Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we...
Transcript of Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we...
![Page 1: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/1.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 1
Mathematical Logic 2016
Lecture 5: Proof methods - Tableaux and Resolution
Instructor: Ashutosh Gupta
TIFR, India
Compile date: 2016-08-21
![Page 2: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/2.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 2
Where are we and where are we going?
We have seen
I propositional logic syntax and semantics
I truth tables for deciding SAT
I normal forms
We will see
I proof methods - tableaux and resolution
I implementation issues in resolution
![Page 3: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/3.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 3
Topic 5.1
Proof Methods
![Page 4: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/4.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 4
Proof methods for entailment
Consider a (in)finite set of formulas Σ and a formula F .
A proof method establishes if the following proof query holds.
Σ |= F
![Page 5: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/5.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 5
Refutation proof method
We will present refutation proof methods that only apply to the followingqueries.
Σ |= ⊥.This is not a restriction.
For queries Σ |= F , we pass the following input to the refutation methods.
Σ ∪ {¬F} |= ⊥
![Page 6: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/6.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 6
Two proof systems
We will present the following two refutation proof methods.
I Tableaux
I Resolution
![Page 7: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/7.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 7
Some additional notation on formulas
We assume that >, ⊥, ⇔ and ⊕ are removed using the following equivalencesI p ⇔ q ≡ (p ∧ q) ∨ (p ∧ q)
I p ⊕ q ≡ (¬p ∧ q) ∨ (p ∧ ¬q)
I > ≡ (p ∧ ¬p) for some p ∈ Vars
I ⊥ ≡ (p ∨ ¬p) for some p ∈ Vars
In order to avoid writing many cases we will use a uniform notation.
Conjunctive Disjunctive
α α1 α2 β β1 β2(F ∧ G ) F G ¬(F ∧ G ) ¬F ¬G¬(F ∨ G ) ¬F ¬G (F ∨ G ) F G¬(F ⇒ G ) F ¬G (F ⇒ G ) ¬F G
A non-literal formula can be from one of the following three types.
I α
I β
I ¬¬F
We can further reduced the number of cases by removing ⇒ andapplying NNF transformation. However, ⇒ helps in human read-ability and NNF destroys the high level structure.
![Page 8: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/8.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 8
Game of symbol pushing
A proof method is essentially a game of symbol pushing with an objective.
In the end, we need to show that the game has a meaning and connects tothe semantics.
![Page 9: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/9.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 9
Topic 5.2
Tableaux
![Page 10: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/10.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 10
Tableaux proof method
Tableaux proof method
I takes a set of formula Σ as input and
I produces a finite labelled tree called tableaux as a proof.
Tableaux
I The tree is labelled with formulas.
I Branching in the tree represents the cases in proofs.
I The goal of the method is to find two nodes in each branch of the treesuch that their labels are F and ¬F for some formula F .
![Page 11: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/11.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 11
Tableaux
Definition 5.1A tableaux T for a set of formulas Σ is a finite labelled tree that is initiallyempty and expanded according to the following tableaux expansion rules.
1. F ∈ Σ labelled node is added as a child to a leaf or root if empty tree
2. Let v be a leaf of T . If an ancestor of v is labelled with F then childrento v are added using the following rules.
I F = ¬¬G : a child is added to v with label GI F = α: a child and grand child to v added with labels α1 and α2
I F = β: two children to v are added with labels β1 and β2
¬¬G
v
G
α
v
α1
α2
β
v
β1 β2
![Page 12: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/12.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 12
Example : Tableaux
Example 5.1
Consider Σ = {¬(p ∨ (q ∨ r)),¬(q ∧ ¬r)}
¬(p ∨ (q ∨ r))
¬(q ∧ ¬r)
¬q ¬¬r
¬p
¬(q ∨ r)
r
All intermediate states are tableaux.
![Page 13: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/13.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 13
Closed Tableaux
Definition 5.2A branch of a tableaux is closed if it contains two nodes with labels F and¬F for a formula F .
Definition 5.3A branch of a tableaux is atomically closed if it contains two nodes withlabels p and ¬p for a variable p.
Definition 5.4A tableaux is (atomically) closed if each branch of the tableaux is(atomically) closed.
Exercise 5.1Prove that a closed tableaux can always be expanded to a atomically closedtableaux.
![Page 14: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/14.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 14
Example : Closed Tableaux
Example 5.2
Consider Σ = {¬(p ∨ (q ∨ r)),¬(¬(q ∨ r) ∧ ¬p)}
¬(p ∨ (q ∨ r))
¬(¬(q ∨ r) ∧ ¬p)
¬p
¬(q ∨ r)
¬¬(q ∨ r) ¬¬p
(q ∨ r) p
![Page 15: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/15.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 15
Proven by tableaux
Definition 5.5For a set of formulas Σ, if there exists a closed tableaux then we say Σ istableaux inconsistent. Otherwise, tableaux consistent.
Definition 5.6If {¬F} is tableaux inconsistent then we write `pt F , i.e., F is a theorem ofthe tableaux proof method. We say a closed tableaux for {¬F} is a tableauxproof of F .
We will later show that |= F iff `pt F .
Exercise 5.2a. Describe the structure of a tableaux of a DNF formulab. Describe the structure of a tableaux of a CNF formula
Commentary: pt stands for “propositional tableaux”.
![Page 16: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/16.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 16
Practice tableaux
Exercise 5.3Prove that the following formulas are theorems in tableaux proof method
1. (p ⇒ q) ∧ (p ∨ q)⇒ q
2. (p ⇒ q) ∧ (q ⇒ r)⇒ ¬(¬r ∧ p)
3. (q ∨ (r ∧ s)) ∧ (q ⇒ t) ∧ (t ⇒ s)⇒ s
![Page 17: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/17.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 17
Topic 5.3
Resolution
![Page 18: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/18.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 18
Clauses as sets
Definition 5.7 (clause redefined)
A clause is a finite set of formulas {F1, . . . ,Fn} and interpreted as F1∨ ..∨Fn.
Here we do not require Fi to be a literal.
For a clause C and a formula F , we will write F ∪ C to denote {F} ∪ C .
Example 5.3
{(p ⇒ q), q} is a clause.
![Page 19: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/19.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 19
Resolution proof method
Resolution proof method takes a set of formulas Σ and produces a sequenceof clauses as a proof.
Clauses in the proof are either from Σ or consequences of previous clauses.
The goal of the proof method is to find the empty clause, which stands forinconsistency.
![Page 20: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/20.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 20
Resolution derivation
Definition 5.8A resolution derivation R for a set of formulas Σ is a finite sequence ofclauses that are generated by the following resolution expansion rules.
1. {F} is appended in R if F ∈ Σ
2. If a clause F ∪ C is in R then new clauses are appended using thefollowing rules
2.1 F = ¬¬G : G ∪ C is appended2.2 F = β: β1 ∪ β2 ∪ C is appended2.3 F = α: α1 ∪ C and α2 ∪ C are appended
3. If clauses F ∪ C and ¬F ∪ D are in R then C ∪ D is appended.
Rules 2.1-3 are expansion of the structure of the formulasRule 3 is called resolution which is valid due to the following implication
(p ∨ q) ∧ (¬p ∨ r) |= (q ∨ r)
p is called the pivot, and (p ∨ q) and (¬p ∨ r) are called resolvents.
![Page 21: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/21.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 21
Example: resolution derivation
Example 5.4
Consider Σ = {¬(p ∨ (¬p ∧ q)), (q ∨ (p ⇒ q))}
Resolution derivation
1. {¬(p ∨ (¬p ∧ q))} // Since ¬(p ∨ (¬p ∧ q)) ∈ Σ
2. {(q ∨ (p ⇒ q))} // Since (q ∨ (p ⇒ q)) ∈ Σ
3. {¬p} // applying α expansion on 1
4. {¬(¬p ∧ q)} // applying α expansion on 1
5. {¬¬p,¬q} // applying β expansion on 4
6. {p,¬q} // applying ¬ expansion on 5
7. {¬q} // applying resolution on p,3,6
![Page 22: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/22.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 22
Closed Resolution
Definition 5.9A resolution derivation is closed if it contains the empty clause.
Definition 5.10A closed resolution derivation is atomically closed if resolution is applied onlyusing literal pivots.
Exercise 5.4Prove that a closed resolution derivation can always be extended to aatomically closed resolution derivation.
![Page 23: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/23.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 23
Example 5.5 (closed resolution)Consider Σ = {¬((p ∧ q) ∨ (r ∧ s)⇒ ((p ∨ (r ∧ s)) ∧ (q ∨ (r ∧ s))))}
1. {¬( (p ∧ q) ∨ (r ∧ s)⇒ ((p ∨ (r ∧ s)) ∧ (q ∨ (r ∧ s))) )}2. {(p ∧ q) ∨ (r ∧ s)} // applying α expansion on 1
3. {¬( (p ∨ (r ∧ s)) ∧ (q ∨ (r ∧ s)) )} // applying α expansion on 1
4. {(p ∧ q), (r ∧ s)} // applying β expansion on 2
5. {¬(p ∨ (r ∧ s)),¬(q ∨ (r ∧ s))} // applying β expansion on 3
6. {¬p,¬(q ∨ (r ∧ s))} // applying α expansion on 5
7. {¬(r ∧ s),¬(q ∨ (r ∧ s))} // applying α expansion on 5
8. {¬p,¬q} // applying α expansion on 6
9. {¬p,¬(r ∧ s)} // applying α expansion on 6
10. {¬(r ∧ s)} // applying α expansion on 7
11. {¬(r ∧ s),¬q} // applying α expansion on 7
12. {p ∧ q} // applying resolution on (r ∧ s),10,4
13. {p} // applying α expansion on 12
14. {q} // applying α expansion on 12
15. {¬q} // applying resolution on p,13,8
16. {} // applying resolution on q,14,15
![Page 24: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/24.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 24
Proven by resolution
Definition 5.11For a set of formulas Σ, if there exists a closed resolution derivation then wesay Σ is resolution inconsistent. Otherwise, resolution consistent.
Definition 5.12If {¬F} is resolution inconsistent then we write `pr F , i.e., F is a theorem ofthe proof method. We say a closed resolution derivation of {¬F} is aresolution proof of F .
We will later show that |= F iff `pr F .
Exercise 5.5a. Describe the structure of a resolution proof of a CNF formulab. Describe the structure of a resolution proof of a DNF formula
Commentary: pr stands for “propositional resolution”.
![Page 25: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/25.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 25
Practice resolution
Exercise 5.6Prove that the following formulas are theorems in the resolution proofmethod
1. (p ⇒ q) ∧ (p ∨ q)⇒ q
2. (p ⇒ q) ∧ (q ⇒ r)⇒ ¬(¬r ∧ p)
3. (q ∨ (r ∧ s)) ∧ (q ⇒ t) ∧ (t ⇒ s)⇒ s
4. (p ∨ q) ∧ (r ∨ s)⇒ ((p ∧ r) ∨ q ∨ s)
![Page 26: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/26.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 26
Topic 5.4
Implementation issues in resolution
![Page 27: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/27.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 27
Efficient implementation of proof methods
A proof method implicitly defines a non-deterministic proof search algorithm
In implementing such a algorithm, one needs to ensure that one is not doingunnecessary work.
We will discuss some simple observations that may cut huge search spaces.
This discussion is a preview of muchdetailed discussion about SAT solvers.
We are ignoring Tableaux here. Similar issues can be discussed for Tableaux.
![Page 28: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/28.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 28
Expand only once
We need not apply structural expansion rules multiple times on sameformulas in resolution derivation.
In a resolution derivation, the order of application of the expansions on aclause is irrelevant.
Example 5.6
Consider clause {α, β}First α then β expansion results in
1. {α, β}2. {α1, β}3. {α2, β}4. {α1, β1, β2}5. {α2, β1, β2}
First β then α expansion results in
1. {α, β}2. {α, β1, β2}3. {α1, β1, β2}4. {α2, β1, β2}
![Page 29: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/29.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 29
Superset clauses are redundant
Theorem 5.1For clauses C and D, if D ⊂ C and the empty clause can be derived using Cthen it can be derived using D.
If clause C is superset of clause D, then C is redundant.
Exercise 5.7Prove the above theorem.
![Page 30: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/30.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 30
Ignore valid clauses in resolution
Definition 5.13If a clause contains both F and ¬F then the clause is syntactically valid.
If a syntactically valid clause contributes in deriving the empty clause, thedescendents clause must participate in some resolution with pivot F .
However, that is impossible.
Example 5.7
{F ,C} {¬F ,F ,D}{F ,C ,D}
Resolution
Note that the resolution fails to remove F in the consequence.
If a syntactically valid clause is generated then we can ignore it for anyfurther expansions, without loss of completeness.
![Page 31: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/31.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 31
Pure literals
Definition 5.14If a literal occurs in a CNF formula and its negation does not then it is apure literal.
Theorem 5.2The removal of clauses containing the pure literals in a CNF preservessatisfiability.
Exercise 5.8Prove the above theorem
![Page 32: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/32.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 32
Unit clause propagation
If {F} occurs in a resolution proof, we can remove ¬F from every clause,which is valid because of the following resolutions.
{F} ¬F ∪ D
{D}Resolution
![Page 33: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/33.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 33
Other proof systems
We are skipping the following important proof systems
I Hilbert system
I Natural deduction
I Sequent Calculus
![Page 34: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/34.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 34
Topic 5.5
Problems
![Page 35: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/35.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 35
Practice
Exercise 5.9Use both tableaux and resolution to prove the following theorems
1. (((p ⇒ q)⇒ q)⇒ q)⇒ (p ⇒ q)
2. (p ⇒ (q ∨ r)) ∨ (r ⇒ ¬p)
Use the full formal notations to write the proofs.
![Page 36: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/36.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 36
Dual
Exercise 5.10Give a dual of Tableaux that proves validity of formulas.
Exercise 5.11Give a dual of Resolution that proves validity of formulas.
![Page 37: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/37.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 37
Resolution rule
Exercise 5.12Give a minimal example that require at least two application of the resolutionrule on some clause to derive the empty clause.
![Page 38: Mathematical Logic 2016akg/courses/2016-logic/lec-05.pdf · In order to avoid writing many cases we will use auniform notation. Conjunctive Disjunctive 1 2 1 2 (F ^G) F G :(F ^G)](https://reader035.fdocuments.in/reader035/viewer/2022070911/5fa9c7f755df3d485315b72f/html5/thumbnails/38.jpg)
cbna Mathematical Logic 2016 Instructor: Ashutosh Gupta TIFR, India 38
End of Lecture 5