Intelligent Systems for Engineers and Scientists: A Practical Guide to Artificial Intelligence

Cover
Half Title
Title Page
Copyright Page
Dedication
Table of Contents
Preface
Acknowledgements
Author
Chapter 1: Introduction
	1.1 Artificial Intelligence and Intelligent Systems
	1.2 A Spectrum of Intelligent Behavior
	1.3 Knowledge-Based Systems (KBSs)
	1.4 The Knowledge Base
		1.4.1 Rules and Facts
		1.4.2 Inference Networks
		1.4.3 Semantic Networks
	1.5 Deduction, Abduction, and Induction
	1.6 The Inference Engine
	1.7 Declarative and Procedural Programming
	1.8 Expert Systems
	1.9 Knowledge Acquisition
	1.10 Search
	1.11 Computational Intelligence (CI)
	1.12 Integration with Other Software
	Further Reading
Chapter 2: Rule-Based Systems
	2.1 Rules and Facts
	2.2 A Rule-Based System for Boiler Control
	2.3 Rule Examination and Rule Firing
	2.4 Maintaining Consistency
	2.5 The Closed-World Assumption
	2.6 Use of Local Variables within Rules
	2.7 Forward Chaining (a Data-Driven Strategy)
		2.7.1 Single and Multiple Instantiation of Local Variables
		2.7.2 Rete Algorithm
	2.8 Conflict Resolution
		2.8.1 First Come, First Served
		2.8.2 Priority Values
		2.8.3 Metarules
	2.9 Backward Chaining (a Goal-Driven Strategy)
		2.9.1 The Backward-Chaining Mechanism
		2.9.2 Implementation of Backward Chaining
		2.9.3 Variations of Backward Chaining
		2.9.4 Format of Backward-Chaining Rules
	2.10 A Hybrid Strategy
	2.11 Explanation Facilities
	2.12 Summary
	Further Reading
Chapter 3: Handling Uncertainty: Probability and Fuzzy Logic
	3.1 Sources of Uncertainty
	3.2 Bayesian Updating
		3.2.1 Representing Uncertainty by Probability
		3.2.2 Direct Application of Bayes’ Theorem
		3.2.3 Likelihood Ratios
		3.2.4 Using the Likelihood Ratios
		3.2.5 Dealing with Uncertain Evidence
		3.2.6 Combining Evidence
		3.2.7 Combining Bayesian Rules with Production Rules
		3.2.8 A Worked Example of Bayesian Updating
		3.2.9 Discussion of the Worked Example
		3.2.10 Advantages and Disadvantages of Bayesian Updating
	3.3 Certainty Theory
		3.3.1 Introduction
		3.3.2 Making Uncertain Hypotheses
		3.3.3 Logical Combinations of Evidence
			3.3.3.1 Conjunction
			3.3.3.2 Disjunction
			3.3.3.3 Negation
		3.3.4 A Worked Example of Certainty Theory
		3.3.5 Discussion of the Worked Example
		3.3.6 Relating Certainty Factors to Probabilities
	3.4 Fuzzy Logic: Type-1
		3.4.1 Crisp Sets and Fuzzy Sets
		3.4.2 Fuzzy Rules
		3.4.3 Defuzzification
			3.4.3.1 Stage 1: Scaling the Membership Functions
			3.4.3.2 Stage 2: Finding the Centroid
			3.4.3.3 Defuzzifying at the Extremes
			3.4.3.4 Sugeno Defuzzification
			3.4.3.5 A Defuzzification Anomaly
	3.5 Fuzzy Control Systems
		3.5.1 Crisp and Fuzzy Control
		3.5.2 Fuzzy Control Rules
		3.5.3 Defuzzification in Control Systems
	3.6 Fuzzy Logic: Type-2
	3.7 Other Techniques
		3.7.1 Dempster–Shafer Theory of Evidence
		3.7.2 Inferno
	3.8 Summary
	Further Reading
Chapter 4: Agents, Objects, and Frames
	4.1 Birds of a Feather: Agents, Objects, and Frames
	4.2 Intelligent Agents
	4.3 Agent Architectures
		4.3.1 Logic-Based Architectures
		4.3.2 Emergent Behavior Architectures
		4.3.3 Knowledge-Level Architectures
		4.3.4 Layered Architectures
	4.4 Multiagent Systems (MASs)
		4.4.1 Benefits of a Multiagent System
		4.4.2 Building a Multiagent System
		4.4.3 Contract Nets
		4.4.4 Cooperative Problem-Solving (CPS)
		4.4.5 Shifting Matrix Management (SMM)
		4.4.6 Comparison of Cooperative Models
		4.4.7 Communication Between Agents
	4.5 Swarm Intelligence
	4.6 Object-Oriented Systems
		4.6.1 Introducing Object-Oriented Programming (OOP)
		4.6.2 An Illustrative Example
		4.6.3 Data Abstraction
			4.6.3.1 Classes
			4.6.3.2 Instances
			4.6.3.3 Attributes (or Data Members)
			4.6.3.4 Operations (or Methods or Member Functions)
			4.6.3.5 Creation and Deletion of Instances
		4.6.4 Inheritance
			4.6.4.1 Single Inheritance
			4.6.4.2 Multiple and Repeated Inheritance
			4.6.4.3 Specialization of Methods
			4.6.4.4 Class Browsers
		4.6.5 Encapsulation
		4.6.6 Unified Modeling Language (UML)
		4.6.7 Dynamic (or Late) Binding
		4.6.8 Message Passing and Function Calls
		4.6.9 Metaclasses
		4.6.10 Type Checking
		4.6.11 Persistence
		4.6.12 Concurrency
		4.6.13 Active Values and Daemons
		4.6.14 Summary of Object-Oriented Systems
	4.7 Objects and Agents
	4.8 Frame-Based Systems
	4.9 Summary: Agents, Objects, and Frames
	Further Reading
Chapter 5: Symbolic Learning
	5.1 Introduction
	5.2 Learning by Induction
		5.2.1 Overview
		5.2.2 Learning Viewed as a Search Problem
		5.2.3 Techniques for Generalization and Specialization
			5.2.3.1 Universalization
			5.2.3.2 Replacing Constants with Variables
			5.2.3.3 Using Conjunctions and Disjunctions
			5.2.3.4 Moving Up or Down a Hierarchy
			5.2.3.5 Chunking
	5.3 Case-Based Reasoning (CBR)
		5.3.1 Storing Cases
			5.3.1.1 Abstraction Links and Index Links
			5.3.1.2 Instance-Of Links
			5.3.1.3 Scene Links
			5.3.1.4 Exemplar Links
			5.3.1.5 Failure Links
		5.3.2 Retrieving Cases
		5.3.3 Adapting Case Histories
			5.3.3.1 Null Adaptation
			5.3.3.2 Parameterization
			5.3.3.3 Reasoning by Analogy
			5.3.3.4 Critics
			5.3.3.5 Reinstantiation
		5.3.4 Dealing with Mistaken Conclusions
	5.4 Summary
	Further Reading
Chapter 6: Single-Candidate Optimization Algorithms
	6.1 Optimization
	6.2 The Search Space
	6.3 Searching the Parameter Space
	6.4 Hill-Climbing and Gradient-Descent Algorithms
		6.4.1 Hill-Climbing
		6.4.2 Steepest Gradient Descent or Ascent
		6.4.3 Gradient-Proportional Descent or Ascent
		6.4.4 Conjugate Gradient Descent or Ascent
		6.4.5 Tabu Search
	6.5 Simulated Annealing
	6.6 Summary
	Further Reading
Chapter 7: Genetic Algorithms for Optimization
	7.1 Introduction: Evolutionary Algorithms
	7.2 The Basic Genetic Algorithm
		7.2.1 Chromosomes
		7.2.2 Algorithm Outline
		7.2.3 Crossover
		7.2.4 Mutation
		7.2.5 Validity Check
	7.3 Selection
		7.3.1 Selection Pitfalls
		7.3.2 Fitness-Proportionate Selection
		7.3.3 Fitness Scaling for Improved Selection
			7.3.3.1 Linear Fitness Scaling
			7.3.3.2 Sigma Scaling
			7.3.3.3 Boltzmann Fitness Scaling
			7.3.3.4 Linear Rank Scaling
			7.3.3.5 Nonlinear Rank Scaling
			7.3.3.6 Probabilistic Nonlinear Rank Scaling
			7.3.3.7 Truncation Selection
			7.3.3.8 Transform Ranking
		7.3.4 Tournament Selection
		7.3.5 Comparison of Selection Methods
	7.4 Elitism
	7.5 Multiobjective Optimization
	7.6 Gray Code
	7.7 Variable-Length Chromosomes
	7.8 Building Block Hypothesis
		7.8.1 Schema Theorem
		7.8.2 Inversion
	7.9 Selecting GA Parameters
	7.10 Monitoring Evolution
	7.11 Finding Multiple Optima
	7.12 Genetic Programming (GP)
	7.13 Other Forms of Population-Based Optimization
	7.14 Summary
	Further Reading
Chapter 8: Shallow Neural Networks
	8.1 Introduction
	8.2 Neural Network Applications
		8.2.1 Classification
		8.2.2 Nonlinear Estimation and Prediction
		8.2.3 Clustering
		8.2.4 Memory and Recall
	8.3 Nodes and Interconnections
	8.4 Single and Multilayer Perceptrons (SLPs and MLPs)
		8.4.1 Network Topology
		8.4.2 Perceptrons as Classifiers
		8.4.3 Training a Perceptron
		8.4.4 Hierarchical Perceptrons
		8.4.5 Buffered Perceptrons
		8.4.6 Some Practical Considerations
			8.4.6.1 Overtraining
			8.4.6.2 Leave-One-Out and K-Fold Cross-Validation
			8.4.6.3 Data Scaling
	8.5 Recurrent Networks
		8.5.1 Simple Recurrent Network (SRN)
		8.5.2 Hopfield Network
		8.5.3 Maxnet
		8.5.4 The Hamming Network
	8.6 Unsupervised Networks
		8.6.1 Adaptive Resonance Theory (ART) Networks
		8.6.2 Kohonen Self-Organizing Networks
		8.6.3 Radial Basis Function (RBF) Networks
	8.7 Spiking Neural Networks (SNNs)
	8.8 Summary
	Further Reading
Chapter 9: Deep Neural Networks
	9.1 Deep Learning
	9.2 Convolutional Neural Networks (CNNs) for Image Recognition
		9.2.1 Origins
		9.2.2 Motivation for Convolutional Networks
		9.2.3 CNN Structure
			9.2.3.1 Input Layer
			9.2.3.2 Feature Maps
			9.2.3.3 Pooling and Classification Layers
		9.2.4 Pretrained Networks and Transfer Learning
		9.2.5 CNNs in Context
	9.3 Generative Networks
		9.3.1 Generative Versus Discriminative Algorithms
		9.3.2 Autoencoder Networks
		9.3.3 Generative Adversarial Networks (GANs)
	9.4 Long Short-Term Memory (LSTM) Networks
	9.5 Summary
	Further Reading
Chapter 10: Hybrid Systems
	10.1 Convergence of Techniques
	10.2 Blackboard Systems for Multifaceted Problems
	10.3 Parameter Setting
		10.3.1 Genetic–Neural Systems
		10.3.2 Genetic–Fuzzy Systems
		10.3.3 Fuzzy–Genetic Systems
	10.4 Capability Enhancement
		10.4.1 Neuro–Fuzzy Systems
		10.4.2 Memetic Algorithms: Genetic Algorithms with Local Search
		10.4.3 Learning Classifier Systems (LCSs)
	10.5 Clarification and Verification of Neural Network Outputs
	10.6 Summary
	Further Reading
Chapter 11: AI Programming Languages and Tools
	11.1 A Range of Intelligent Systems Tools
	11.2 Features of AI Languages
		11.2.1 Lists
		11.2.2 Other Data Types
		11.2.3 Programming Environments
	11.3 Lisp
		11.3.1 Background
		11.3.2 Lisp Functions
		11.3.3 A Worked Example
	11.4 Prolog
		11.4.1 Background
		11.4.2 A Worked Example
		11.4.3 Backtracking in Prolog
	11.5 Python
		11.5.1 Background
		11.5.2 A Worked Example
	11.6 Comparison of AI Languages
	11.7 Summary
	Further Reading
		Lisp
		Prolog
		Python
Chapter 12: Systems for Interpretation and Diagnosis
	12.1 Introduction
	12.2 Deduction and Abduction for Diagnosis
		12.2.1 Exhaustive Testing
		12.2.2 Explicit Modeling of Uncertainty
		12.2.3 Hypothesize-and-Test
	12.3 Depth of Knowledge
		12.3.1 Shallow Knowledge
		12.3.2 Deep Knowledge
		12.3.3 Combining Shallow and Deep Knowledge
	12.4 Model-Based Reasoning
		12.4.1 The Limitations of Rules
		12.4.2 Modeling Function, Structure, and State
			12.4.2.1 Function
			12.4.2.2 Structure
			12.4.2.3 State
		12.4.3 Using the Model
		12.4.4 Monitoring
		12.4.5 Tentative Diagnosis
			12.4.5.1 The Shotgun Approach
			12.4.5.2 Structural Isolation
			12.4.5.3 The Heuristic Approach
		12.4.6 Fault Simulation
		12.4.7 Fault Repair
		12.4.8 Using Problem Trees
		12.4.9 Summary of Model-Based Reasoning
	12.5 Case Study: A Blackboard System for Interpreting Ultrasonic Images
		12.5.1 Ultrasonic Imaging
		12.5.2 Agents in DARBS
		12.5.3 Rules in DARBS
		12.5.4 The Stages of Image Interpretation
			12.5.4.1 Arc Detection Using the Hough Transform
			12.5.4.2 Gathering the Evidence
			12.5.4.3 Defect Classification
		12.5.5 The Use of Neural Networks
			12.5.5.1 Defect Classification Using a Neural Network
			12.5.5.2 Echodynamic Classification Using a Neural Network
			12.5.5.3 Combining the Two Applications of Neural Networks
		12.5.6 Rules for Verifying Neural Networks
	12.6 Summary
	Further Reading
Chapter 13: Systems for Design and Selection
	13.1 The Design Process
	13.2 Design as a Search Problem
	13.3 Computer-Aided Design
	13.4 The Product Design Specification (PDS): A Telecommunications Case Study
		13.4.1 Background
		13.4.2 Alternative Views of a Network
		13.4.3 Implementation
		13.4.4 The Classes
			13.4.4.1 Network
			13.4.4.2 Link
			13.4.4.3 Information Stream
			13.4.4.4 Site
			13.4.4.5 Equipment
		13.4.5 Summary of PDS Case Study
	13.5 Conceptual Design
	13.6 Constraint Propagation and Truth Maintenance
	13.7 Case Study: The Design of a Lightweight Beam
		13.7.1 Conceptual Design
		13.7.2 Optimization and Evaluation
		13.7.3 Detailed Design
	13.8 Design as a Selection Exercise
		13.8.1 Overview
		13.8.2 Merit Indices
		13.8.3 The Polymer Selection Example
		13.8.4 Two-Stage Selection
		13.8.5 Constraint Relaxation
		13.8.6 A Naive Approach to Scoring
		13.8.7 A Better Approach to Scoring
		13.8.8 Case Study: The Design of a Kettle
		13.8.9 Reducing the Search Space by Classification
	13.9 Failure Mode and Effects Analysis (FMEA)
	13.10 Summary
	Further Reading
Chapter 14: Systems for Planning
	14.1 Introduction
	14.2 Classical Planning Systems
	14.3 Stanford Research Institute Problem Solver (STRIPS)
		14.3.1 General Description
		14.3.2 An Example Problem
		14.3.3 A Simple Planning System in Prolog
	14.4 Considering the Side Effects of Actions
		14.4.1 Maintaining a World Model
		14.4.2 Deductive Rules
	14.5 Hierarchical Planning
		14.5.1 Description
		14.5.2 Benefits of Hierarchical Planning
		14.5.3 Hierarchical Planning with ABSTRIPS
	14.6 Postponement of Commitment
		14.6.1 Partial Ordering of Plans
		14.6.2 The Use of Planning Variables
	14.7 Job-Shop Scheduling
		14.7.1 The Problem
		14.7.2 Some Approaches to Scheduling
	14.8 Constraint-Based Analysis (CBA)
		14.8.1 Constraints and Preferences
		14.8.2 Formalizing the Constraints
		14.8.3 Identifying the Critical Sets of Operations
		14.8.4 Sequencing in the Disjunctive Case
		14.8.5 Sequencing in the Nondisjunctive Case
		14.8.6 Updating Earliest Start Times and Latest Finish Times
		14.8.7 Applying Preferences
		14.8.8 Using Constraints and Preferences
	14.9 Replanning and Reactive Planning
	14.10 Summary
	Further Reading
Chapter 15: Systems for Control
	15.1 Introduction
	15.2 Low-Level Control
		15.2.1 Open-Loop Control
		15.2.2 Feedforward Control
		15.2.3 Feedback Control
		15.2.4 First- and Second-Order Models
		15.2.5 Algorithmic Control: The PID Controller
		15.2.6 Bang-Bang Control
	15.3 Requirements of High-Level (Supervisory) Control
	15.4 Blackboard Maintenance
	15.5 Time-Constrained Reasoning
		15.5.1 Prioritization of Processes
		15.5.2 Approximation
			15.5.2.1 Approximate Search
			15.5.2.2 Data Approximations
			15.5.2.3 Knowledge Approximations
		15.5.3 Single and Multiple Instantiation
	15.6 Fuzzy Control
	15.7 The BOXES Controller
		15.7.1 The Conventional BOXES Algorithm
		15.7.2 Fuzzy BOXES
	15.8 Neural Network Controllers
		15.8.1 Direct Association of State Variables with Action Variables
		15.8.2 Estimation of Critical State Variables
	15.9 Statistical Process Control (SPC)
		15.9.1 Applications
		15.9.2 Collecting the Data
		15.9.3 Using the Data
	15.10 Summary
	Further Reading
Chapter 16: The Future of Intelligent Systems
	16.1 Benefits
	16.2 Trends in Implementation
	16.3 Intelligent Systems and the Internet
	16.4 Computational Power
	16.5 Ubiquitous Intelligent Systems
	16.6 Ethics
	16.7 Conclusions
References
Index