Intelligent Systems for Engineers and Scientists

The third edition of this bestseller examines the principles of artificial intelligence and their application to engineering and science, as well as techniques for developing intelligent systems to solve practical problems. Covering the full spectrum of intelligent systems techniques, it incorporates knowledge-based systems, computational intelligence, and their hybrids. Using clear and concise language, Intelligent Systems for Engineers and Scientists, Third Edition features updates and improvements throughout all chapters. It includes expanded and separated chapters on genetic algorithms and single-candidate optimization techniques, while the chapter on neural networks now covers spiking networks and a range of recurrent networks. The book also provides extended coverage of fuzzy logic, including type-2 and fuzzy control systems. Example programs using rules and uncertainty are presented in an industry-standard format, so that you can run them yourself. The first part of the book describes key techniques of artificial intelligence--including rule-based systems, Bayesian updating, certainty theory, fuzzy logic (types 1 and 2), frames, objects, agents, symbolic learning, case-based reasoning, genetic algorithms, optimization algorithms, neural networks, hybrids, and the Lisp and Prolog languages. The second part describes a wide range of practical applications in interpretation and diagnosis, design and selection, planning, and control. The author provides sufficient detail to help you develop your own intelligent systems for real applications. Whether you are building intelligent systems or you simply want to know more about them, this book provides you with detailed and up-to-date guidance.

Table of Contents:
Introduction Intelligent Systems A Spectrum of Intelligent Behavior Knowledge-Based Systems The Knowledge Base      Rules and Facts      Inference Networks      Semantic Networks Deduction, Abduction, and Induction The Inference Engine Declarative and Procedural Programming Expert Systems Knowledge Acquisition Search Computational Intelligence Integration with Other Software Further Reading Rule-Based Systems Rules and Facts A Rule-Based System for Boiler Control Rule Examination and Rule Firing Maintaining Consistency The Closed-World Assumption Use of Local Variables within Rules Forward Chaining (a Data-Driven Strategy)      Single and Multiple Instantiation of Local Variables      Rete Algorithm Conflict Resolution      First Come, First Served      Priority Values      Metarules Backward Chaining (a Goal-Driven Strategy)      The Backward-Chaining Mechanism      Implementation of Backward Chaining      Variations of Backward Chaining      Format of Backward-Chaining Rules A Hybrid Strategy Explanation Facilities Summary Further Reading Handling Uncertainty: Probability and Fuzzy Logic Sources of Uncertainty Bayesian Updating      Representing Uncertainty by Probability      Direct Application of Bayes’ Theorem      Likelihood Ratios      Using the Likelihood Ratios      Dealing with Uncertain Evidence      Combining Evidence      Combining Bayesian Rules with Production Rules      A Worked Example of Bayesian Updating      Discussion of the Worked Example      Advantages and Disadvantages of Bayesian Updating Certainty Theory      Introduction      Making Uncertain Hypotheses      Logical Combinations of Evidence           Conjunction           Disjunction           Negation      A Worked Example of Certainty Theory      Discussion of the Worked Example      Relating Certainty Factors to Probabilities Fuzzy Logic: Type-1      Crisp Sets and Fuzzy Sets      Fuzzy Rules      Defuzzification           Stage 1: Scaling the Membership Functions           Stage 2: Finding the Centroid      Defuzzifying at the Extremes      Sugeno Defuzzification      A Defuzzification Anomaly Fuzzy Control Systems      Crisp and Fuzzy Control      Fuzzy Control Rules      Defuzzification in Control Systems Fuzzy Logic: Type-2 Other Techniques      Dempster–Shafer Theory of Evidence      Inferno Summary Further Reading Agents, Objects, and Frames Birds of a Feather: Agents, Objects, and Frames Intelligent Agents Agent Architectures      Logic-Based Architectures      Emergent Behavior Architectures      Knowledge-Level Architectures       Layered Architectures Multiagent Systems      Benefits of a Multiagent System      Building a Multiagent System      Contract Nets      Cooperative Problem-Solving (CPS)      Shifting Matrix Management (SMM)      Comparison of Cooperative Models      Communication between Agents Swarm Intelligence Object-Oriented Systems       Introducing OOP      An Illustrative Example      Data Abstraction           Classes           Instances           Attributes (or Data Members)            Operations (or Methods or Member Functions)           Creation and Deletion of Instances      Inheritance           Single Inheritance           Multiple and Repeated Inheritance           Specialization of Methods           Class Browsers      Encapsulation      Unified Modeling Language (UML)      Dynamic (or Late) Binding      Message Passing and Function Calls      Metaclasses      Type Checking      Persistence      Concurrency      Active Values and Daemons      OOP Summary Objects and Agents Frame-Based Systems Summary: Agents, Objects, and Frames Further Reading Symbolic Learning Introduction Learning by Induction      Overview      Learning Viewed as a Search Problem      Techniques for Generalization and Specialization           Universalization           Replacing Constants with Variables           Using Conjunctions and Disjunctions           Moving up or down a Hierarchy           Chunking Case-Based Reasoning (CBR)      Storing Cases           Abstraction Links and Index Links           Instance-of Links           Exemplar Links           Failure Links      Retrieving Cases      Adapting Case Histories           Null Adaptation           Parameterization           Reasoning by Analogy           Critics           Reinstantiation           Dealing with Mistaken Conclusions Summary Further Reading Single-Candidate Optimization Algorithms  Optimization The Search Space Searching the Parameter Space Hill-Climbing and Gradient Descent Algorithms           Hill-Climbing           Steepest Gradient Descent or Ascent           Gradient-Proportional Descent or Ascent           Conjugate Gradient Descent or Ascent           Tabu Search Simulated Annealing Summary Further Reading Genetic Algorithms for Optimization Introduction The Basic GA           Chromosomes           Algorithm Outline           Crossover           Mutation           Validity Check Selection           Selection Pitfalls           Fitness-Proportionate Selection           Fitness Scaling for Improved Selection                     Linear Fitness Scaling                     Sigma Scaling                     Linear Rank Scaling           Nonlinear Rank Scaling           Probabilistic Nonlinear Rank Scaling           Truncation Selection           Transform Ranking           Tournament Selection           Comparison of Selection Methods Elitism Multiobjective Optimization Gray Code Building Block Hypothesis           Schema Theorem           Inversion Selecting GA Parameters Monitoring Evolution Genetic Programming Other Forms of Population-Based Optimization Summary Further Reading Neural Networks Introduction Neural Network Applications           Classification           Nonlinear Estimation           Clustering           Content-Addressable Memory Nodes and Interconnections Single and Multilayer Perceptrons           Network Topology           Perceptrons as Classifiers           Training a Perceptron           Buffered Perceptrons            Some Practical Considerations Recurrent Networks           Simple Recurrent Network (SRN)           Hopfield Network           MAXNET           The Hamming Network Unsupervised Networks           Adaptive Resonance Theory (ART) Networks           Kohonen Self-Organizing Networks           Radial Basis Function Networks Spiking Neural Networks Summary Further Reading Hybrid Systems Convergence of Techniques Blackboard Systems for Multifaceted Problems Parameter Setting           Genetic–Neural Systems           Genetic–Fuzzy Systems Capability Enhancement           Neuro–Fuzzy Systems           Baldwinian and Lamarckian Inheritance in Genetic Algorithms           Learning Classifier Systems Clarification and Verification of Neural Network Outputs Summary Further Reading Artificial Intelligence Programming Languages A Range of Intelligent Systems Tools Features of AI Languages           Lists           Other Data Types           Programming Environments Lisp           Background           Lisp Functions           A Worked Example Prolog           Background           Backtracking in Prolog Comparison of AI Languages Summary Further Reading Systems for Interpretation and Diagnosis Introduction Deduction and Abduction for Diagnosis           Exhaustive Testing           Explicit Modeling of Uncertainty           Hypothesize-and-Test  Depth of Knowledge           Shallow Knowledge           Deep Knowledge           Shallow and Deep Knowledge Model-Based Reasoning           The Limitations of Rules           Modeling Function, Structure, and State                     Function                     Structure                     State           Using the Model           Monitoring           Tentative Diagnosis                     The Shotgun Approach                     Structural Isolation                     The Heuristic Approach           Fault Simulation           Fault Repair           Using Problem Trees           Summary of Model-Based Reasoning Case Study: A Blackboard System for Interpreting Ultrasonic Images           Ultrasonic Imaging           Agents in DARBS           Rules in DARBS           The Stages of Image Interpretation                     Arc Detection Using the Hough Transform                     Gathering the Evidence                     Defect Classification           The Use of Neural Networks                     Defect Classification Using a Neural Network                     Echodynamic Classification Using a Neural Network                      Combining the Two Applications of Neural Networks           Rules for Verifying Neural Networks Summary Further Reading Systems for Design and Selection The Design Process Design as a Search Problem Computer-Aided Design The Product Design Specification (PDS): A Telecommunications Case Study           Background           Alternative Views of a Network           The Classes                     Network                     Link                     Site                      Information Stream                     Equipment           Summary of PDS Case Study Conceptual Design Constraint Propagation and Truth Maintenance Case Study: Design of a Lightweight Beam           Conceptual Design           Optimization and Evaluation           Detailed Design Design as a Selection Exercise           Overview           Merit Indices           The Polymer Selection Example           Two-Stage Selection           Constraint Relaxation           A Naive Approach to Scoring           A Better Approach to Scoring           Case Study: Design of a Kettle           Reducing the Search Space by Classification Failure Mode and Effects Analysis (FMEA) Summary Further Reading Systems for Planning Introduction Classical Planning Systems STRIPS           General Description           An Example Problem           A Simple Planning System in Prolog Considering the Side Effects of Actions           Maintaining a World Model           Deductive Rules Hierarchical Planning           Description           Benefits of Hierarchical Planning           Hierarchical Planning with ABSTRIPS Postponement of Commitment           Partial Ordering of Plans           The Use of Planning Variables Job-Shop Scheduling           The Problem           Some Approaches to Scheduling Constraint-Based Analysis           Constraints and Preferences           Formalizing the Constraints           Identifying the Critical Sets of Operations           Sequencing in Disjunctive Case           Sequencing in Nondisjunctive Case           Updating Earliest Start Times and Latest Finish Times           Using Constraints and Preferences Replanning and Reactive Planning Summary Further Reading Systems for Control Introduction Low-Level Control           Open-Loop Control           Feedforward Control           Feedback Control           First- and Second-Order Models           Algorithmic Control: The PID Controller           Bang-Bang Control Requirements of High-Level (Supervisory) Control Blackboard Maintenance Time-Constrained Reasoning           Prioritization of Processes           Approximation                     Approximate Search                     Data Approximations                     Knowledge Approximations                     Single and Multiple Instantiation Fuzzy Control The BOXES Controller           The Conventional BOXES Algorithm           Fuzzy BOXES Neural Network Controllers           Direct Association of State Variables with Action Variables           Estimation of Critical State Variables Statistical Process Control (SPC)           Applications           Collecting the Data           Using the Data Summary Further Reading The Future of Intelligent Systems Benefits Trends in Implementation  Intelligent Systems and the Internet Ubiquitous Intelligent Systems Conclusion References Index

About the Author :
Adrian Hopgood earned his BSc from the University of Bristol, PhD from the University of Oxford, and MBA from the Open University. After completing his PhD in 1984, he spent 2 years developing applied intelligent systems for Systems Designers PLC. That experience set the direction of his career toward the investigation of intelligent systems and their practical applications. After leaving Systems Designers, he spent 14 years at the Open University and remains attached as a visiting professor. During that period, he also spent 2 years at Telstra Research Laboratories in Australia, investigating the role of intelligent systems in telecommunications. He has subsequently worked for Nottingham Trent University, De Montfort University, and Sheffield Hallam University. Despite assuming senior management positions, he has not lost his passion for intelligent systems. He has recently led the development of an open-source blackboard system, DARBS. His Website is www.adrianhopgood.com.