Artificial Intelligence: A Modern Approach, 4/e (美國原版) (人工智慧：現代方法（第四版）)

《人工智慧：現代方法》是一本最全面、最新的人工智慧理論與實踐介紹。這本期待已久的第四版書籍探索了人工智慧領域的廣度和深度。新版更新了最新技術，以更統一的方式呈現概念，並增加了機器學習、深度學習、遷移學習、多智能體系統、機器人學、自然語言處理、因果關係、概率編程、隱私、公平性和安全人工智慧等內容的涵蓋範圍。

特點：
- 提供最全面、最新的人工智慧理論與實踐介紹。
- 非技術性的學習材料以直觀的解釋介紹主要概念，再進入數學或算法細節。非技術性的語言使得這本書更容易被更廣泛的讀者理解。
- 統一的人工智慧方法向學生展示了各個子領域如何結合起來構建實際有用的程式。
- 更新了對AI系統基本定義的概括，消除了對智能代理人的標準假設，即目標是固定且已知的；相反，代理人可能對其操作的人類真正目標感到不確定。
- 深入涵蓋了基礎和高級主題，使學生在不降低複雜性和深度的情況下對人工智慧的前沿有基本的理解。
- 作者維護的網站（http://aima.cs.berkeley.edu/）包含與書籍相關的評論和討論、練習題、在線程式碼庫、教學資源等。
- 面向學生的互動式練習現在可以在網站上進行，以實現持續更新和添加。
- 在線軟體提供了更多完成項目的機會，包括書中算法的實現，以及Python、Java和JavaScript的補充編碼示例和應用。
- 新增的教學視頻教程加深了學生的參與度，使關鍵概念更加生動。
- 彈性的格式使得這本書適應不同教師的偏好。

保持與最新技術同步，以更統一的方式呈現概念。
- 新增章節擴展了對概率編程、多智能體決策、深度學習以及自然語言處理的涵蓋範圍。
- 增加了對機器學習的涵蓋範圍。
- 更新了對機器人學的材料，包括與人類互動的機器人和將強化學習應用於機器人學。
- 新增了由Judea Pearl撰寫的因果關係部分。
- 新增了關於遊戲和機器人學的蒙特卡洛搜索部分。
- 新增了關於深度學習和自然語言處理的遷移學習部分。
- 新增了隱私、公平性、未來工作和安全人工智慧的部分。
- 廣泛涵蓋了人工智慧應用的最新進展。
- 修訂了對計算機視覺、自然語言處理和機器學習的涵蓋範圍。

Stuart Russell was born in 1962 in Portsmouth, England. He received his B.A. with first-class honours in physics from Oxford University in 1982, and his Ph.D. in computer science from Stanford in 1986. He then joined the faculty of the University of California at Berkeley, where he is a professor and former chair of computer science, director of the Center for Human-Compatible AI, and holder of the Smith–Zadeh Chair in Engineering. In 1990, he received the Presidential Young Investigator Award of the National Science Foundation, and in 1995 he was co-winner of the Computers and Thought Award. He is a Fellow of the American Association for Artificial Intelligence, the Association for Computing Machinery, and the American Association for the Advancement of Science, and Honorary Fellow of Wadham College, Oxford, and an Andrew Carnegie Fellow. He held the Chaire Blaise Pascal in Paris from 2012 to 2014. He has published over 300 papers on a wide range of topics in artificial intelligence. His other books include: The Use of Knowledge in Analogy and Induction, Do the Right Thing: Studies in Limited Rationality (with Eric Wefald), and Human Compatible: Artificial Intelligence and the Problem of Control.

Peter Norvig is currently Director of Research at Google, Inc., and was the director responsible for the core Web search algorithms from 2002 to 2005. He is a Fellow of the American Association for Artificial Intelligence and the Association for Computing Machinery. Previously, he was head of the Computational Sciences Division at NASA Ames Research Center, where he oversaw NASA’s research and development in artificial intelligence and robotics, and chief scientist at Junglee, where he helped develop one of the first Internet information extraction services. He received a B.S. in applied mathematics from Brown University and a Ph.D. in computer science from the University of California at Berkeley. He received the Distinguished Alumni and Engineering Innovation awards from Berkeley and the Exceptional Achievement Medal from NASA. He has been a professor at the University of Southern California and a research faculty member at Berkeley. His other books are: Paradigms of AI Programming: Case Studies in Common Lisp, Verbmobil: A Translation System for Face-to-Face Dialog, and Intelligent Help Systems for UNIX.

The two authors shared the inaugural AAAI/EAAI Outstanding Educator award in 2016.

Stuart Russell於1962年出生於英國朴茨茅斯。他於1982年獲得牛津大學物理學一等榮譽學士學位，並於1986年獲得斯坦福大學計算機科學博士學位。之後，他加入了加州大學伯克利分校的教職，擔任計算機科學教授和前任系主任，並擔任人類相容人工智能中心主任和史密斯-扎德工程講座教授。1990年，他獲得了美國國家科學基金會的總統青年研究員獎，並於1995年獲得了計算機與思維獎的共同獲獎者。他是美國人工智能協會、計算機協會和美國科學促進協會的會士，牛津大學沃德姆學院的名譽會士，以及安德魯·卡內基研究員。他於2012年至2014年擔任巴黎布萊斯·帕斯卡講座。他在人工智能領域發表了300多篇論文。他的其他著作包括《在類比和歸納中使用知識》、《做正確的事情：有限理性研究》（與埃里克·韋法爾德合著）和《人類相容：人工智能與控制問題》。

Peter Norvig目前擔任Google公司的研究總監，並在2002年至2005年期間負責核心網絡搜索算法的研發。他是美國人工智能協會和計算機協會的會士。此前，他曾擔任NASA艾姆斯研究中心計算科學部門主任，負責監督NASA在人工智能和機器人領域的研究和開發，並擔任Junglee的首席科學家，協助開發了最早的互聯網信息提取服務之一。他獲得了布朗大學應用數學學士學位和加州大學伯克利分校計算機科學博士學位。他獲得了伯克利的傑出校友和工程創新獎，以及NASA的傑出成就獎章。他曾任職於南加州大學教授和伯克利研究教職成員。他的其他著作包括《AI編程範例：Common Lisp案例研究》、《Verbmobil：面對面對話翻譯系統》和《UNIX智能幫助系統》。

這兩位作者於2016年共同獲得了AAAI/EAAI傑出教育家獎。

Part I: Artificial Intelligence
1. Introduction
    1.1  What Is AI?
    1.2  The Foundations of Artificial Intelligence
    1.3  The History of Artificial Intelligence
    1.4  The State of the Art
    1.5  Risks and Benefits of AI
2. Intelligent Agents
    2.1  Agents and Environments
    2.2  Good Behavior: The Concept of Rationality
    2.3  The Nature of Environments
    2.4  The Structure of Agents
 
Part II: Problem Solving
3. Solving Problems by Searching
    3.1  Problem-Solving Agents
    3.2  Example Problems
    3.3  Search Algorithms
    3.4  Uninformed Search Strategies
    3.5  Informed (Heuristic) Search Strategies
    3.6  Heuristic Functions
4. Search in Complex Environments
    4.1  Local Search and Optimization Problems
    4.2  Local Search in Continuous Spaces
    4.3  Search with Nondeterministic Actions
    4.4  Search in Partially Observable Environments
    4.5  Online Search Agents and Unknown Environments
5. Adversarial Search and Games
    5.1  Game Theory
    5.2  Optimal Decisions in Games
    5.3  Heuristic Alpha--Beta Tree Search
    5.4  Monte Carlo Tree Search
    5.5  Stochastic Games
    5.6  Partially Observable Games
    5.7  Limitations of Game Search Algorithms
6. Constraint Satisfaction Problems
    6.1  Defining Constraint Satisfaction Problems
    6.2  Constraint Propagation: Inference in CSPs
    6.3  Backtracking Search for CSPs
    6.4  Local Search for CSPs
    6.5  The Structure of Problems
 
Part III: Knowledge and Reasoning
7. Logical Agents
    7.1  Knowledge-Based Agents
    7.2  The Wumpus World
    7.3  Logic
    7.4  Propositional Logic: A Very Simple Logic
    7.5  Propositional Theorem Proving
    7.6  Effective Propositional Model Checking
    7.7  Agents Based on Propositional Logic
8. First-Order Logic
    8.1  Representation Revisited
    8.2  Syntax and Semantics of First-Order Logic
    8.3  Using First-Order Logic
    8.4  Knowledge Engineering in First-Order Logic
9. Inference in First-Order Logic
    9.1  Propositional vs.~First-Order Inference
    9.2  Unification and First-Order Inference
    9.3  Forward Chaining
    9.4  Backward Chaining
    9.5  Resolution
10. Knowledge Representation
    10.1  Ontological Engineering
    10.2  Categories and Objects
    10.3  Events
    10.4  Mental Objects and Modal Logic
    10.5  Reasoning Systems for Categories
    10.6  Reasoning with Default Information
11. Automated Planning
    11.1  Definition of Classical Planning
    11.2  Algorithms for Classical Planning
    11.3  Heuristics for Planning
    11.4  Hierarchical Planning
    11.5  Planning and Acting in Nondeterministic Domains
    11.6  Time, Schedules, and Resources
    11.7  Analysis of Planning Approaches
12. Quantifying Uncertainty
    12.1  Acting under Uncertainty
    12.2  Basic Probability Notation
    12.3  Inference Using Full Joint Distributions
    12.4  Independence
    12.5  Bayes' Rule and Its Use
    12.6  Naive Bayes Models
    12.7  The Wumpus World Revisited
 
Part IV: Uncertain Knowledge and Reasoning
13. Probabilistic Reasoning
    13.1  Representing Knowledge in an Uncertain Domain
    13.2  The Semantics of Bayesian Networks
    13.3  Exact Inference in Bayesian Networks
    13.4  Approximate Inference for Bayesian Networks
    13.5  Causal Networks
14. Probabilistic Reasoning over Time
    14.1  Time and Uncertainty
    14.2  Inference in Temporal Models
    14.3  Hidden Markov Models
    14.4  Kalman Filters
    14.5  Dynamic Bayesian Networks
15. Probabilistic Programming
    15.1  Relational Probability Models
    15.2  Open-Universe Probability Models
    15.3  Keeping Track of a Complex World
    15.4  Programs as Probability Models
16. Making Simple Decisions
    16.1  Combining Beliefs and Desires under Uncertainty
    16.2  The Basis of Utility Theory
    16.3  Utility Functions
    16.4  Multiattribute Utility Functions
    16.5  Decision Networks
    16.6  The Value of Information
    16.7  Unknown Preferences
17. Making Complex Decisions
    17.1  Sequential Decision Problems
    17.2  Algorithms for MDPs
    17.3  Bandit Problems
    17.4  Partially Observable MDPs
    17.5  Algorithms for solving POMDPs
 
Part V: Learning
18. Multiagent Decision Making
    18.1  Properties of Multiagent Environments
    18.2  Non-Cooperative Game Theory
    18.3  Cooperative Game Theory
    18.4  Making Collective Decisions
19. Learning from Examples
    19.1  Forms of Learning
    19.2  Supervised Learning
    19.3  Learning Decision Trees
    19.4  Model Selection and Optimization
    19.5  The Theory of Learning
    19.6  Linear Regression and Classification
    19.7  Nonparametric Models
    19.8  Ensemble Learning
    19.9  Developing Machine Learning Systems
20. Learning Probabilistic Models
    20.1  Statistical Learning
    20.2  Learning with Complete Data
    20.3  Learning with Hidden Variables: The EM Algorithm
21. Deep Learning
    21.1  Simple Feedforward Networks
    21.2  Mixing and matching models, loss functions and optimizers
    21.3  Loss functions
    21.4  Models
    21.5  Optimization Algorithms
    21.6  Generalization
    21.7  Recurrent neural networks
    21.8  Unsupervised, semi-supervised and transfer learning
    21.9  Applications
 
Part VI: Communicating, Perceiving, and Acting
22. Reinforcement Learning
    22.1  Learning from Rewards
    22.2  Passive Reinforcement Learning
    22.3  Active Reinforcement Learning
    22.4  Safe Exploration
    22.5  Generalization in Reinforcement Learning
    22.6  Policy Search
    22.7  Applications of Reinforcement Learning
23. Natural Language Processing
    23.1  Language Models
    23.2  Grammar
    23.3  Parsing
    23.4  Augmented Grammars
    23.5  Complications of Real Natural Language
    23.6  Natural Language Tasks
24. Deep Learning for Natural Language Processing
    24.1  Limitations of Feature-Based NLP Models
    24.2  Word Embeddings
    24.3  Recurrent Neural Networks
    24.4  Sequence-to-sequence Models
    24.5  The Transformer Architecture
    24.6  Pretraining and Transfer Learning
    24.7  Introduction
    24.8  Image Formation
    24.9  Simple Image Features
    24.10 Classifying Images
    24.11 Detecting Objects
    24.12 The 3D World
    24.13 Using Computer Vision
25. Robotics
    25.1  Robots
    25.2  Robot Hardware
    25.3  What kind of problem is robotics solving?
    25.4  Robotic Perception
    25.5  Planning and Control
    25.6  Planning Uncertain Movements
    25.7  Reinforcement Learning in Robotics
    25.8  Humans and Robots
    25.9  Alternative Robotic Frameworks
    25.10 Application Domains
 
Part VII: Conclusions
26. Philosophy and Ethics of AI
    26.1  Weak AI: What are the Limits of AI?
    26.2  Strong AI: Can Machines Really Think?
    26.3  The Ethics of AI
27. The Future of AI
    27.1  AI Components
    27.2  AI Architectures
 

Appendix A: Mathematical Background
    A.1  Complexity Analysis and O() Notation
    A.2  Vectors, Matrices, and Linear Algebra
    A.3  Probability Distributions
Appendix B: Notes on Languages and Algorithms
    B.1  Defining Languages with Backus--Naur Form (BNF)
    B.2  Describing Algorithms with Pseudocode
    B.3  Online Supplemental Material

第一部分：人工智慧
1. 簡介
1.1 什麼是人工智慧？
1.2 人工智慧的基礎
1.3 人工智慧的歷史
1.4 目前的技術水平
1.5 人工智慧的風險與益處
2. 智能代理
2.1 代理與環境
2.2 良好行為：理性的概念
2.3 環境的性質
2.4 代理的結構

第二部分：問題解決
3. 通過搜索解決問題
3.1 解決問題的代理
3.2 示例問題
3.3 搜索算法
3.4 無信息搜索策略
3.5 有信息（啟發式）搜索策略
3.6 啟發函數
4. 在複雜環境中的搜索
4.1 局部搜索和優化問題
4.2 在連續空間中的局部搜索
4.3 具有非確定性動作的搜索
4.4 在部分可觀察環境中的搜索
4.5 在線搜索代理和未知環境

第三部分：知識和推理
7. 邏輯代理
7.1 基於知識的代理
7.2 Wumpus世界
7.3 邏輯
7.4 命題邏輯：一種非常簡單的邏輯
7.5 命題定理證明
7.6 有效的命題模型檢查
7.7 基於命題邏輯的代理
8. 一階邏輯
8.1 重新思考表示
8.2 一階邏輯的語法和語義
8.3 使用一階邏輯
8.4 一階邏輯中的知識工程
9. 一階邏輯推理
9.1 命題邏輯與一階邏輯推理
9.2 合一和一階邏輯推理
9.3 正向鏈接
9.4 反向鏈接
9.5 解析
10. 知識表示
10.1 本體工程
10.2 類別和對象
10.3 事件
10.4 心智對象和模態邏輯
10.5 類別的推理系統