69913-2 / 9783540699132