Similar documents
Fundamentals of Logic

0.1 Equivalence between Natural Deduction and Axiomatic Systems

Notes on Natural Logic

Threshold logic proof systems

2 Deduction in Sentential Logic

Tableau Theorem Prover for Intuitionistic Propositional Logic

Tableau Theorem Prover for Intuitionistic Propositional Logic

Cut-free sequent calculi for algebras with adjoint modalities

SAT and DPLL. Introduction. Preliminaries. Normal forms DPLL. Complexity. Espen H. Lian. DPLL Implementation. Bibliography.

SAT and DPLL. Espen H. Lian. May 4, Ifi, UiO. Espen H. Lian (Ifi, UiO) SAT and DPLL May 4, / 59

Computing Unsatisfiable k-sat Instances with Few Occurrences per Variable

arxiv: v1 [math.lo] 24 Feb 2014

Practical SAT Solving

Computing Unsatisfiable k-sat Instances with Few Occurrences per Variable

Security issues in contract-based computing

Mathematical Logic Final Exam 14th June PROPOSITIONAL LOGIC

A Translation of Intersection and Union Types

A Syntactic Realization Theorem for Justification Logics

Binary Decision Diagrams

Binary Decision Diagrams

5 Deduction in First-Order Logic

Logic and Artificial Intelligence Lecture 24

CIS 500 Software Foundations Fall October. CIS 500, 6 October 1

TR : Knowledge-Based Rational Decisions

Another Variant of 3sat. 3sat. 3sat Is NP-Complete. The Proof (concluded)

An Adaptive Characterization of Signed Systems for Paraconsistent Reasoning

TABLEAU-BASED DECISION PROCEDURES FOR HYBRID LOGIC

Arborescent Architecture for Decentralized Supervisory Control of Discrete Event Systems

Horn-formulas as Types for Structural Resolution

In this lecture, we will use the semantics of our simple language of arithmetic expressions,

A semantics for concurrent permission logic. Stephen Brookes CMU

Implications as rules

Lecture 14: Basic Fixpoint Theorems (cont.)

Semantics with Applications 2b. Structural Operational Semantics

Lecture 2: The Neoclassical Growth Model

δ j 1 (S j S j 1 ) (2.3) j=1

Lecture Notes on Bidirectional Type Checking

Continuous images of closed sets in generalized Baire spaces ESI Workshop: Forcing and Large Cardinals

2. This algorithm does not solve the problem of finding a maximum cardinality set of non-overlapping intervals. Consider the following intervals:

The Subjective and Personalistic Interpretations

Determinacy models and good scales at singular cardinals

Tableau-based Decision Procedures for Hybrid Logic

A Knowledge-Theoretic Approach to Distributed Problem Solving

Lecture 2: The Simple Story of 2-SAT

MAT25 LECTURE 10 NOTES. = a b. > 0, there exists N N such that if n N, then a n a < ɛ

Notes to The Resurrection Axioms

Level by Level Inequivalence, Strong Compactness, and GCH

Algorithmic Game Theory and Applications. Lecture 11: Games of Perfect Information

Econ 424/CFRM 462 Portfolio Risk Budgeting

Another Variant of 3sat

Generalising the weak compactness of ω

Lecture Notes on Type Checking

Introduction to Greedy Algorithms: Huffman Codes

CS792 Notes Henkin Models, Soundness and Completeness

Maximally Consistent Extensions

Strong Subgraph k-connectivity of Digraphs

Partial Fractions. A rational function is a fraction in which both the numerator and denominator are polynomials. For example, f ( x) = 4, g( x) =

Harvard School of Engineering and Applied Sciences CS 152: Programming Languages

10 5 The Binomial Theorem

Definition: A proposition is a sentence that is either true or false, but not both.

The Traveling Salesman Problem. Time Complexity under Nondeterminism. A Nondeterministic Algorithm for tsp (d)

Levin Reduction and Parsimonious Reductions

CATEGORICAL SKEW LATTICES

Automated Policy Combination for Secure Data Sharing in Cross-Organizational Collaborations

Algebra homework 8 Homomorphisms, isomorphisms

Focusing on contraction

Outline Introduction Game Representations Reductions Solution Concepts. Game Theory. Enrico Franchi. May 19, 2010

Continuous distributions. Lecture 6: Probability. Probabilities from continuous distributions. From histograms to continuous distributions

A relative of the approachability ideal, diamond and non-saturation

Part A: Questions on ECN 200D (Rendahl)

Harvard School of Engineering and Applied Sciences CS 152: Programming Languages

THE NUMBER OF UNARY CLONES CONTAINING THE PERMUTATIONS ON AN INFINITE SET

ExpTime Tableau Decision Procedures for Regular Grammar Logics with Converse

You Have an NP-Complete Problem (for Your Thesis)

Failure and Rescue in an Interbank Network

CS 4110 Programming Languages and Logics Lecture #2: Introduction to Semantics. 1 Arithmetic Expressions

Lecture Outline. Scheduling aperiodic jobs (cont d) Scheduling sporadic jobs

8. Propositional Logic Natural deduction - negation. Solved problems

Online Algorithms SS 2013

Inequality, welfare and the progressivity (?) of the UK tax/benefit system. Sam Hinds

Introduction to Game Theory Lecture Note 5: Repeated Games

UNIFORM BOUNDS FOR BLACK SCHOLES IMPLIED VOLATILITY

IDENTIFYING BROAD AND NARROW FINANCIAL RISK FACTORS VIA CONVEX OPTIMIZATION: PART II

Strong normalisation and the typed lambda calculus

Philipp Moritz Lücke

Covering properties of derived models

A Logic-based Approach to Decision Making. Magdalena Ivanovska and Martin Giese

Lecture 37 Sections 11.1, 11.2, Mon, Mar 31, Hampden-Sydney College. Independent Samples: Comparing Means. Robb T. Koether.

Forward Risk Adjusted Probability Measures and Fixed-income Derivatives

Isabelle/FOL First-Order Logic

Introduction An example Cut elimination. Deduction Modulo. Olivier Hermant. Tuesday, December 12, Deduction Modulo

Application of Stochastic Calculus to Price a Quanto Spread

Bilateral trading with incomplete information and Price convergence in a Small Market: The continuous support case

Remarks on Probability

Two Notions of Sub-behaviour for Session-based Client/Server Systems

Cook s Theorem: the First NP-Complete Problem

Satisfaction in outer models

Game Theory: Normal Form Games

Core and Top Trading Cycles in a Market with Indivisible Goods and Externalities

Budget Management In GSP (2018)

Transcription:

1 FUNDAMENTALS OF LOGIC NO.5 SOUNDNESS AND COMPLETENESS Tatsuya Hagino hagino@sfc.keio.ac.jp lecture URL https://vu5.sfc.keio.ac.jp/slide/

2 So Far Propositional Logic Logical Connectives(,,, ) Truth Table Tautology Normal From Disjunctive Normal Form Conjunctive Normal Form Restricting Logical Connectives Proof Axiom and Theorem LK Framework

3 Tautology for Sequent Extend the notion of tautology to sequent. Let Γ be a sequence of formulae A 1,..., A m : Γ = A 1 A m when m > 0 when m = 0 Γ = A 1 A m when m > 0 when m = 0 Sequent Γ Δ is a tautology Γ Δ is a tautology.

4 Soundness and Completeness Γ Δ is a tautology: For any assignment v, v(γ Δ ) = T For any assignment of truth values to propositional variables, we can calculate the truth value of Γ Δ using the truth tables and check whether it is always true. Γ Δ is provable in LK: There is a proof figure of which end sequent is Γ Δ. Starting from initial sequents, we can construct a proof figure of Γ Δ. Soundness Any sequent which is provable in LK is a tautology. Completeness Any tautology is provable in LK.

5 LK Soundness Theorem Theorem: For any sequent Γ Δ, if Γ Δ is provable in LK, Γ Δ is a tautology. Proof: We need to show the following two things: 1. An initial sequent is a tautology. 2. For each inference rule, if all the sequents above are tautologies, the bottom sequent is also a tautology. For (1), an initial sequent A A is A A, and it is a tautology. For (2), we need to check for each inference rule. For example, Γ 1 Δ 1, A B, Γ 2 Δ 2 A B, Γ 1, Γ 2 Δ 1, Δ ( L) 2 we need to show when Γ 1 Δ 1 A and B Γ 2 Δ 2 are tautologies, ( A B Γ 1 Γ 2 ) (Δ 1 Δ 2 ) is also a tautology.

6 cont. For an assignment v if v( A B Γ 1 Γ 2 ) = T, then v(a B) = T, v(γ 1 ) = T, and v(γ 2 ) = T. 1. If v(a) = F, from v(γ 1 Δ 1 A) = T and v(γ 1 ) = T, v(δ 1 ) = T. 2. If v(a) = T, from v(a B) = T, v(b) = T. Since v(b Γ 2 Δ 2 ) = T and v(γ 2 ) = T, v(δ 2 ) = T. In both cases, v(δ 1 Δ 2 ) = T. Therefore, ( A B Δ 2 ) is a tautology Γ 1 Γ 2 ) (Δ 1 For other rules, we can show the bottom sequent is a tautology when the above sequents are tautologies. Therefore LK framework is sound. QED

7 LK Completeness Theorem Theorem: If Γ Δ is a tautology, Γ Δ is provable in LK without using cut inference rule. Proof: First, define the decomposition of a sequent as follows: Γ Δ 1, A B, Δ 2 Γ 1, A B, Γ 2 Δ Γ Δ 1, A B, Δ 2 Γ 1, A B, Γ 2 Δ Γ Δ 1, A, Δ 2 Γ Δ 1, B, Δ 2 Γ 1, A, B, Γ 2 Δ Γ Δ 1, A, B, Δ 2 Γ 1, A, Γ 2 Δ Γ 1, B, Γ 2 Δ Γ Δ 1, A B, Δ 2 Γ 1, A B, Γ 2 Δ Γ Δ 1, A, Δ 2 Γ 1, A, Γ 2 Δ A, Γ Δ 1, B, Δ 2 Γ 1, Γ 2 Δ, A Γ 1, B, Γ 2 Δ A, Γ Δ 1, Δ 2 Γ 1, Γ 2 Δ, A A sequent is decomposed into one or two sequents. Two things hold for any decomposition: The number of logical connectives decreases. If the original sequent is a tautology, the decompositions are tautologies.

8 cont. Apply decomposition repeatedly: Γ Δ Γ 1 Δ 1 Γ 2 Δ 2 Γ 3 Δ 3 Γ 4 Δ 4 Γ 5 Δ 5 Sequents at the bottom row do not contain logical connectives. Complete decomposition tree If Γ Δ is a tautology, the bottom row of the complete decomposition contains tautology sequents containing propositional variables (no logical connectives). A sequent p 1,, p m q 1, q n containing only propositional variables is a tautology if and only if one of p i matches with one of q j. Γ 1, A, Γ 2 Δ 1, A, Δ 2 is provable without cut rule. Apply weakening and exchange rules to initial sequent A A.

9 cont. When Γ Δ is decomposed into Γ 1 Δ 1 and Γ 2 Δ 2, if Γ 1 Δ 1 and Γ 2 Δ 2 are provable without cut, Γ Δ is also provable without cut. For each decomposition, use rules for logical connectives, exchange and contraction. Decomposition Γ 1, A B, Γ 2 Δ Γ 1, Γ 2 Δ, A Γ 1, B, Γ 2 Δ Inference rule Γ 1 Δ 1, A B, Γ 2 Δ 2 A B, Γ 1, Γ 2 Δ 1, Δ 2 ( L) Therefore, if a sequent is a tautology, the bottom row sequents of its complete decomposition are tautologies and they are provable without cut. By traversing back the decomposition, we can see that the original sequent is also provable without cut. On the other hand, if a given sequent is not a tautology, no matter how we may decompose it, there exists a non tautology sequent at the bottom row. LK framework is complete. QED This proof provides an algorithm to check whether a sequent is provable or not.

Example Find a complete decomposition of p q p p and create a proof figure without cut. 10 Decomposition p q p p Proof p q p p

11 Cut Elimination Theorem Theorem: If Γ Δ is provable in LK, there exists a proof figure of Γ Δ without cut. Proof: Since any provable sequent is a tautology, it is provable without cut. QED A proof figure with cut A A A, A A, A A A A B, A A A A, A A B A A A B A A, A A A A B (B A) A A B (B A) A A A A A B (B A) A B (B A) A A A, A A, A A, A B Theorem: Any formula in a cut free proof figure of Γ Δ is a sub formula of Γ Δ.

12 Exercise Find a proof figure of A B B A without cut.

Duality Duality in inference rules: ( L) and ( R) is symmetrical by exchanging left and right of Γ Δ. 13 Γ Δ, A A, Γ Δ ( L) A, Γ Δ Γ Δ, A ( R) Inference rules for and are symmetrical by replacing left and right and and. A, Γ Δ A B, Γ Δ ( L 1 ) Γ Δ, A Γ Δ, A B ( R 1 ) B, Γ Δ A B, Γ Δ ( L 2 ) Γ Δ, B Γ Δ, A B ( R 2 ) A, Γ 1 Δ 1 B, Γ 2 Δ 2 A B, Γ 1, Γ 2 Δ 1, Δ 2 ( L) Γ 1 Δ 1, A Γ 2 Δ 2, B Γ 1, Γ 2 Δ 1, Δ 2, A B ( R) Duality Theorem: Let A and B be formulae without. Let A and B be formulae A and B but replacing and. If A B is provable, B A is also provable.

14 Summary Soundness of LK framework Any sequent which is provable in LK is a tautology. Completeness of LK framework A tautology is provable in LK. There is an algorithm of checking whether a formula is provable or not. Properties of LK framework cut elimination theorem duality theorem