Figures. Agents in the World: What are Agents and How Can They be Built? 1

Transcription

1 Table of Figures v xv I Agents in the World: What are Agents and How Can They be Built? 1 1 Artificial Intelligence and Agents What is Artificial Intelligence? Artificial and Natural Intelligence A Brief History of Artificial Intelligence Relationship to Other Disciplines Agents Situated in Environments Designing Agents Design Time, Offline and Online Computation Tasks Defining a Solution Representations Agent Design Space Modularity Planning Horizon Representation Computational Limits Learning v

2 Table of vi Uncertainty Preference Number of Agents Interaction Interaction of the Dimensions Prototypical Applications An Autonomous Delivery Robot A Diagnostic Assistant An Intelligent Tutoring System A Trading Agent Smart House Overview of the Book Review References and Further Reading Exercises Agent Architectures and Hierarchical Control Agents Agent Systems The Agent Function Hierarchical Control Acting with Reasoning Agents Modeling the World Knowledge and Acting Design Time and Offline Computation Online Computation Review References and Further Reading Exercises II Reasoning, Planning and Learning with Certainty 75 3 Searching for Solutions Problem Solving as Search State Spaces Graph Searching Formalizing Graph Searching A Generic Searching Algorithm Uninformed Search Strategies Breadth-First Search Depth-First Search Iterative Deepening... 94

3 Table of vii Lowest-Cost-First Search Heuristic Search A Search Designing a Heuristic Function Pruning the Search Space Cycle Pruning Multiple-Path Pruning Summary of Search Strategies More Sophisticated Search Branch and Bound Direction of Search Dynamic Programming Review References and Further Reading Exercises Reasoning with Constraints Possible Worlds, Variables, and Constraints Variables and Worlds Constraints Constraint Satisfaction Problems Generate-and-Test Algorithms Solving CSPs Using Search Consistency Algorithms Domain Splitting Variable Elimination Local Search Iterative Best Improvement Randomized Algorithms Local Search Variants Evaluating Randomized Algorithms Random Restart Population-Based Methods Optimization Systematic Methods for Optimization Local Search for Optimization Review References and Further Reading Exercises Propositions and Inference Propositions Syntax of Propositional Calculus

4 Table of viii Semantics of the Propositional Calculus Propositional Constraints Clausal Form for Consistency Algorithms Exploiting Propositional Structure in Local Search Propositional Definite Clauses Questions and Answers Proofs Knowledge Representation Issues Background Knowledge and Observations Querying the User Knowledge-Level Explanation Knowledge-Level Debugging Proving by Contradiction Horn Clauses Assumables and Conflicts Consistency-Based Diagnosis Reasoning with Assumptions and Horn Clauses Complete Knowledge Assumption Non-monotonic Reasoning Proof Procedures for Negation as Failure Abduction Causal Models Review References and Further Reading Exercises Planning with Certainty Representing States, Actions, and Goals Explicit State-Space Representation The STRIPS Representation Feature-Based Representation of Actions Initial States and Goals Forward Planning Regression Planning Planning as a CSP Action Features Partial-Order Planning Review References and Further Reading Exercises Supervised Machine Learning Learning Issues

5 Table of ix 7.2 Supervised Learning Evaluating Predictions Types of Errors Point Estimates with No Input Features Basic Models for Supervised Learning Learning Decision Trees Linear Regression and Classification Overfitting Pseudocounts Regularization Cross Validation Neural Networks and Deep Learning Composite Models Random Forests Ensemble Learning Case-Based Reasoning Learning as Refining the Hypothesis Space Version-Space Learning Probably Approximately Correct Learning Review References and Further Reading Exercises III Reasoning, Learning and Acting with Uncertainty Reasoning with Uncertainty Probability Semantics of Probability Axioms for Probability Conditional Probability Expected Values Information Independence Belief Networks Observations and Queries Constructing Belief Networks Probabilistic Inference Variable Elimination for Belief Networks Representing Conditional Probabilities and Factors Sequential Probability Models Markov Chains Hidden Markov Models

6 Table of x Algorithms for Monitoring and Smoothing Dynamic Belief Networks Time Granularity Probabilistic Models of Language Stochastic Simulation Sampling from a Single Variable Forward Sampling in Belief Networks Rejection Sampling Likelihood Weighting Importance Sampling Particle Filtering Markov Chain Monte Carlo Review References and Further Reading Exercises Planning with Uncertainty Preferences and Utility Axioms for Rationality Factored Utility Prospect Theory One-Off Decisions Single-Stage Decision Networks Sequential Decisions Decision Networks Policies Variable Elimination for Decision Networks The Value of Information and Control Decision Processes Policies Value Iteration Policy Iteration Dynamic Decision Networks Partially Observable Decision Processes Review References and Further Reading Exercises Learning with Uncertainty Probabilistic Learning Learning Probabilities Probabilistic Classifiers MAP Learning of Decision Trees

7 Table of xi Description Length Unsupervised Learning k-means Expectation Maximization for Soft Clustering Learning Belief Networks Learning the Probabilities Hidden Variables Missing Data Structure Learning General Case of Belief Network Learning Bayesian Learning Review References and Further Reading Exercises Multiagent Systems Multiagent Framework Representations of Games Normal Form Games Extensive Form of a Game Multiagent Decision Networks Computing Strategies with Perfect Information Reasoning with Imperfect Information Computing Nash Equilibria Group Decision Making Mechanism Design Review References and Further Reading Exercises Learning to Act Reinforcement Learning Problem Evolutionary Algorithms Temporal Differences Q-learning Exploration and Exploitation Evaluating Reinforcement Learning Algorithms On-Policy Learning Model-Based Reinforcement Learning Reinforcement Learning with Features SARSA with Linear Function Approximation Multiagent Reinforcement Learning Perfect-Information Games

8 Table of xii Learning to Coordinate Review References and Further Reading Exercises IV Reasoning, Learning and Acting with Individuals and Relations Individuals and Relations Exploiting Relational Structure Symbols and Semantics Datalog: A Relational Rule Language Semantics of Ground Datalog Interpreting Variables Queries with Variables Proofs and Substitutions Instances and Substitutions Bottom-up Procedure with Variables Unification Definite Resolution with Variables Function Symbols Proof Procedures with Function Symbols Applications in Natural Language Using Definite Clauses for Context-Free Grammars Augmenting the Grammar Building Structures for Non-terminals Canned Text Output Enforcing Constraints Building a Natural Language Interface to a Database Limitations Equality Allowing Equality Assertions Unique Names Assumption Complete Knowledge Assumption Complete Knowledge Assumption Proof Procedures Review References and Further Reading Exercises Ontologies and Knowledge-Based Systems Knowledge Sharing Flexible Representations

9 Table of xiii Choosing Individuals and Relations Graphical Representations Classes Ontologies and Knowledge Sharing Uniform Resource Identifiers Description Logic Top-Level Ontologies Implementing Knowledge-Based Systems Base Languages and Metalanguages A Vanilla Meta-Interpreter Expanding the Base Language Depth-Bounded Search Meta-Interpreter to Build Proof Trees Delaying Goals Review References and Further Reading Exercises Relational Planning, Learning, and Probabilistic Reasoning Planning with Individuals and Relations Situation Calculus Event Calculus Relational Learning Structure Learning: Inductive Logic Programming Learning Hidden Properties: Collaborative Filtering Statistical Relational Artificial Intelligence Relational Probabilistic Models Review References and Further Reading Exercises V Retrospect and Prospect Retrospect and Prospect Dimensions of Complexity Revisited Social and Ethical Consequences References and Further Reading Exercises A Mathematical Preliminaries and Notation 745 A.1 Discrete Mathematics A.2 Functions, Factors and Arrays

10 Table of xiv A.3 Relations and the Relational Algebra References 751 Index 773