人工智能:复杂问题求解的结构和策略(英文影印版.第6版)

.3.1 graph theory
3.2 strategies for state space search
3.3 using the state space to represent reasoning with the predicate calculus
3.4 epilogue and references
3.5 exercises
4 heuristic search
4.0 introduction
4.1 hill climbing and dynamic programming
4.2 the best-first search algorithm
4.3 admissibility, monotonicity, and informedness
4.4 using heuristics in games
4.5 complexity issues
4.6 epilogue and references
4.7 exercises
5 stochastic methods
5.0 introduction
5.1 the elements of counting
5.2 elements of probability theory
5.3 applications of the stochastic methodology
5.4 bayes' theorem
5.5 epilogue and references
5.6 exercises
6 controland implementation of state space search
6.0 introduction
6.1 recursion-based search
6.2 production systems
6.3 the blackboard architecture for problem solving
6.4 epilogue and references
6.5 exercises
part iii capturing intelligence: the ai challenge
7 knowledge representation
7.0 issues in knowledge representation
7.1 a brief history of ai representational systems
7.2 conceptual graphs: a network language
7.3 alternative representations and ontologies
7.4 agent based and distributed problem solving
7.5 epilogue and references
7.6 exercises
8 strong method problem solving
8.0 introduction
8.1 overview of expert system technology
8.2 rule-based expert systems
8.3 model-based, case based, and hybrid systems
8.4 planning
8.5 epilogue and references
8.6 exercises ..
9 reasoning in uncertain situations
9.0 introduction
9.1 logic-based abductive inference
9.2 abduction: alternatives to logic
9.3 the stochastic approach to uncertainty
9.4 epilogue and references
9.5 exercises
part iv machine learning
10 machine learning: symbol-based
10.0 introduction
10.1 a framework for symbol-based learning
10.2 version space search
10.3 the id3 decision tree induction algorithm
10.4 inductive bias and learnability
10.5 knowledge and learning
10.6 unsupervised learning
10.7 reinforcement learning
10.8 epilogue and references
10.9 exercises
11 machine learning: connectionist
11.0 introduction
11.1 foundations for connectionist networks
11.2 perceptron learning
11.3 backpropagation learning
11.4 competitive learning
11.5 hebbian coincidence learning
11.6 attractor networks or "memories"
11.7 epilogue and references
11.8 exercises
12 machine learning: genetic and emergent
12.0 genetic and emergent models of learning
12.1 the genetic algorithm
12.2 classifier systems and genetic programming
12.3 artificial life and society-based learning
12.4 epilogue and references
12.5 exercises
13 machine learning: probabilistic
13.0 stochastic and dynamic models of learning
13.1 hidden markov models (hmms)
13.2 dynamic bayesian networks and learning
13.3 stochastic extensions to reinforcement learning
13.4 epilogue and references
13.5 exercises
part v advanced topics for al problem solving
14 automated reasoning
14.0 introduction to weak methods in theorem proving
14.1 the general problem solver and difference tables
14.2 resolution theorem proving
14.3 prolog and automated reasoning
14.4 further issues in automated reasoning
14.5 epilogue and references
14.6 exercises
15 understanding natural language
15.0 the natural language understanding problem
15.1 deconstmcting language: an analysis
15.2 syntax
15.3 transition network parsers and semantics
15.4 stochastic tools for language understanding
15.5 natural language applications
15.6 epilogue and references
15.7 exercises
part vi epilogue
16 artificial intelligence as empirical enquiry
16.0 introduction
16.1 artificial intelligence: a revised definition
16.2 the science oflntelligent systems
16.3 al: current challanges and future direstions
16.4 epilogue and references
bibliography
author index
subject index ...

.　　Our third commitment was made at the beginning of this book's life cycle: to place artificial intelligence within the context of empirical science. In the spirit of the Newell and Simon (1976) Turing award lecture we quote from an earlier edition:
　　... AI is not some strange aberration from the scientific tradition, but.., part of a general quest for knowledge about, and the understanding of, intelligence itself. Furthermore, our AI programming tools, along with the exploratory programming methodology.., are ideal for exploring an environment. Our tools give us a medium for both understanding and questions. We come to appreciate and know phenomena constructively, that is, by progressive approximation.
　　Thus we see each design and program as an experiment with nature: we propose a representation, we generate a search algorithm, and we question the adequacy of our characterization to account for part of the phenomenon of intelligence. And the natural world gives a response to our query. Our experiment can be deconstmcted, revised, extended, and run again. Our model can be refined, our understanding extended.
　　New with The Sixth Edition
　　The biggest change for the sixth edition is the extension of the stochastic approaches to AI. To accomplish this we revised Section 9.3 and added a new chapter (13) introducing probability-based machine learning. Our presentation of stochastic AI tools and their application to learning and natural language is now more comprehensive.
　　From probability theory's foundations in set theory and counting we develop the notions of probabilities, random variables, and independence. We present and use Bayes' theorem first with one symptom and one disease and then in its full general form. We examine the hypotheses that underlie the use of Bayes and then present the argmax and naive Bayes approaches. We present examples of stochastic reasoning, including the analysis of language phenomena and the Vierbi algorithm. We also introduce the idea of conditional independence that leads to Bayesian belief networks, the BBN, in Chapter 9.
　　In Chapter 13 we introduce hidden Markov models, HMMs, and show their use in several examples. We also present several HMM variants, including the auto-regressive and hierarchical HMMs. We present dynamic Bayesian networks, DBNs, and demonstrate their use. We discuss parameter and structure learning and present the expectation maximization algorithm and demonstrate its use with loopy belief propagation. Finally, we present Markov decision processes, the MDP, and partially observable Markov decision process, the POMDP, in the context of an extension to the earlier presentation of reinforce- ment learning.
　　We include several more examples of probabilistic finite state machines and probabilistic acceptors, as well as the use of dynamic programming, especially with stochastic measures (the Viterbi algorithm). We added a stochastic English language parser (based on the work of Mark Steedman at the University of Edinburgh) as well as the use of dynamic programming with the Earley parser.
　　We made a major decision to remove the Prolog and Lisp chapters from the book. Part of the reason for this is that these were getting too large. We have also accumulated a number of AI algorithms written in Java. When we added the new Chapter 13 on stochastic approaches to machine learning, we determined that the book was getting too large/cumbersome. Thus the sixth edition is more than 150 pages smaller than the fifth and the AI algorithms in Prolog, Lisp, and Java are being released as supplementary materials. From our earliest days in AI we have always felt that the way to understand the power (and limitations) of AI algorithms is constructively - that is, by building them! We encourage our present generation of readers to do exactly this: to visit the supplementary matedals: to build and experiment directly with the algorithms we present.
　　Finally, we have done the usual updating of references and materials that a new edition warrants. In a revised Chapter 16, we return to the deeper questions on the nature of intelligence and the possibility of creating intelligent machines.
　　Sixth Edition: The Contents
　　Chapter 1 introduces artificial intelligence, beginning with a brief history of attempts to understand mind and intelligence in philosophy, psychology, and other areas of research. In an important sense, AI is an old science, tracing its roots back at least to Aristotle. An appreciation of this background is essential for an understanding of the issues addressed in modern research. We also present an overview of some of the important application areas in Al. Our goal in Chapter 1 is to provide both background and a motivation for the theory and applications that follow. ..
　　Chapters 2, 3, 4, 5, and 6 (Part II) introduce the research tools for A1 problem solving. These include, in Chapter 2, the predicate calculus presented both as a mathematical system as well as a representation language to describe the essential features of a problem. Search, and the algorithms and data structures used to implement search, are introduced in Chapter 3, to organize the exploration of problem situations. In Chapter 4, we discuss the essential role of heuristics in focusing and constraining search-based problem solving. In Chapter 5, we introduce the stochastic methodology, important technology for reasoning in situations of uncertainty. In Chapter 6, we present a number of software architectures, including the blackboard and production system, for implementing these search algorithms.
　　Chapters 7, 8, and 9 make up Part III: representations for Al, knowledge-intensive problem solving, and reasoning in changing and ambiguous situations. In Chapter 7 we present the evolving story of Al representational schemes. We begin with a discussion of association-based networks and extend this model to include conceptual dependency theory, frames, and scripts. We then present an in-depth examination of a particular formalism, conceptual graphs, emphasizing the epistemological issues involved in representing knowledge and showing how these issues are addressed in a modern representation language. Expanding on this formalism in Chapter 14, we show how conceptual graphs can be used to implement a natural language database front end. We conclude Chapter 7 with more modern approaches to representation, including Copycat and agentoriented architectures.
　　Chapter 8 presents the rule-based expert system along with case-based and modelbased reasoning, including examples from the NASA space program. These approaches to problem solving are presented as a natural evolution of the material in Part II: using a production system of predicate calculus expressions to orchestrate a graph search. We end with an analysis of the strengths and weaknesses of each of these approaches to knowledge-intensive problem solving.
　　Chapter 9 presents models for reasoning with uncertainty as well as the use of unreliable information. We introduce Bayesian models, belief networks, Dempster-Shafer, causal models, and the Stanford certainty algebra for reasoning in uncertain situations. Chapter 9 also contains algorithms for truth maintenance, reasoning with minimum models, logic-based abduction, and the clique-tree algorithm for Bayesian belief networks.
　　Part IV, Chapters 10 through 13, is an extensive presentation of issues in machine learning. In Chapter 10 we offer a detailed look at algorithms for symbol-based learning, a fruitful area of research spawning a number of different problems and solution approaches. These learning algorithms vary in their goals, the training data considered, their learning strategies, and the knowledge representations they employ. Symbol-based learning includes induction, concept learning, version-space search, and ID3. The role of inductive bias is considered, generalizations from patterns of data, as well as the effective use of knowledge to learn from a single example in explanation-based learning. Category learning, or conceptual clustering, is presented with unsupervised learning. Reinforcement learning, or the ability to integrate feedback from the environment into a policy for making new decisions concludes the chapter.
　　In Chapter 11 we present neural networks, often referred to as sub-symbolic or connectionist models of learning. In a neural net, information is implicit in the organization and weights on a set of connected processors, and learning involves a re-arrangement and modification of the overall weighting of nodes and structure of the system. We present a number of connectionist architectures, including perceptron learning, backpropagation, and counterpropagation. We demonstrate Kohonen, Grossberg, and Hebbian models. We present associative learning as well as attractor models, including examples of Hopfield networks.
　　Genetic algorithms and evolutionary approaches to learning are introduced in Chapter 12. On this viewpoint, learning is cast as an emerging and adaptive process. After several examples of problem solutions based on genetic algorithms, we introduce the application of genetic techniques to more general problem solvers. These include classifier systems and genetic programming. We then describe society-based learning with examples from artificial life, called a-life, research. We conclude the chapter with an example of emergent computation from research at the Santa Fe Institute.
　　Chapter 13 presents stochastic approaches to machine learning. We begin with a definition of hidden markov models and then present several important variations including the auto-regressive and hierarchical HMM. We then present dynamic Bayesian networks, a generalization of the HMM, and also able to track systems across periods of time. These techniques are useful for modeling the changes in complex environments as is required for diagnostic and prognostic reasoning. Finally, we add a probabilistic component to reinforcement learning first introduced in Chapter 10. This includes presentation of the Markov decision process (or MDP) and the partially observed Markov decision process (or POMDP).
　　Part V, Chapters 14 and 15, presents automated reasoning and natural language understanding. Theorem proving, often referred to as automated reasoning, is one of the oldest areas of AI research. In Chapter 14, we discuss the first programs in this area, including the Logic Theorist and the General Problem Solver. The primary focus of the chapter is binary resolution proof procedures, especially resolution refutations. More advanced inferencing with hyper-resolution and paramodulation is also presented. Finally, we describe the Prolog interpreter as a Horn clause and resolution-based inferencing system, and see Prolog computing as an instance of the logic programming paradigm.
　　Chapter 15 presents natural language understanding. Our traditional approach to language understanding, exemplified by many of the semantic structures presented in Chapter 7, is complemented with the stochastic approach. These include using Markov models, CART trees, CHART parsing (the Earley algorithm), mutual information clustering, and statistics-based parsing. The chapter concludes with examples applying natural language techniques to database query generation, a text summarization systems well as.the use of machine learning to generalize extracted results from the WWW.
　　Finally, Chapter 16 serves as an epilogue for the book. It addresses the issue of the possibility of a science of intelligent systems, and considers contemporary challenges to AI; it discusses AI's current limitations, and projects its exciting future.
　　Using This Book
　　Artificial intelligence is a big field, and consequently, this is a large book. Although it would require more than a single semester to cover all of the material offered, we have designed our book so that a number of paths may be taken through the material. By selecting subsets of the material, we have used this text for single semester and full year (two semester) courses.
　　We assume that most students will have had introductory courses in discrete mathematics, including predicate calculus, set theory, counting, and graph theory. If this is not true, the instructor should spend more time on these concepts in the "optional" sections at the beginning of the introductory chapters (2.1,3.1, and 5.1). We also assume that students have had courses in data structures including trees, graphs, and recursion-based search, using stacks, queues, and priority queues. If they have not, then spend more time on the beginning sections of Chapters 3, 4, and 6.
　　In a one quarter or one semester course, we go quickly through the first two parts of the book. With this preparation, students are able to appreciate the material in Part III. We then consider the Prolog, Lisp, or .the Java code in the supplementary materials for the book and require students to build many of the representation and search techniques of the second part of the book. Alternatively, one of the languages, Prolog, for example, can be introduced early in the course and be used to test out the data structures and search techniques as that are encountered. We feel the meta-interpreters presented in the language materials are very helpful for building rule-based and other knowledge-intensive problem solvers. Prolog, Lisp, and Java are excellent tools for building natural language understanding and learning systems; these architectures are presented in Parts II and III and there are examples of them in the supplementary course materials.
　　In a two-semester or three-quarter course, we are able to cover the application areas of Parts IV and V, especially the machine learning chapters, in appropriate detail. We also expect a much more detailed programming project from students. We also think that it is very important in the second semester for students to revisit many of the primary sources of the AI literature. It is crucial for students to see both where we are in the evolution of the AI enterprise, as well as how we got here, and to have an appreciation of the future promises of artificial intelligence. We use materials from the WWW for this purpose or select a collected set of readings, such as, Computation and Intelligence (Luger 1995).
　　The algorithms of our book are described using a Pascal-like pseudo-code. This notation uses the control structures of Pascal along with English descriptions of the tests and operations. We have added two useful constructs to the Pascal control structures. The first is a modified case statement that, rather than comparing the value of a variable with constant case labels, as in standard Pascal, lets each item be labeled with an arbitrary boolean test. The case evaluates these tests in order until one of them is true and then performs the associated action; all other actions are ignored. Those familiar with Lisp will note that this has the same semantics as the Lisp cond statement.
　　The other addition to our pseudo code language is a return statement which takes one argument and can appear anywhere within a procedure or function. When the return is encountered, it causes the program to immediately exit the function, returning its argument as a result. Other than these modifications we used Pascal structure, with a reliance on the English descriptions, to make the algorithms clear.
　　Supplemental Material Available
　　The sixth edition has an attached web site maintained by my graduate students. This site, built originally by two UNM students, Alejandro CdeBaca and Cheng Liu, includes supplementary ideas for most chapters, some sample problems with their solutions, and ideas for student projects. Besides the Prolog, Lisp, and Java programs in the supplementary materials for this book, we have included many other Al algorithms in Java and C++ on the web site. Students are welcome to use these and supplement them with their own comments, code, and critiques. The web url is www. cs.unm.edu/～luger/ai-final/.
　　The Prolog, Lisp, and Java programs implementing many of the AI data structures and search algorithms of the book are available through your Addison-Wesley Pearson Education representative. There is also an Instructor's Guide available which has many of the book's exercises worked out, several practice tests with solutions, a sample syllabus, and ideas supporting teaching the material. There are also a full set of PowerPointpresenration materials for use by instructors adopting this book. Again, consult your local A-W Pearson representative for access and visit www. aw. com/luger.
　　My e-mail address is luger@cs.unm.edu, and I enjoy hearing from my readers.
　　Acknowledgements
　　Although I am the sole author of the sixth edition, this book has always been the product of my efforts as Professor of Computer Science, Psychology, and Linguistics at the University of New Mexico along with my fellow faculty, professional colleagues, graduate students, and friends. The sixth edition is also the product of the many readers that have emailed comments, corrections, and suggestions. The book will continue this way, reflecting a community effort; consequently, I will continue using the prepositions we, our, and us when presenting material.
　　I thank Bill Stubblefield, the co-author for the first three editions, for more than twenty years of contributions, but even more importantly, for his friendship. I also thank the many reviewers that have helped develop this and earlier editions. These include Dennis Bahler, Leonardo Bottaci, Skona Brittain, Philip Chan, Peter Collingwood, Mehdi Dastard, John Donald, Sarah Douglas, Christophe Giraud-Carrier, Andrew Kosoresow, Terran Lane, Chris Malcolm, Ray Mooney, Marek Perkowski, Barak Pearmutter, Dan Pless, Bruce Porter, Stuart Shapiro, Julian Richardson, Jude Shavlik, John Sheppard, Carl Stern, Leon van der Torre, Marco Valtorta, and Bob Veroff. We also appreciate the numerous suggestions and comments sent directly by e-mail by readers. Finally, Chris Malcolm, Brendan McGonnigle, and Akasha Tang, critiqued Chapter 16.
　　From our UNM colleagues, we thank Dan Pless, Nikita Sakhanenko, Roshan Rammohan, and Chayan Chakrabarti for their major role in developing materials for Chapters 5, 9, and 13; Joseph Lewis for his efforts on Chapters 9 and 16; Carl Stern for his help in developing Chapter 11 on connectionist learning; Bob Veroff for his critique of the automated reasoning material in Chapter 14; and Jared Saia, Stan Lee, and Paul dePalma for helping with the stochastic approaches to natural language understanding of Chapter 15.
　　We thank Academic Press for permission to reprint much of the material of Chapter 11; this first appeared in the book Cognitive Science: The Science oflntelligent Systems (Luger 1994). Finally, we thank more than two decades of students who have used various versions of this book and software at UNM for their help in expanding our horizons, as well as in removing typos and bugs.
　　We thank our many friends at Benjamin-Cummings, Addison-Wesley-Longman, and Pearson Education for their support and encouragement in completing the writing task of our six editions, especially Alan Apt in helping us with the first edition, Lisa Moller and Mary Tudor for their help on the second, Victoria Henderson, Louise Wilson, and Karen Mosman for their assistance on the third, Keith Mansfield, Karen Sutherland, and Anita Atkinson for support on the fourth, Keith Mansfield, Owen Knight, Anita Atkinson, and Mary Lince for their help on the fifth edition, and Simon Plumtree, Matt Goldstein, Joe Vetere, and Sarah Mihnore for their help on the sixth. Katherine Haratunian of Addison Wesley has had a huge role in seeing that Professors received their Instructor Guides and PowerPoint presentation materials (These are maintained by Addison-Wesley Pearson and available only through your local sales representative). Linda Cicarella of the University of New Mexico helped prepare many of the figures.
　　We thank Thomas Barrow, internationally recognized artist and University of New Mexico Professor of Art (emeritus), who created the photograms for this book.
　　Artificial intelligence is an exciting and rewarding discipline; may you enjoy your study as you come to appreciate its power and challenges. ...
　　George Luger
　　1 January 2008
　　Albuquerque