Register · · Password forgotten? · |
PEARSON SERIES IN ARTIFICIAL INTELLIGENCE
|
![]() |
Home » Books and magazines » Computer literature » Other programs |
DL-List and Torrent activity | |
Size: 33 MB | Registered: 3 years 7 months | Completed: 1 time | |
Seeder not seen: 7 months 7 days |
|
|
Author | Message | |||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Download Office ® Gender: Longevity: 9 years Posts: 2275 |
Artificial Intelligence: A Modern Approach, 4th Global Edition
Год издания: 2022 Автор: Russell Stuart, Norvig Peter Издательство: Pearson ISBN: 978-1-292-40113-3 Серия: PEARSON SERIES IN ARTIFICIAL INTELLIGENCE Язык: Английский Формат: PDF Качество: Издательский макет или текст (eBook) Интерактивное оглавление: Да Количество страниц: 1167 Описание: The most comprehensive, up-to-date introduction to the theory and practice of artificial intelligence. The long-anticipated revision of Artificial Intelligence: A Modern Approach explores the full breadth and depth of the field of artificial intelligence (AI). The 4th Edition brings readers up to date on the latest technologies, presents concepts in a more unified manner, and offers new or expanded coverage of machine learning, deep learning, transfer learning, multiagent systems, robotics, natural language processing, causality, probabilistic programming, privacy, fairness, and safe AI. Author's (Companion) website: http://aima.eecs.berkeley.edu/global-index.html ОглавлениеI Artificial Intelligence1 Introduction 19 1.1 What Is AI? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.2 The Foundations of Artificial Intelligence . . . . . . . . . . . . . . . . . . 23 1.3 The History of Artificial Intelligence . . . . . . . . . . . . . . . . . . . . 35 1.4 The State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 1.5 Risks and Benefits of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 53 2 Intelligent Agents 54 2.1 Agents and Environments . . . . . . . . . . . . . . . . . . . . . . . . . . 54 2.2 Good Behavior: The Concept of Rationality . . . . . . . . . . . . . . . . 57 2.3 The Nature of Environments . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.4 The Structure of Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 78 II Problem-solving 3 Solving Problems by Searching 81 3.1 Problem-Solving Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 3.2 Example Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 3.3 Search Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 3.4 Uninformed Search Strategies . . . . . . . . . . . . . . . . . . . . . . . . 94 3.5 Informed (Heuristic) Search Strategies . . . . . . . . . . . . . . . . . . . 102 3.6 Heuristic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 124 4 Search in Complex Environments 128 4.1 Local Search and Optimization Problems . . . . . . . . . . . . . . . . . . 128 4.2 Local Search in Continuous Spaces . . . . . . . . . . . . . . . . . . . . . 137 4.3 Search with Nondeterministic Actions . . . . . . . . . . . . . . . . . . . 140 4.4 Search in Partially Observable Environments . . . . . . . . . . . . . . . . 144 4.5 Online Search Agents and Unknown Environments . . . . . . . . . . . . 152 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 160 5 Constraint Satisfaction Problems 164 5.1 Defining Constraint Satisfaction Problems . . . . . . . . . . . . . . . . . 164 5.2 Constraint Propagation: Inference in CSPs . . . . . . . . . . . . . . . . . 169 5.3 Backtracking Search for CSPs . . . . . . . . . . . . . . . . . . . . . . . . 175 5.4 Local Search for CSPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 5.5 The Structure of Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 188 6 Adversarial Search and Games 192 6.1 Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 6.2 Optimal Decisions in Games . . . . . . . . . . . . . . . . . . . . . . . . 194 6.3 Heuristic Alpha–Beta Tree Search . . . . . . . . . . . . . . . . . . . . . 202 6.4 Monte Carlo Tree Search . . . . . . . . . . . . . . . . . . . . . . . . . . 207 6.5 Stochastic Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 6.6 Partially Observable Games . . . . . . . . . . . . . . . . . . . . . . . . . 214 6.7 Limitations of Game Search Algorithms . . . . . . . . . . . . . . . . . . 219 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 221 III Knowledge, reasoning, and planning 7 Logical Agents 226 7.1 Knowledge-Based Agents . . . . . . . . . . . . . . . . . . . . . . . . . . 227 7.2 The Wumpus World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 7.3 Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 7.4 Propositional Logic: A Very Simple Logic . . . . . . . . . . . . . . . . . 235 7.5 Propositional Theorem Proving . . . . . . . . . . . . . . . . . . . . . . . 240 7.6 Effective Propositional Model Checking . . . . . . . . . . . . . . . . . . 250 7.7 Agents Based on Propositional Logic . . . . . . . . . . . . . . . . . . . . 255 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 265 8 First-Order Logic 269 8.1 Representation Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . 269 8.2 Syntax and Semantics of First-Order Logic . . . . . . . . . . . . . . . . . 274 8.3 Using First-Order Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . 283 8.4 Knowledge Engineering in First-Order Logic . . . . . . . . . . . . . . . . 289 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 296 9 Inference in First-Order Logic 298 9.1 Propositional vs. First-Order Inference . . . . . . . . . . . . . . . . . . . 298 9.2 Unification and First-Order Inference . . . . . . . . . . . . . . . . . . . . 300 9.3 Forward Chaining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304 9.4 Backward Chaining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 9.5 Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 328 10 Knowledge Representation 332 10.1 Ontological Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . 332 10.2 Categories and Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 10.3 Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 10.4 Mental Objects and Modal Logic . . . . . . . . . . . . . . . . . . . . . . 344 10.5 Reasoning Systems for Categories . . . . . . . . . . . . . . . . . . . . . 347 10.6 Reasoning with Default Information . . . . . . . . . . . . . . . . . . . . 351 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 356 11 Automated Planning 362 11.1 Definition of Classical Planning . . . . . . . . . . . . . . . . . . . . . . . 362 11.2 Algorithms for Classical Planning . . . . . . . . . . . . . . . . . . . . . . 366 11.3 Heuristics for Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . 371 11.4 Hierarchical Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374 11.5 Planning and Acting in Nondeterministic Domains . . . . . . . . . . . . . 383 11.6 Time, Schedules, and Resources . . . . . . . . . . . . . . . . . . . . . . . 392 11.7 Analysis of Planning Approaches . . . . . . . . . . . . . . . . . . . . . . 396 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 398 IV Uncertain knowledge and reasoning 12 Quantifying Uncertainty 403 12.1 Acting under Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 403 12.2 Basic Probability Notation . . . . . . . . . . . . . . . . . . . . . . . . . . 406 12.3 Inference Using Full Joint Distributions . . . . . . . . . . . . . . . . . . . 413 12.4 Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415 12.5 Bayes’ Rule and Its Use . . . . . . . . . . . . . . . . . . . . . . . . . . . 417 12.6 Naive Bayes Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420 12.7 The Wumpus World Revisited . . . . . . . . . . . . . . . . . . . . . . . . 422 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 426 13 Probabilistic Reasoning 430 13.1 Representing Knowledge in an Uncertain Domain . . . . . . . . . . . . . 430 13.2 The Semantics of Bayesian Networks . . . . . . . . . . . . . . . . . . . . 432 13.3 Exact Inference in Bayesian Networks . . . . . . . . . . . . . . . . . . . 445 13.4 Approximate Inference for Bayesian Networks . . . . . . . . . . . . . . . 453 13.5 Causal Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 472 14 Probabilistic Reasoning over Time 479 14.1 Time and Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479 14.2 Inference in Temporal Models . . . . . . . . . . . . . . . . . . . . . . . . 483 14.3 Hidden Markov Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 491 14.4 Kalman Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497 14.5 Dynamic Bayesian Networks . . . . . . . . . . . . . . . . . . . . . . . . 503 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 515 15 Making Simple Decisions 518 15.1 Combining Beliefs and Desires under Uncertainty . . . . . . . . . . . . . 518 15.2 The Basis of Utility Theory . . . . . . . . . . . . . . . . . . . . . . . . . 519 15.3 Utility Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522 15.4 Multiattribute Utility Functions . . . . . . . . . . . . . . . . . . . . . . . 530 15.5 Decision Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534 15.6 The Value of Information . . . . . . . . . . . . . . . . . . . . . . . . . . 537 15.7 Unknown Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 547 16 Making Complex Decisions 552 16.1 Sequential Decision Problems . . . . . . . . . . . . . . . . . . . . . . . . 552 16.2 Algorithms for MDPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562 16.3 Bandit Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571 16.4 Partially Observable MDPs . . . . . . . . . . . . . . . . . . . . . . . . . 578 16.5 Algorithms for Solving POMDPs . . . . . . . . . . . . . . . . . . . . . . 580 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 586 17 Multiagent Decision Making 589 17.1 Properties of Multiagent Environments . . . . . . . . . . . . . . . . . . . 589 17.2 Non-Cooperative Game Theory . . . . . . . . . . . . . . . . . . . . . . . 595 17.3 Cooperative Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 616 17.4 Making Collective Decisions . . . . . . . . . . . . . . . . . . . . . . . . 622 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 636 18 Probabilistic Programming 641 18.1 Relational Probability Models . . . . . . . . . . . . . . . . . . . . . . . . 642 18.2 Open-Universe Probability Models . . . . . . . . . . . . . . . . . . . . . 648 18.3 Keeping Track of a Complex World . . . . . . . . . . . . . . . . . . . . . 655 18.4 Programs as Probability Models . . . . . . . . . . . . . . . . . . . . . . . 660 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 665 V Machine Learning 19 Learning from Examples 669 19.1 Forms of Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669 19.2 Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671 19.3 Learning Decision Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . 675 19.4 Model Selection and Optimization . . . . . . . . . . . . . . . . . . . . . 683 19.5 The Theory of Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . 690 19.6 Linear Regression and Classification . . . . . . . . . . . . . . . . . . . . 694 19.7 Nonparametric Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 704 19.8 Ensemble Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714 19.9 Developing Machine Learning Systems . . . . . . . . . . . . . . . . . . . 722 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 732 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 733 20 Knowledge in Learning 739 20.1 A Logical Formulation of Learning . . . . . . . . . . . . . . . . . . . . . 739 20.2 Knowledge in Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . 747 20.3 Explanation-Based Learning . . . . . . . . . . . . . . . . . . . . . . . . 750 20.4 Learning Using Relevance Information . . . . . . . . . . . . . . . . . . . 754 20.5 Inductive Logic Programming . . . . . . . . . . . . . . . . . . . . . . . . 758 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 768 21 Learning Probabilistic Models 772 21.1 Statistical Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 772 21.2 Learning with Complete Data . . . . . . . . . . . . . . . . . . . . . . . . 775 21.3 Learning with Hidden Variables: The EM Algorithm . . . . . . . . . . . . 788 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 797 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 798 22 Deep Learning 801 22.1 Simple Feedforward Networks . . . . . . . . . . . . . . . . . . . . . . . 802 22.2 Computation Graphs for Deep Learning . . . . . . . . . . . . . . . . . . 807 22.3 Convolutional Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . 811 22.4 Learning Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 816 22.5 Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 819 22.6 Recurrent Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . 823 22.7 Unsupervised Learning and Transfer Learning . . . . . . . . . . . . . . . 826 22.8 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 833 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 835 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 836 23 Reinforcement Learning 840 23.1 Learning from Rewards . . . . . . . . . . . . . . . . . . . . . . . . . . . 840 23.2 Passive Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . 842 23.3 Active Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . 848 23.4 Generalization in Reinforcement Learning . . . . . . . . . . . . . . . . . 854 23.5 Policy Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 861 23.6 Apprenticeship and Inverse Reinforcement Learning . . . . . . . . . . . . 863 23.7 Applications of Reinforcement Learning . . . . . . . . . . . . . . . . . . 866 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 869 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 870 VI Communicating, perceiving, and acting 24 Natural Language Processing 874 24.1 Language Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874 24.2 Grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 884 24.3 Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 886 24.4 Augmented Grammars . . . . . . . . . . . . . . . . . . . . . . . . . . . . 892 24.5 Complications of Real Natural Language . . . . . . . . . . . . . . . . . . 896 24.6 Natural Language Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . 900 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 902 25 Deep Learning for Natural Language Processing 907 25.1 Word Embeddings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 907 25.2 Recurrent Neural Networks for NLP . . . . . . . . . . . . . . . . . . . . 911 25.3 Sequence-to-Sequence Models . . . . . . . . . . . . . . . . . . . . . . . 915 25.4 The Transformer Architecture . . . . . . . . . . . . . . . . . . . . . . . . 919 25.5 Pretraining and Transfer Learning . . . . . . . . . . . . . . . . . . . . . . 922 25.6 State of the art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 929 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 929 26 Robotics 932 26.1 Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 932 26.2 Robot Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 933 26.3 What kind of problem is robotics solving? . . . . . . . . . . . . . . . . . 937 26.4 Robotic Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 938 26.5 Planning and Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 945 26.6 Planning Uncertain Movements . . . . . . . . . . . . . . . . . . . . . . . 963 26.7 Reinforcement Learning in Robotics . . . . . . . . . . . . . . . . . . . . 965 26.8 Humans and Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 968 26.9 Alternative Robotic Frameworks . . . . . . . . . . . . . . . . . . . . . . 975 26.10 Application Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 978 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 981 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 982 27 Computer Vision 988 27.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 988 27.2 Image Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 989 27.3 Simple Image Features . . . . . . . . . . . . . . . . . . . . . . . . . . . 995 27.4 Classifying Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1002 27.5 Detecting Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1006 27.6 The 3D World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1008 27.7 Using Computer Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . 1013 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1026 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 1027 VII Conclusions 28 Philosophy, Ethics, and Safety of AI 1032 28.1 The Limits of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1032 28.2 Can Machines Really Think? . . . . . . . . . . . . . . . . . . . . . . . . 1035 28.3 The Ethics of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1037 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1056 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 1057 29 The Future of AI 1063 29.1 AI Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063 29.2 AI Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1069 A Mathematical Background 1074 A.1 Complexity Analysis and O() Notation . . . . . . . . . . . . . . . . . . . 1074 A.2 Vectors, Matrices, and Linear Algebra . . . . . . . . . . . . . . . . . . . 1076 A.3 Probability Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 1078 Bibliographical and Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . 1080 B Notes on Languages and Algorithms 1081 B.1 Defining Languages with Backus–Naur Form (BNF) . . . . . . . . . . . . 1081 B.2 Describing Algorithms with Pseudocode . . . . . . . . . . . . . . . . . . 1082 B.3 Online Supplemental Material . . . . . . . . . . . . . . . . . . . . . . . . 1083 Bibliography 1084 Index 1119
|
|||||||||||||||||||||
![]() |
Home » Books and magazines » Computer literature » Other programs |
Current time is: 24-Feb 14:25
All times are UTC + 2
You cannot post new topics in this forum
You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You can download files in this forum |