Automated Deduction in Frege-Hilbert Calculi Elmar Eder University - - PowerPoint PPT Presentation

automated deduction in frege hilbert calculi
SMART_READER_LITE
LIVE PREVIEW

Automated Deduction in Frege-Hilbert Calculi Elmar Eder University - - PowerPoint PPT Presentation

Introduction Proof by Composition of Rules Conclusion Automated Deduction in Frege-Hilbert Calculi Elmar Eder University of Salzburg Department of Computer Sciences 8th WSEAS International Conference on Applied Computer and Applied


slide-1
SLIDE 1

Introduction Proof by Composition of Rules Conclusion

Automated Deduction in Frege-Hilbert Calculi

Elmar Eder

University of Salzburg Department of Computer Sciences

8th WSEAS International Conference on Applied Computer and Applied Computational Science (ACACOS ’09) Hangzhou, May 20–22, 2009 Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-2
SLIDE 2

Introduction Proof by Composition of Rules Conclusion

Outline

1

Introduction

2

Proof by Composition of Rules

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-3
SLIDE 3

Introduction Proof by Composition of Rules Conclusion

Outline

1

Introduction

2

Proof by Composition of Rules

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-4
SLIDE 4

Introduction Proof by Composition of Rules Conclusion

Motivation

Main Objectives Automated deduction in classical first order logic Find short proofs Find them quickly Use a calculus which can express powerful proof principles

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-5
SLIDE 5

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculi

Gottlob Frege 1879: Begriffsschrift (concept language) Axioms A → B → A (A → B → C) → (A → B) → A → C (¬A → ¬B) → B → A ∀xF → F x

t

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-6
SLIDE 6

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculi

Rules A A → B B modus ponens A → F x

p

A → ∀xF Constraint: The parameter p must not occur in the conclusion A → ∀xF.

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-7
SLIDE 7

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculi

A Proof Tree

P → P P → Q → P (P → Q → P) → P → P P → (Q → P) → P (P → (Q → P) → P) → (P → Q → P) → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-8
SLIDE 8

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculi

difficult to use for automated deduction Very inefficient to construct a proof by forward reasoning Better: backward reasoning But backward application of modus ponens not unique A A → B B have to guess the cut formula A cut rule

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-9
SLIDE 9

Introduction Proof by Composition of Rules Conclusion

Sequent Calculus

Gerhard Gentzen 1935 Sequents of formulas instead of formulas Cut elimination theorem Sequent calculus without cut rule allows analytic backward reasoning efficient proof search

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-10
SLIDE 10

Introduction Proof by Composition of Rules Conclusion

Conventional Automated Deduction

Proof procedure Construct the formulas of a proof one by one. Calculi used J.A. Robinson 1965 Resolution Calculi based on backward reasoning in the cut-free sequent calculus

Gerhard Gentzen 1935 Sequent calculus without cut-rule Wolfgang Bibel 1982 Connection method Evert W. Beth, Raymond M. Smullyan 1955–1971 Tableau calculus

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-11
SLIDE 11

Introduction Proof by Composition of Rules Conclusion

Cost of Cut Elimination

Theorem (R. Statman 1979, V.P . Orevkov 1982) There is a sequence (Fn) of formulas and a polynomial p such that each Fn has a proof of length ≤ p(n) in the full sequent calculus, but the shortest proof of Fn in the cut-free sequent calculus has length ≥ 22...22

n times

.

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-12
SLIDE 12

Introduction Proof by Composition of Rules Conclusion

Case Distinction

in a proof Case 1: Assume A. <Proof of B> Case 2: Assume ¬A. <Proof of B> is a cut rule and equivalent to the general cut rule. Cut and case distinction are essential parts of human reasoning.

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-13
SLIDE 13

Introduction Proof by Composition of Rules Conclusion

Outline

1

Introduction

2

Proof by Composition of Rules

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-14
SLIDE 14

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

Calculus with cut rule (Frege-Hilbert calculus) Construction of the proof Construct formula schemes and compositions of rules Later instantiate them to obtain the final proof. Inference rule Composition of rules of the Frege-Hilbert calculus

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-15
SLIDE 15

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

Calculus with cut rule (Frege-Hilbert calculus) Construction of the proof Construct formula schemes and compositions of rules Later instantiate them to obtain the final proof. Inference rule Composition of rules of the Frege-Hilbert calculus

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-16
SLIDE 16

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

Calculus with cut rule (Frege-Hilbert calculus) Construction of the proof Construct formula schemes and compositions of rules Later instantiate them to obtain the final proof. Inference rule Composition of rules of the Frege-Hilbert calculus

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-17
SLIDE 17

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

P → P P → C → P (P → C → P) → P → P P → (C → P) → P (P → (C → P) → P) → (P → C → P) → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-18
SLIDE 18

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

P → P A A → P → P P → (C → P) → P (P → (C → P) → P) → (P → C → P) → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-19
SLIDE 19

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

P → P A A → P → P B B → A → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-20
SLIDE 20

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

P → P P → D P → ( D ) → P → P P → D → P (P → D → P) → (P → D ) → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-21
SLIDE 21

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

P → P P → C → P (P → C → P) → P → P P → (C → P) → P (P → (C → P) → P) → (P → C → P) → P → P

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-22
SLIDE 22

Introduction Proof by Composition of Rules Conclusion

Our Proof Procedure

A, B, C are meta-symbols standing for formulas formula schemes composition of rules reasoning in arbitrary directions

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-23
SLIDE 23

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

F1 . . . Fm Gk G1 . . . Gk−1 Gk Gk+1 . . . Gn H

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-24
SLIDE 24

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

A A → B B B B → C C results in the rule A A → B B → C C

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-25
SLIDE 25

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

A proof using the rules ∀xF → F x

t

and A → F x

p

A → ∀xF ∀xP(x) → P(p) ∀xP(x) → ∀yP(y).

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-26
SLIDE 26

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

∀xF → F x

t

A → F x

p

A → ∀xF with the constraint p / ∈ AF

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-27
SLIDE 27

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

∀xF → F x

t

A → Gy

p

A → ∀yG with the constraint p / ∈ AG

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-28
SLIDE 28

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

∀xF → F x

t

∀xF → Gy

p

∀xF → ∀yG with the constraints F x

t = Gy p and p /

∈ FG

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-29
SLIDE 29

Introduction Proof by Composition of Rules Conclusion

Composition of Rules

∀xF → ∀yG with the constraints F x

t = Gy p and p /

∈ FG Also yields a proof of ∀xP(x) → ∀yP(f(y))

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-30
SLIDE 30

Introduction Proof by Composition of Rules Conclusion

The Most General Form of a Rule

Φ1 . . . Φn Ψ with a sequence C of constraints

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-31
SLIDE 31

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculus

Axioms

A → B → A (A → B → C) → (A → B) → A → C (¬A → ¬B) → B → A ∀xF → E with constraint E = F x

t .

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-32
SLIDE 32

Introduction Proof by Composition of Rules Conclusion

Frege-Hilbert Calculus

Rules

A A → B B A → E A → ∀xF with constraints E = F x

p,

p / ∈ AF.

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-33
SLIDE 33

Introduction Proof by Composition of Rules Conclusion

Rules

Premises and conclusion are terms of a free term algebra. Composition by standardizing apart and unification Merge constraints

Elmar Eder Automated Deduction in Frege-Hilbert Calculi

slide-34
SLIDE 34

Introduction Proof by Composition of Rules Conclusion

Conclusion

Automated deduction feasible in calculi with cut Short proofs Expressiveness A student at our department is implementing a system. Future: sequent calculus

Elmar Eder Automated Deduction in Frege-Hilbert Calculi