lecture 14 of 41 first-order logic and theorem proving
TRANSCRIPT
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Lecture 14 of 41Lecture 14 of 41
First-Order Logicand Theorem Proving
Wednesday, 22 September 2004
William H. HsuDepartment of Computing and Information Sciences, KSU
http://www.kddresearch.orghttp://www.cis.ksu.edu/~bhsu
Reading:Sections 8.1-8.3, Russell and Norvig 2e
Review: Chapter 6, R&N 2e
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Lecture OutlineLecture Outline
• Today’s Reading– Chapter 8, Russell and Norvig– Recommended references: Nilsson and Genesereth (excerpt of Chapter 5 online)
• Next Week’s Reading: Chapters 9-10, R&N• Previously: Introduction to Propositional and First-Order Logic
– Last Friday (17 Sep 2004)• FOL agents, issues: frame, ramification, qualification problems• Solutions: situation calculus, circumscription by successor state axioms
– Monday (20 Sep 2004)• First-order logic (FOL): predicates, functions, quantifiers• Sequent rules, proof by refutation
• Today: FOL Knowledge Bases and Theorem Proving– Forward Chaining with And-Introduction, Universal Elimination, Modus Ponens– Ontology, History of Logic, Russell’s Paradox– Unification, Logic Programming Basics
• Next Week: Resolution, Logic Programming, Decidability of SAT
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
InIn--Class Discussion:Class Discussion:Problem Set 2Problem Set 2
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Taking Stock:Taking Stock:FOL InferenceFOL Inference
• Previously: Logical Agents and Calculi• Review: FOL in Practice
– Agent “toy” world: 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• FOL Knowledge Bases• FOL Inference
– Proofs– Pattern-matching: unification– Theorem-proving as search
• Generalized Modus Ponens (GMP)• Forward Chaining and Backward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Automated Deduction (Chapters 8Automated Deduction (Chapters 8--10 R&N)10 R&N)
Adapted from slides by S. Russell, UC Berkeley
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Example ProofExample Proof
• ???• Apply Sequent Rules to Generate New Assertions
• Modus Ponens And Introduction Universal Elimination
Adapted from slides by S. Russell, UC Berkeley
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Search with Primitive Inference RulesSearch with Primitive Inference Rules
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
A Brief History of Reasoning:A Brief History of Reasoning:Chapter 8 End Notes, R&NChapter 8 End Notes, R&N
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
KKnowledge nowledge EEngineeringngineering
• KE: Process of– Choosing logical language (basis of KR)– Building KB– Implementing proof theory– Inferring new facts
• Analogy: Programming Languages / Software Engineering– Choosing programming language (basis of software engineering)– Writing program– Choosing / writing compiler– Running program
• Example Domains– Electronic circuits (Section 8.3 R&N)– Exercise
• Look up, read about protocol analysis• Find example and think about KE process for your project domain
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
OntologyOntology
• Ontology: “What Objects Exist and Are Symbolically Representable?”• Issue: Grouping Objects and Describing Families
– Grouping objects and describing families– Example: sets of sets
• Russell’s paradox: http://plato.stanford.edu/entries/russell-paradox/• (Four) responses: types, formalism, intuitionism, Zermelo-Fraenkel set theory
– Sidebar: natural kinds (p. 232)• Issue: Reasoning About Time
– Modal logics (CIS 301)– Interval logics (Section 8.4 R&N p. 238-241)
• Example Domains– Grocery shopping (Section 8.5 R&N); similar example in Winston 3e– Data models for knowledge discovery in databases (KDD)
• Data dictionaries• See grocery example, especially p. 249 - 252
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Unification:Unification:Definitions and Idea SketchDefinitions and Idea Sketch
Adapted from slides by S. Russell, UC Berkeley
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
GGeneralized eneralized MModus odus PPonensonens
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Soundness of GMPSoundness of GMP
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Forward ChainingForward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Example:Example:Forward ChainingForward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Backward ChainingBackward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Example:Example:Backward ChainingBackward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
• Question: How Does This Relate to Proof by Refutation?• Answer
– Suppose ¬Query, For The Sake Of Contradiction (FTSOC)
– Attempt to prove that KB ∧ ¬Query ⊥
Adapted from slides by S. Russell, UC Berkeley
Review:Review:Backward ChainingBackward Chaining
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Completeness Completeness ReduxRedux
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Completeness in FOLCompleteness in FOL
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Resolution Inference RuleResolution Inference Rule
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Adapted from slides by S. Russell, UC Berkeley
Digression:Digression:Decidability and Formal LanguagesDecidability and Formal Languages
• See: Hopcroft and Ullman 2e, Lewis and Papadimitriou 3e• Formal Languages (See: CIS 540, Other Automata Theory Course)
– Member of Turing hierarchy• Finite state automata: regular languages• Pushdown automata: context-free languages• Linear bounded automata: context-sensitive languages• Turing machines: recursive languages
– Recursive languages• ∃ computational model for decision problem, halts in finite number of steps• REC: set of all recursive languages• Example: finite searches (convert to decision problem of checking solution)• Closed under complementation (consequence?)
– Recursive enumerable but not recursive (RE - REC)– Not recursive (∉ RE)
• What Are FOL-VALID, FOL-NOT-SAT, FOL-SAT, FOL-NOT-VALID?
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
Summary PointsSummary Points
• Applications of Knowledge Bases (KBs) and Inference Systems• “Industrial Strength” KBs
– Building KBs• Knowledge Engineering (KE) and protocol analysis• Inductive Logic Programming (ILP) and other machine learning techniques
– Components• Ontologies• Fact and rule bases
– Using KBs• Systems of Sequent Rules: GMP/AI/UE, Resolution• Methodology of Inference
– Inference as search– Forward and backward chaining– Fan-in, fan-out
Kansas State UniversityDepartment of Computing and Information SciencesCIS 730: Introduction to Artificial Intelligence
TerminologyTerminology
• Logical Languages: WFFs, Quantification• Properties of Knowledge Bases (KBs)
– Satisfiability and validity– Entailment and provability
• Properties of Proof Systems: Soundness and Completeness• Knowledge Bases in Practice
– Knowledge Engineering– Ontologies
• Sequent Rules– (Generalized) Modus Ponens– And-Introduction– Universal-Elimination
• Methodology of Inference– Forward and backward chaining– Fan-in, fan-out (wax on, wax off…)