Well-quasi-orders in Logic Sylvain Schmitz Panhellenic Logic - - PowerPoint PPT Presentation

well quasi orders in logic
SMART_READER_LITE
LIVE PREVIEW

Well-quasi-orders in Logic Sylvain Schmitz Panhellenic Logic - - PowerPoint PPT Presentation

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory Well-quasi-orders in Logic Sylvain Schmitz Panhellenic Logic Symposium, June 29, 2019 1/18 Well-Quasi-Orders Verification Proof Theory Finite Model Theory


slide-1
SLIDE 1

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-quasi-orders in Logic

Sylvain Schmitz Panhellenic Logic Symposium, June 29, 2019

1/18

slide-2
SLIDE 2

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Outline

well-quasi-orders (wqo):

◮ robust notion ◮ selection of applications:

verification algorithm termination proof theory relevance logic finite model theory preservation theorems database theory certain answers

2/18

slide-3
SLIDE 3

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Outline

well-quasi-orders (wqo):

◮ robust notion ◮ selection of applications:

verification algorithm termination proof theory relevance logic finite model theory preservation theorems database theory certain answers

2/18

slide-4
SLIDE 4

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

A One-Player Game

◮ over Q0 × Q0 ◮ given initially (x0,y0) ◮ Eloise plays (xj,yj) s.t.

∀0 i < j, xi > xj or yi > yj

(x1,y1) (x2,y2) (x0,y0)

◮ Can Eloise win, i.e. play indefinitely?

3/18

slide-5
SLIDE 5

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

A One-Player Game

◮ over Q0 × Q0 ◮ given initially (x0,y0) ◮ Eloise plays (xj,yj) s.t.

∀0 i < j, xi > xj or yi > yj

(x1,y1) (x2,y2) (x0,y0)

◮ Can Eloise win, i.e. play indefinitely?

3/18

slide-6
SLIDE 6

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

A One-Player Game

◮ over Q0 × Q0 ◮ given initially (x0,y0) ◮ Eloise plays (xj,yj) s.t.

∀0 i < j, xi > xj or yi > yj

(x1,y1) (x2,y2) (x0,y0)

◮ Can Eloise win, i.e. play indefinitely?

3/18

slide-7
SLIDE 7

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

If (x0,y0) (0,0), then choosing (xj,yj) = (x0

2j , y0 2j ) wins.

4/18

slide-8
SLIDE 8

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

A One-Player Game

◮ over N × N ◮ given initially (x0,y0) ◮ Eloise plays (xj,yj) s.t.

∀0 i < j, xi > xj or yi > yj

(x1,y1) (x2,y2) (x0,y0)

◮ Can Eloise win, i.e. play indefinitely?

5/18

slide-9
SLIDE 9

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Assume there exists an infinite sequence (xj,yj)j of moves over N2.

6/18

slide-10
SLIDE 10

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Assume there exists an infinite sequence (xj,yj)j of moves over N2. Consider the pairs of indices i < j: color (i,j) purple if xi > xj but yi yj, red if xi > xj and yi > yj,

  • range if yi > yj but xi xj.

(3,4) (5,2) (2,3) ...

6/18

slide-11
SLIDE 11

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Assume there exists an infinite sequence (xj,yj)j of moves over N2. Consider the pairs of indices i < j: color (i,j) purple if xi > xj but yi yj, red if xi > xj and yi > yj,

  • range if yi > yj but xi xj.

(3,4) (5,2) (2,3) ... By the infinite Ramsey Theorem, there exists an infinite monochromatic subset of indices.

6/18

slide-12
SLIDE 12

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Assume there exists an infinite sequence (xj,yj)j of moves over N2. Consider the pairs of indices i < j: color (i,j) purple if xi > xj but yi yj, red if xi > xj and yi > yj,

  • range if yi > yj but xi xj.

(3,4) (5,2) (2,3) ... By the infinite Ramsey Theorem, there exists an infinite monochromatic subset of indices. In all cases, it implies the existence of an infinite decreasing sequence in N, a contradiction.

6/18

slide-13
SLIDE 13

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions ◮ algebraic constructions

7/18

slide-14
SLIDE 14

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions: (X,) wqo iff

◮ bad sequences are finite: x0,x1,... is bad if ∀i < j, xi xj ◮ is well-founded and has no infinite antichains ◮ finite basis property: ∅ U ⊆ X has at least one and finitely

many minimal elements

◮ ascending chain condition: any chain U0 U1 ··· of

upwards-closed sets is finite

◮ etc.

◮ algebraic constructions

7/18

slide-15
SLIDE 15

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions: (X,) wqo iff

◮ bad sequences are finite: x0,x1,... is bad if ∀i < j, xi xj ◮ is well-founded and has no infinite antichains ◮ finite basis property: ∅ U ⊆ X has at least one and finitely

many minimal elements

◮ ascending chain condition: any chain U0 U1 ··· of

upwards-closed sets is finite

◮ etc.

◮ algebraic constructions

7/18

slide-16
SLIDE 16

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions: (X,) wqo iff

◮ bad sequences are finite: x0,x1,... is bad if ∀i < j, xi xj ◮ is well-founded and has no infinite antichains ◮ finite basis property: ∅ U ⊆ X has at least one and finitely

many minimal elements

◮ ascending chain condition: any chain U0 U1 ··· of

upwards-closed sets is finite

◮ etc.

◮ algebraic constructions

7/18

slide-17
SLIDE 17

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions: (X,) wqo iff

◮ bad sequences are finite: x0,x1,... is bad if ∀i < j, xi xj ◮ is well-founded and has no infinite antichains ◮ finite basis property: ∅ U ⊆ X has at least one and finitely

many minimal elements

◮ ascending chain condition: any chain U0 U1 ··· of

upwards-closed sets is finite

◮ etc.

◮ algebraic constructions

7/18

slide-18
SLIDE 18

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions: (X,) wqo iff

◮ bad sequences are finite: x0,x1,... is bad if ∀i < j, xi xj ◮ is well-founded and has no infinite antichains ◮ finite basis property: ∅ U ⊆ X has at least one and finitely

many minimal elements

◮ ascending chain condition: any chain U0 U1 ··· of

upwards-closed sets is finite

◮ etc.

◮ algebraic constructions

7/18

slide-19
SLIDE 19

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Well-Quasi-Orders

◮ multiple equivalent definitions ◮ algebraic constructions

◮ Cartesian products (Dickson’s Lemma), ◮ finite sequences (Higman’s Lemma), ◮ disjoint sums, ◮ finite sets with Hoare’s quasi-ordering, ◮ finite trees (Kruskal’s Tree Theorem), ◮ graphs with minors (Robertson and Seymour’s Graph Minor

Theorem),

◮ etc. 7/18

slide-20
SLIDE 20

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Example: Ordinals

  • rdinal: well-founded linear
  • rder

bad sequences are descending sequences: α β iff α > β

8/18

slide-21
SLIDE 21

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Example: Dickson’s Lemma

Lemma (Dickson 1913) If (X,X) and (Y,Y) are two wqos, then (X × Y,×) is a wqo, where × is the product order- ing: x,y × x′,y′

def

⇔ x X x′ ∧ y Y y′ . Example

◮ (Nd,×) using the product ordering ◮ (M(X),m) for finite multiset embedding over finite X

9/18

slide-22
SLIDE 22

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Example: Higman’s Lemma

Lemma (Higman 1952) If (X,) is a wqo, then (X∗,∗) is a wqo where ∗ is the subword embedding ordering: a1 ···am ∗ b1 ···bn

def

  • ∃1 i1 < ··· < im n,

m

j=1 aj A bij .

Example aba ∗ baaacabbab

10/18

slide-23
SLIDE 23

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Example: Bounded Tree-Depth

Lemma (Ding 1992) For all k, (Graphs \ ↑Pk,⊆) is wqo. Non-Examples

... ...

11/18

slide-24
SLIDE 24

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Algorithm Termination

simple (a,b) c ← − 1 while a > 0 ∧ b > 0 a,b,c ← − a − 1,b,2c

  • r

a,b,c ← − 2c,b − 1,1 end

a0,b0,c0 a1,b1,c1 . . . ai,bi,ci . . . aj,bj,cj ×

◮ in any execution, a0,b0,...,an,bn is a bad sequence

  • ver (N2,×),

◮ (N2,×) is a wqo: all the runs are finite

12/18

slide-25
SLIDE 25

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Algorithm Termination

simple (a,b) c ← − 1 while a > 0 ∧ b > 0 a,b,c ← − a − 1,b,2c

  • r

a,b,c ← − 2c,b − 1,1 end

a0,b0,c0 a1,b1,c1 . . . ai,bi,ci . . . aj,bj,cj ×

◮ in any execution, a0,b0,...,an,bn is a bad sequence

  • ver (N2,×),

◮ (N2,×) is a wqo: all the runs are finite

12/18

slide-26
SLIDE 26

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Algorithm Termination

simple (a,b) c ← − 1 while a > 0 ∧ b > 0 a,b,c ← − a − 1,b,2c

  • r

a,b,c ← − 2c,b − 1,1 end

a0,b0,c0 a1,b1,c1 . . . ai,bi,ci . . . aj,bj,cj ×

◮ in any execution, a0,b0,...,an,bn is a bad sequence

  • ver (N2,×),

◮ (N2,×) is a wqo: all the runs are finite

12/18

slide-27
SLIDE 27

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Algorithm Termination

simple (a,b) c ← − 1 while a > 0 ∧ b > 0 a,b,c ← − a − 1,b,2c

  • r

a,b,c ← − 2c,b − 1,1 end

a0,b0,c0 a1,b1,c1 . . . ai,bi,ci . . . aj,bj,cj ×

◮ in any execution, a0,b0,...,an,bn is a bad sequence

  • ver (N2,×),

◮ (N2,×) is a wqo: all the runs are finite

12/18

slide-28
SLIDE 28

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Quasi-ranking Function

Definition A function f from Confs to a wqo (X,) is a quasi-ranking function if, for all executions c0,c1,..., ∀i < j, f(ci) f(cj). Proposition If an algorithm has a quasi-ranking function, then it terminates. Proof. The sequence of ranks f(c0),f(c1),... is a bad sequence

  • ver the wqo (X,).

c.f. Podelski & Rybalchenko’s transition invariants

13/18

slide-29
SLIDE 29

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Quasi-ranking Function

Definition A function f from Confs to a wqo (X,) is a quasi-ranking function if, for all executions c0,c1,..., ∀i < j, f(ci) f(cj). Proposition If an algorithm has a quasi-ranking function, then it terminates. Proof. The sequence of ranks f(c0),f(c1),... is a bad sequence

  • ver the wqo (X,).

c.f. Podelski & Rybalchenko’s transition invariants

13/18

slide-30
SLIDE 30

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Example (A → (B → A)) “if it’s raining (A), then if your favorite color is green (B) then it’s raining (A)” A theorem in classical logic, not in relevance logic. Gentzen-style sequent calculus A,B,C formulæ; Γ,∆ multisets of formulæ; no weakening A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

14/18

slide-31
SLIDE 31

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Example (A → (B → A)) “if it’s raining (A), then if your favorite color is green (B) then it’s raining (A)” A theorem in classical logic, not in relevance logic. Gentzen-style sequent calculus A,B,C formulæ; Γ,∆ multisets of formulæ; no weakening A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

14/18

slide-32
SLIDE 32

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Gentzen-style sequent calculus A,B,C formulæ; Γ,∆ multisets of formulæ; no weakening A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R) Problem (provability) Given a sequent Γ ⊢ A, is it provable? Theorem (Kripke 1959) Provability is decidable in implicational relevance logic.

14/18

slide-33
SLIDE 33

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Theorem (Kripke 1959) Provability is decidable in implicational relevance logic. Proof Sketch.

◮ subformula property ◮ irredundant proof searches

◮ (C) and (→R) commute: (C)’s only below a (→L) ◮ rewrite proofs to apply (C) whenever possible

◮ irredundant proof branches are bad sequences for contraction ◮ ...which is wqo over the subformulæ of Γ ⊢ A

14/18 A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

slide-34
SLIDE 34

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Theorem (Kripke 1959) Provability is decidable in implicational relevance logic. Proof Sketch.

◮ subformula property ◮ irredundant proof searches

◮ (C) and (→R) commute: (C)’s only below a (→L) ◮ rewrite proofs to apply (C) whenever possible

◮ irredundant proof branches are bad sequences for contraction ◮ ...which is wqo over the subformulæ of Γ ⊢ A

14/18 A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

slide-35
SLIDE 35

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Theorem (Kripke 1959) Provability is decidable in implicational relevance logic. Proof Sketch.

◮ subformula property ◮ irredundant proof searches

◮ (C) and (→R) commute: (C)’s only below a (→L) ◮ rewrite proofs to apply (C) whenever possible

◮ irredundant proof branches are bad sequences for contraction ◮ ...which is wqo over the subformulæ of Γ ⊢ A

14/18 A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

slide-36
SLIDE 36

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Relevance Logic

Theorem (Kripke 1959) Provability is decidable in implicational relevance logic. Proof Sketch.

◮ subformula property ◮ irredundant proof searches

◮ (C) and (→R) commute: (C)’s only below a (→L) ◮ rewrite proofs to apply (C) whenever possible

◮ irredundant proof branches are bad sequences for contraction ◮ ...which is wqo over the subformulæ of Γ ⊢ A

14/18 A ⊢ A (Id) Γ,A,A ⊢ B Γ,A ⊢ B (C) Γ ⊢ A ∆,B ⊢ C Γ,∆,A → B ⊢ C (→L) Γ,A ⊢ B Γ ⊢ A → B (→R)

slide-37
SLIDE 37

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Formula ϕ(x,y) = ∃z.x G − → y ∧ ¬(y R − → z)

a1 a2 b1 b2 G G

S

15/18

slide-38
SLIDE 38

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Formula ϕ(x,y) = ∃z.x G − → y ∧ ¬(y R − → z)

a1 a2 b1 b2 G G

S | = ϕ(a1,b1)

15/18

slide-39
SLIDE 39

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Formula ϕ(x,y) = ∃z.x G − → y ∧ ¬(y R − → z)

a1 a2 b1 b2 G G

S | = ϕ(a1,b1)

a′

1

a′

2

b′

1

b′

2

c G G R B

S′

strong injective homomorphism

15/18

slide-40
SLIDE 40

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Formula ϕ(x,y) = ∃z.x G − → y ∧ ¬(y R − → z)

a1 a2 b1 b2 G G

S | = ϕ(a1,b1)

a′

1

a′

2

b′

1

b′

2

c G G R B

S′

strong injective homomorphism

| = ϕ(a′

1,b′ 1)

15/18

slide-41
SLIDE 41

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Formula ϕ(x,y) = ∃z.x G − → y ∧ ¬(y R − → z) ∈ ∃FO

a1 a2 b1 b2 G G

S | = ϕ(a1,b1)

a′

1

a′

2

b′

1

b′

2

c G G R B

S′

strong injective homomorphism

| = ϕ(a′

1,b′ 1)

15/18

slide-42
SLIDE 42

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

logic L example homL ∃FO ∃z.x G − → y ∧ ¬(y R − → z) strong injective ∃FO+() ∃yy′.x R − → y ∧ y′ B − → z ∧ y y′ injective ∃FO+ ∃y.x G − → y all

Fact

If ψ ∈ L, h ∈ homL, and D | = ψ(x), then h(D) | = ψ(h(x)).

15/18

slide-43
SLIDE 43

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

logic L example homL ∃FO ∃z.x G − → y ∧ ¬(y R − → z) strong injective ∃FO+() ∃yy′.x R − → y ∧ y′ B − → z ∧ y y′ injective ∃FO+ ∃y.x G − → y all

Definition

D L D′ if ∃h ∈ homL s.t. D′ = h(D).

15/18

slide-44
SLIDE 44

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

logic L example homL ∃FO ∃z.x G − → y ∧ ¬(y R − → z) strong injective ∃FO+() ∃yy′.x R − → y ∧ y′ B − → z ∧ y y′ injective ∃FO+ ∃y.x G − → y all

Over arbitrary structures Theorem (Ło´ s, Lyndon, Tarski)

If ϕ is an FO-sentence s.t. ϕ is upwards-closed for L, then there exists ψ ∈ L with ϕ = ψ.

15/18

slide-45
SLIDE 45

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

logic L example homL ∃FO ∃z.x G − → y ∧ ¬(y R − → z) strong injective ∃FO+() ∃yy′.x R − → y ∧ y′ B − → z ∧ y y′ injective ∃FO+ ∃y.x G − → y all

Over finite (relational) structures?

15/18

slide-46
SLIDE 46

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

logic L example homL ∃FO ∃z.x G − → y ∧ ¬(y R − → z) strong injective ∃FO+() ∃yy′.x R − → y ∧ y′ B − → z ∧ y y′ injective ∃FO+ ∃y.x G − → y all

Over finite (relational) structures?

15/18

no [Tait 1959] no [Ajtai & Gurevich 1994] yes [Rossman 2008]

slide-47
SLIDE 47

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures

15/18

slide-48
SLIDE 48

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

15/18

slide-49
SLIDE 49

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K ϕ

(upwards-closed inside K) 15/18

slide-50
SLIDE 50

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

↑A1

ϕ

(upwards-closed inside K)

ϕ ∩ K ⊆ ↑{A1,...,An}

find finitely many structures A1,...,An ∈ K s.t. 15/18

slide-51
SLIDE 51

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

↑A1 ↑A2

ϕ

(upwards-closed inside K)

ϕ ∩ K ⊆ ↑{A1,...,An}

find finitely many structures A1,...,An ∈ K s.t. 15/18

slide-52
SLIDE 52

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

↑A1 ↑A2 ↑A3

ϕ

(upwards-closed inside K)

ϕ ∩ K ⊆ ↑{A1,...,An}

find finitely many structures A1,...,An ∈ K s.t. 15/18

slide-53
SLIDE 53

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

(downwards-closed)

↑A1 ↑A2 ↑A3

ϕ

(upwards-closed inside K)

ϕ ∩ K ⊆ ↑{A1,...,An}

find finitely many structures A1,...,An ∈ K s.t. 15/18

slide-54
SLIDE 54

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Preservation Theorems

Over finite (relational) structures? finite σ-structures class K

(downwards-closed)

↑A1 ↑A2 ↑A3

ϕ

(upwards-closed inside K)

ϕ ∩ K ⊆ ↑{A1,...,An}

find finitely many structures A1,...,An ∈ K s.t.

◮ by finite basis property: if (K,L) wqo and

downwards-closed then such A1,...,An exist

◮ associate ψi ∈ L to each Ai s.t. ψi = ↑Ai

15/18

slide-55
SLIDE 55

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Certain Answers

incomplete database I possible completions (I)

ϕ

query

16/18

slide-56
SLIDE 56

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Certain Answers

incomplete database I possible completions (I)

ϕ

query certain answers: true in all completions

16/18

slide-57
SLIDE 57

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Application: Certain Answers

incomplete database I possible completions (I)

ϕ

query certain answers: true in all completions certainI(ϕ) =

  • D∈

(I)

{x | D | = ϕ(x)}

16/18

slide-58
SLIDE 58

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ... 17/18

slide-59
SLIDE 59

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0

17/18

slide-60
SLIDE 60

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0 S1

↑S0 ⊆ ↑S1

17/18

slide-61
SLIDE 61

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0 S1

↑S0 ⊆ ↑S1

S2

⊆ ↑S2

17/18

slide-62
SLIDE 62

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0 S1

↑S0 ⊆ ↑S1

S2

⊆ ↑S2 ≡

...

17/18

slide-63
SLIDE 63

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0 S1

↑S0 ⊆ ↑S1

S2

⊆ ↑S2 ≡

...

17/18

◮ over a wqo: by ascending chain condition,

↑ S ⊆ ↑ S

1

⊆ · · · a l w a y s s t a b i l i s e s t

S

◮ certainI(ϕ) = (domI)∗ ∩

B ∈ S⋆

{ x | B | = ϕ ( x ) }

slide-64
SLIDE 64

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Chase of x G

− → z = ⇒ ∃y . x R − → y ∧ y B − → z for ϕ ∈ ∃FO+()

a1 a2 b1 b2 G G a1 a2 b1 b2 ⊥1 G G R B a1 a2 b1 b2 ⊥2 G G R B a1 a2 b1 b2 ⊥1 G G R B R B a1 a2 b1 b2 ⊥1 ⊥2 G G R B R B ... ... ... ...

S0 S1

↑S0 ⊆ ↑S1

S2

⊆ ↑S2 ≡

...

17/18

◮ over a wqo: by ascending chain condition,

↑ S ⊆ ↑ S

1

⊆ · · · a l w a y s s t a b i l i s e s t

S

◮ certainI(ϕ) = (domI)∗ ∩

B ∈ S⋆

{ x | B | = ϕ ( x ) }

slide-65
SLIDE 65

Well-Quasi-Orders Verification Proof Theory Finite Model Theory Database Theory

Questions?

18/18