Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 42 Monday, 08 December 2003 William H. Hsu Department of Computing and Information Sciences, KSU Reading: None Final Review: Chapters 1-15, 18-19, 23, 24 R&N (emphasis on 14-15, 18-19) Consciousness Final Review Part 1 of 2

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 1: The Intelligent Agent Framework Artificial Intelligence (AI) –Operational definition: study / development of systems capable of “thought processes” (reasoning, learning, problem solving) –Constructive definition: expressed in artifacts (design and implementation) Intelligent Agents Topics and Methodologies –Knowledge representation Logical Uncertain (probabilistic) Other (rule-based, fuzzy, neural, genetic) –Search –Machine learning –Planning Applications –Problem solving, optimization, scheduling, design –Decision support, data mining –Natural language processing, conversational and information retrieval agents –Pattern recognition and robot vision

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 2: Agents and Problem Solving Agent Frameworks –Reactivity vs. state –From goals to preferences (utilities) Applications and Automation Case Studies –Search: game-playing systems, problem solvers –Planning, design, scheduling systems –Control and optimization systems –Machine learning: pattern recognition, data mining (business decision support) Things to Check Out Online –Resources page: –Yahoo! Group discussions: groups.yahoo.com/group/ksu-cis730-fall2001groups.yahoo.com/group/ksu-cis730-fall2001 –Suggested project topics, resources – posted in YG

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 3: Search and Constraints Today’s Reading: Sections , Russell and Norvig Solving Problems by Searching –Problem solving agents: design, specification, implementation –Specification components Problems – formulating well-defined ones Solutions – requirements, constraints –Measuring performance Formulating Problems as (State Space) Search Example Search Problems –Toy problems: 8-puzzle, 8-queens, cryptarithmetic, toy robot worlds, constraints –Real-world problems: layout, scheduling Data Structures Used in Search Uninformed Search Algorithms: BFS, DFS, Branch-and-Bound Next Class: Informed Search Strategies –State space search handout (Winston) –Search handouts (Ginsberg, Rich and Knight)

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 4: Uninformed Search Algorithms Search –Problem formulation: state space (initial / operator / goal test / cost), graph –State space search approaches Blind (uninformed) Heuristic (informed) Applications –Problem solving Optimization Scheduling Design –Machine learning (hypothesis space search) More Resources Online – –See also (“REFERENCES”) Course Project Guidelines Posted in YG –Part I: format –Part II: writing quality and criteria –Part III: resources and suggested topics

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 5: Heuristic Search Algorithms – Greedy, A* More Heuristic Search –Best-First Search Greedy A/A* –Search as function maximization Problems: ridge; foothill; plateau, jump discontinuity Solutions: macro operators; global optimization Constraint Satisfaction Search Next Class: IDA*, Hill-Climbing, Iterative Improvement –Gradient descent –Global search MCMC: intuition Some examples of state-of-the-art applications Properties and tradeoffs

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 6: More Heuristic Search – A*, Hill-Climbing / SA More Heuristic Search –Best-First Search: A/A* concluded –Iterative improvement Hill-climbing Simulated annealing (SA) –Search as function maximization Problems: ridge; foothill; plateau, jump discontinuity Solutions: macro operators; global optimization (genetic algorithms / SA) Next Class: Constraint Satisfaction Search, Heuristic Search Next Week: Adversarial Search (e.g., Game Tree Search) –Competitive problems –Minimax algorithm

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 7: Constraint Satisfaction Problems Constraint Satisfaction Problems (CSPs) –Problem definition Domain Constraints –Examples: N-queens, cryptarithmetic, etc. Issues to be Covered Later –Knowledge representation: how to express domain, constraints –Relational constraints In classical logic (propositional, predicate, first-order) In uncertain reasoning Solving CSPs –Propositional constraints: satisfiability solver –First-order relational constraints: difficulties – later –Speeding up CSPs: iterative improvement Gradient (hill-climbing) optimization Simulated annealing

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 8: Game Tree Search: Minimax Game Graph Search –Frameworks Two-player versus multi-player Zero-sum versus cooperative Perfect information versus partially-observable (hidden state) –Concepts Utility and representations (e.g., static evaluation function) Reinforcements: possible role for machine learning Game tree: node/move correspondence, search ply Family of Algorithms for Game Trees: Minimax –Propagation of credit –Imperfect decisions –Issues Quiescence Horizon effect –Need for (alpha-beta) pruning

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 9: More Game Tree Search:  - , Expectiminimax Games as Search Problems –Frameworks –Concepts: utility, reinforcements, game trees –Static evaluation under resource limitations Family of Algorithms for Game Trees: Minimax –Static evaluation algorithm To arbitrary ply To fixed ply Sophistications: iterative deepening, pruning –Credit propagation Intuitive concept Basis for simple (delta-rule) learning algorithms State of The Field Uncertainty in Games: Expectiminimax and Other Algorithms

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 10: Logical Agents and Knowledge Representations Logical Agents –Knowledge Bases (KB) –Logic in general Representation languages, syntax Inference systems –Calculi Propositional First-order (FOL, FOPC) Possible Worlds –Entailment –Models IA Toy Worlds –Wumpus world –Blocks world

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 11: Propositional and Predicate Logic Logical Frameworks –Knowledge Bases (KB) –Logic in general: representation languages, syntax, semantics –Propositional logic –First-order logic (FOL, FOPC) –Model theory, domain theory: possible worlds semantics, entailment Normal Forms –Conjunctive Normal Form (CNF) –Disjunctive Normal Form (DNF) –Horn Form Proof Theory and Inference Systems –Sequent calculi: rules of proof theory –Derivability or provability –Properties Soundness (derivability implies entailment) Completeness (entailment implies derivability)

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 12: Foundations of First-Order Logic FOL in Practice –FOL agents –Example: Wumpus World in FOL –Situation calculus –Frame problem and variants (see R&N sidebar) Representational vs. inferential frame problems Qualification problem: “what if?” Ramification problem: “what else?” (side effects) –Successor-state axioms Logical Languages –Propositional logic –Predicates, terms, functions, atoms (atomic sentences / atomic WFFs), WFFs –First-order logic (FOL, FOPC): universal and existential quantification

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 13: First-Order Knowledge Bases Properties of Knowledge Bases (KBs) –Satisfiability and validity –Entailment and provability Properties of Proof Systems: Soundness and Completeness Normal Forms: CNF, DNF, Horn; Clauses vs. Terms Frame, Ramification, Qualification Problems

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 14: Resolution Theorem Proving Resolution Theorem Proving –Conjunctive Normal Form (clausal form) –Inference rule Single-resolvent form General form –Proof procedure: refutation –Decidability properties FOL-SAT FOL-NOT-SAT (language of unsatisfiable sentences; complement of FOL-SAT) FOL-VALID FOL-NOT-VALID Next Class –More Prolog –Implementing unification

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 15: Logic Programming Techniques Properties of Proof Systems (Again) –Soundness and completeness –Decidability, semi-decidability, undecidability Resolution Refutation Satisfiability, Validity Unification –Occurs check –Most General Unifier Prolog: Tricks of The Trade –Demodulation, paramodulation –Unit resolution, set of support, input / linear resolution, subsumption –Indexing (table-based, tree-based)

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 16: Classical Planning Classical Planning –Planning versus search –Problematic approaches to planning Forward chaining Situation calculus –Representation Initial state Goal state / test Operators Efficient Representations –STRIPS axioms Components: preconditions, postconditions (ADD, DELETE lists) Clobbering / threatening –Reactive plans and policies –Markov decision processes

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Lecture 17: Partial-Order Planning Classical Planning Framework –Planning versus search –Representation: initial state, goal state / test, operators STRIPS Operators –Components: preconditions, postconditions (ADD, DELETE lists) –STRIPS and interference Clobbering / threatening Promotion / demotion –Partial-Order Planners (POP systems) Next Week –Hierarchical abstraction planning: ABSTRIPS –Conditional plans –Reactive plans and policies –Markov decision processes

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Adapted from slides by S. Russell, UC Berkeley Lecture 18: STRIPS and ABSTRIPS Classical Planning Framework –Planning versus search –Representation: initial state, goal state / test, operators STRIPS Operators –Components: preconditions, postconditions (ADD, DELETE lists) –STRIPS and interference Clobbering / threatening Promotion / demotion –Partial-Order Planners (POP systems) Next Week –Hierarchical abstraction planning: ABSTRIPS –Conditional plans –Reactive plans and policies –Markov decision processes

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Classical Planning Framework –Planning versus search –Representation: initial state, goal state / test, operators –STRIPS operators –Partial versus total-order: property of plans –Interleaved vs. noninterleaved: property of planners Last Week –Hierarchical abstraction planning: ABSTRIPS –Conditional plans This Week –Monitoring and replanning –Reactive plans and policies Later –Decision theory –Markov decision processes Lecture 19: Reaction and Replanning

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Introduction to Probabilistic Reasoning –Framework: using probabilistic criteria to search H –Probability foundations Definitions: subjectivist, objectivist; Bayesian, frequentist, logicist Kolmogorov axioms Bayes’s Theorem –Definition of conditional (posterior) probability –Product rule Maximum A Posteriori (MAP) and Maximum Likelihood (ML) Hypotheses –Bayes’s Rule and MAP –Uniform priors: allow use of MLE to generate MAP hypotheses –Relation to version spaces, candidate elimination Next Week: Chapter 15, Russell and Norvig –Later: Bayesian learning: MDL, BOC, Gibbs, Simple (Naïve) Bayes –Categorizing text and documents, other applications Lecture 20: Reasoning under Uncertainty

Kansas State University Department of Computing and Information Sciences CIS 730: Introduction to Artificial Intelligence Graphical Models of Probability –Bayesian belief networks (BBNs) aka belief networks aka causal networks –Conditional independence, causal Markovity –Inference and learning using Bayesian networks Representation of distributions: conditional probability tables (CPTs) Learning polytrees (singly-connected BBNs) and tree-structured BBNs (trees) BBN Inference –Type of probabilistic reasoning –Finds answer to query about P(x) - aka QA Learning in BBNs: In Two Weeks –Known structure –Partial observability Lecture 21: Introduction to Bayesian Networks