Logics of independence and dependence GTS for FO GTS for IF Team - - PowerPoint PPT Presentation

logics of independence and dependence
SMART_READER_LITE
LIVE PREVIEW

Logics of independence and dependence GTS for FO GTS for IF Team - - PowerPoint PPT Presentation

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics Logics of independence and dependence GTS for FO GTS for IF Team semantics for FO Jonni Virtema Team semantics for IF University of Helsinki, Finland


slide-1
SLIDE 1

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

1/ 46 Logics of independence and dependence

Jonni Virtema

University of Helsinki, Finland

Phileth seminar @ Hokkaido University, 11th of May 2017

slide-2
SLIDE 2

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

2/ 46 Outline

Part 0: Introduction to first-order logic Part 1: Game theoretic semantics for first-order logic Part 2: Game theoretic semantics for independence-friendly logic Part 3: Team semantics for independence-friendly logic Part 4: Team semantics via database theory and dependence logic Part 5: Expressive power of dependence logic and IF-logic

slide-3
SLIDE 3

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

3/ 46

Introduction

slide-4
SLIDE 4

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

4/ 46 What is logic ?

slide-5
SLIDE 5

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

4/ 46 What is logic ?

◮ Mathematics, philosophy, linguistics, and theoretical computer science.

slide-6
SLIDE 6

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

4/ 46 What is logic ?

◮ Mathematics, philosophy, linguistics, and theoretical computer science. ◮ Logic is an abstract machinery that can be used in describing the state and

behaviour of systems, and deduction related to these systems.

slide-7
SLIDE 7

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

4/ 46 What is logic (model theoretic view)?

◮ Mathematics, philosophy, linguistics, and theoretical computer science. ◮ Logic is an abstract machinery that can be used in describing the state and

behaviour of systems, and deduction related to these systems.

◮ The core notions are models and formulae.

slide-8
SLIDE 8

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

5/ 46 What are models and formulae?

◮ A model is an abstraction of some state of affairs. ◮ A collection of points together with some arrows from points to points is a

model (e.g., modelling a rail network).

slide-9
SLIDE 9

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

5/ 46 What are models and formulae?

◮ A model is an abstraction of some state of affairs. ◮ A collection of points together with some arrows from points to points is a

model (e.g., modelling a rail network).

◮ A formula is an abstraction of a claim related to some state of affairs. ◮ A sentence that describes a possible property of a model is a formula (e.g.,

a statement that there is a rail connection from Tokyo to Sapporo).

slide-10
SLIDE 10

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

6/ 46 First-order models

◮ An n-ary relation R over a set A is a subset of An. ◮ For simplicity we now consider only 1-ary and 2-ary relations, i.e., subsets of

A and {(a, b) | a, b ∈ A}, respectively.

slide-11
SLIDE 11

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

6/ 46 First-order models

◮ An n-ary relation R over a set A is a subset of An. ◮ For simplicity we now consider only 1-ary and 2-ary relations, i.e., subsets of

A and {(a, b) | a, b ∈ A}, respectively.

◮ (We omit function symbols and constant symbols)

slide-12
SLIDE 12

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

6/ 46 First-order models

◮ An n-ary relation R over a set A is a subset of An. ◮ For simplicity we now consider only 1-ary and 2-ary relations, i.e., subsets of

A and {(a, b) | a, b ∈ A}, respectively.

◮ (We omit function symbols and constant symbols)

Definition

Let τ = {P, R} where P is a 1-ary relation symbol and R a 2-ary relation

  • symbol. A τ-model A is a tuple (A, P, R), where

◮ A is a nonempty set called the domain of A, ◮ R ⊆ A × A is a binary relation over A, and ◮ P ⊆ A is a unary relation over A.

slide-13
SLIDE 13

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

7/ 46 First-order models

Example

◮ Let A be the set of all cities in Japan. ◮ Let R = {(Sapporo, Hakodate), (Hakodate, Hirosaki), (Hirosaki, Sapporo)}. ◮ Let P = {Hakodate, Hirosaki}.

Now (A, R, P) is a first-order model that describes my travel during the Golden Week.

slide-14
SLIDE 14

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

8/ 46 First-order language

Definition

The formulae for first-order logic FO over vocabulary {R, P} is generated by the following grammar: ϕ ::= P(x) | R(x, y) | x = y | ¬ϕ | ϕ ∧ ϕ | ϕ ∨ ϕ | ∃xϕ | ∀xϕ, where x and y are variable symbols.

slide-15
SLIDE 15

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

8/ 46 First-order language

Definition

The formulae for first-order logic FO over vocabulary {R, P} is generated by the following grammar: ϕ ::= P(x) | R(x, y) | x = y | ¬ϕ | ϕ ∧ ϕ | ϕ ∨ ϕ | ∃xϕ | ∀xϕ, where x and y are variable symbols.

Example

Using the example model of the last slide, the sentence: I visited two Japanese cities during the Golden Week can be written in first-oder logic as follows: ∃x∃y(¬x = y ∧ P(x) ∧ P(y))

slide-16
SLIDE 16

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

9/ 46 The connection between models and formulae

A model describes some system and a formula describes some possible property

  • f that system.
slide-17
SLIDE 17

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

9/ 46 The connection between models and formulae

A model describes some system and a formula describes some possible property

  • f that system.

The question: Does the property described hold in the system or not?

slide-18
SLIDE 18

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

9/ 46 The connection between models and formulae

A model describes some system and a formula describes some possible property

  • f that system.

The question: Does the property described hold in the system or not? Is the formula true in the model?

slide-19
SLIDE 19

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

9/ 46 The connection between models and formulae

A model describes some system and a formula describes some possible property

  • f that system.

The question: Does the property described hold in the system or not? Is the formula true in the model? Formally: Does A, s | = ϕ hold?

slide-20
SLIDE 20

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

10/ 46 Tarski semantics of first-order logic (Tarski 1930s)

An assignment s : Var → A is a function that gives a value for each variable symbol in Var.

Definition

Let A be a {R, P} model and s an assignment. The satisfaction relation A, s | = ϕ for FO is defined as follows. A, s | = x = y ⇔ s(x) = s(y). A, s | = P(x) ⇔ s(x) ∈ P. A, s | = R(x, y) ⇔ (s(x), s(y)) ∈ R. A, s | = ¬ϕ ⇔ A, s | = ϕ. A, s | = (ϕ ∧ ψ) ⇔ A, s | = ϕ and A, s | = ψ. A, s | = (ϕ ∨ ψ) ⇔ A, s | = ϕ or A, s | = ψ. A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A. A, s | = ∀xϕ ⇔ A, s(x → a) | = ϕ for every a ∈ A.

slide-21
SLIDE 21

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

11/ 46 Toy example

We can now use Tarski semantics to check that, indeed, I visited two Japanense cities during Golden Week, since A, s | = ∃x∃y(¬x = y ∧ P(x) ∧ P(y)), where A = (A, R, P) is the first-order model defined such that

◮ A is the set of all cities in Japan. ◮ R = {(Sapporo, Hakodate), (Hakodate, Hirosaki), (Hirosaki, Sapporo)}. ◮ P = {Hakodate, Hirosaki}.

slide-22
SLIDE 22

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

12/ 46

PART 1

GTS semantics for FO

slide-23
SLIDE 23

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

13/ 46 Game theoretic semantics of FO (Hintikka 1968)

◮ An alternative way to give meaning for FO-formulae. ◮ The game G(ϕ, (A, s)) is a two-player game. ◮ Players: (∀) Abelard and (∃) Eloise.

slide-24
SLIDE 24

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

13/ 46 Game theoretic semantics of FO (Hintikka 1968)

◮ An alternative way to give meaning for FO-formulae. ◮ The game G(ϕ, (A, s)) is a two-player game. ◮ Players: (∀) Abelard and (∃) Eloise. ◮ Abelard wants to establish that ϕ is not true in A, s, while Eloise wishes to

show that ϕ is true in A, s.

◮ Abelard controls ∀ and ∧, while Eloise controls ∃ and ∨.

slide-25
SLIDE 25

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

13/ 46 Game theoretic semantics of FO (Hintikka 1968)

◮ An alternative way to give meaning for FO-formulae. ◮ The game G(ϕ, (A, s)) is a two-player game. ◮ Players: (∀) Abelard and (∃) Eloise. ◮ Abelard wants to establish that ϕ is not true in A, s, while Eloise wishes to

show that ϕ is true in A, s.

◮ Abelard controls ∀ and ∧, while Eloise controls ∃ and ∨. ◮ (Negation ¬ switches the roles of Abelard and Eloise. However in order to

simplify things we assume that negations may occur only in front of atomic formulae.)

slide-26
SLIDE 26

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)).

slide-27
SLIDE 27

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

slide-28
SLIDE 28

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (ψ1 ∨ ψ2, (A, s)), Eloise chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

slide-29
SLIDE 29

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (ψ1 ∨ ψ2, (A, s)), Eloise chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (∀xψ, (A, s)), Abelard chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

slide-30
SLIDE 30

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (ψ1 ∨ ψ2, (A, s)), Eloise chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (∀xψ, (A, s)), Abelard chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

◮ In position (∃xψ, (A, s)), Eloise chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

slide-31
SLIDE 31

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

14/ 46 Rules of G(ϕ, (A, s))

◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (ψ1 ∨ ψ2, (A, s)), Eloise chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (∀xψ, (A, s)), Abelard chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

◮ In position (∃xψ, (A, s)), Eloise chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

◮ The game is played until a position (ψ, (A, t)) is reached, where ψ is a

literal.

slide-32
SLIDE 32

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

15/ 46 End Game

◮ Eloise wins the play of the game if

◮ (x = y, (A, t)) is reached and t(x) = t(y), ◮ (¬x = y, (A, t)) is reached and t(x) = t(y)

slide-33
SLIDE 33

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

15/ 46 End Game

◮ Eloise wins the play of the game if

◮ (x = y, (A, t)) is reached and t(x) = t(y), ◮ (¬x = y, (A, t)) is reached and t(x) = t(y) ◮ (P(x), (A, t)) is reached and t(x) ∈ P, ◮ (¬P(x), (A, t)) is reached and t(x) ∈ P, ◮ (R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R, ◮ ¬(R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R,

◮ otherwise Abelard wins the play.

slide-34
SLIDE 34

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

15/ 46 End Game

◮ Eloise wins the play of the game if

◮ (x = y, (A, t)) is reached and t(x) = t(y), ◮ (¬x = y, (A, t)) is reached and t(x) = t(y) ◮ (P(x), (A, t)) is reached and t(x) ∈ P, ◮ (¬P(x), (A, t)) is reached and t(x) ∈ P, ◮ (R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R, ◮ ¬(R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R,

◮ otherwise Abelard wins the play.

To win the Game it is not enough to win a single play. To win the game the player has to be able to win EVERY play.

slide-35
SLIDE 35

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

16/ 46 Strategies

◮ A strategy for Eloise in the game G(ϕ, (A, s)) is a function that maps every

possible position that she controls and that may occur in some play of the game to a single next position of the play.

slide-36
SLIDE 36

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

16/ 46 Strategies

◮ A strategy for Eloise in the game G(ϕ, (A, s)) is a function that maps every

possible position that she controls and that may occur in some play of the game to a single next position of the play.

◮ A strategy for Abelard in the game G(ϕ, (A, s)) is a function that maps

every possible position that he controls and that may occur in some play of the game to a single next position of the play.

slide-37
SLIDE 37

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

16/ 46 Strategies

◮ A strategy for Eloise in the game G(ϕ, (A, s)) is a function that maps every

possible position that she controls and that may occur in some play of the game to a single next position of the play.

◮ A strategy for Abelard in the game G(ϕ, (A, s)) is a function that maps

every possible position that he controls and that may occur in some play of the game to a single next position of the play.

◮ A stategy of a player is a winning strategy for that player if she/he wins

every play of the game in which she/he plays according to her/his strategy.

slide-38
SLIDE 38

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

17/ 46 Winner of the game

◮ Eloise wins the game G(ϕ, (A, s)) if she has a winning strategy for that

game.

slide-39
SLIDE 39

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

17/ 46 Winner of the game

◮ Eloise wins the game G(ϕ, (A, s)) if she has a winning strategy for that

game.

◮ A formula ϕ is true in A, s under GTS (denoted by A, s |

=GTS ϕ) if Eloise wins the game G(ϕ, (A, s)).

slide-40
SLIDE 40

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

17/ 46 Winner of the game

◮ Eloise wins the game G(ϕ, (A, s)) if she has a winning strategy for that

game.

◮ A formula ϕ is true in A, s under GTS (denoted by A, s |

=GTS ϕ) if Eloise wins the game G(ϕ, (A, s)).

◮ A, s |

=GTS ϕ holds if and only if A, s | = ϕ holds (uses the axiom of choice).

slide-41
SLIDE 41

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

18/ 46 Toy examples

◮ A, s |

=GTS ∀x∃y(x = y) is always true.

◮ A, s |

=GTS ∀x∃y(¬x = y) is true if the model has at least two elements.

slide-42
SLIDE 42

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

19/ 46 Games of perfect information

◮ Both players in the plays of GTS for FO, when making a move, remember

all previous moves made by both players in that play.

◮ Next we move to a variant of GTS in which this is not the case.

slide-43
SLIDE 43

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

20/ 46

PART 2

GTS semantics for IF-logic

slide-44
SLIDE 44

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

21/ 46 Independence-Friendly logic (Hintikka and Sandu 1989)

Syntax for IF-logic: ϕ ::= x = y | ¬x = y | R(x) | ¬R(x) | ϕ ∧ ϕ | ϕ ∨ ϕ | ∃x/W ϕ | ∀xϕ where W is a set of variable symbols.

◮ The reading of ∃x/{y}ϕ is that there exists x independently of y such that

ϕ holds.

slide-45
SLIDE 45

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

21/ 46 Independence-Friendly logic (Hintikka and Sandu 1989)

Syntax for IF-logic: ϕ ::= x = y | ¬x = y | R(x) | ¬R(x) | ϕ ∧ ϕ | ϕ ∨ ϕ | ∃x/W ϕ | ∀xϕ where W is a set of variable symbols.

◮ The reading of ∃x/{y}ϕ is that there exists x independently of y such that

ϕ holds.

◮ This kind of language is used, e.g., in many definitions in mathematics.

◮ Continuity: a function f : D → R is continuous, if for all a ∈ D and for all

ǫ > 0 there exists δ > 0 such that for all x ∈ D, if |x − a| < δ, then |f (x) − f (a)| < ǫ.

◮ Uniformly continuity: a function f : D → R is uniformly continuous, if for all

a ∈ D and for all ǫ > 0 there exists δ > 0 independent of a such that for all x ∈ D, if |x − a| < δ, then |f (x) − f (a)| < ǫ.

slide-46
SLIDE 46

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

22/ 46 Game theoretic semantics of IF

◮ The game GIF(ϕ, (A, s)) is a two-player game of imperfect information. ◮ Players: (∀) Abelard and (∃) Eloise. ◮ Abelard wants to establish that ϕ is not true in A, s, while Eloise wishes to

show that ϕ is true in A, s.

◮ Abelard controls ∀ and ∧, while Eloise controls ∃ and ∨.

slide-47
SLIDE 47

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

23/ 46 Rules of GIF(ϕ, (A, s))

◮ Plays of GIF(ϕ, (A, s)) and G(ϕ, (A, s)) are identical.

slide-48
SLIDE 48

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

23/ 46 Rules of GIF(ϕ, (A, s))

◮ Plays of GIF(ϕ, (A, s)) and G(ϕ, (A, s)) are identical. ◮ A play of the game starts from the position (ϕ, (A, s)). ◮ In position (ψ1 ∧ ψ2, (A, s)), Abelard chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (ψ1 ∨ ψ2, (A, s)), Eloise chooses i ∈ {1, 2} and the game

continues from position (ψi, (A, s)).

◮ In position (∀xψ, (A, s)), Abelard chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

◮ In position (∃x/W ψ, (A, s)), Eloise chooses a ∈ A and the game continues

from position (ψ, (A, s(x → a))).

◮ The game is played until a position (ψ, (A, t)) is reached, where ψ is a

literal.

slide-49
SLIDE 49

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

24/ 46 End Game

◮ Eloise wins the play of the game if

◮ (x = y, (A, t)) is reached and t(x) = t(y), ◮ (¬x = y, (A, t)) is reached and t(x) = t(y) ◮ (P(x), (A, t)) is reached and t(x) ∈ P, ◮ (¬P(x), (A, t)) is reached and t(x) ∈ P, ◮ (R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R, ◮ ¬(R(x, y), (A, t)) is reached and (t(x), t(y)) ∈ R,

◮ otherwise Abelard wins the play.

To win the Game it is not enough to win a single play.

slide-50
SLIDE 50

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

25/ 46 Strategies

◮ A strategy for Eloise in the game GIF(ϕ, (A, s)) is a function that maps

every possible position that she controls and that may occur in some play of the game to a single next position of the play.

slide-51
SLIDE 51

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

25/ 46 Strategies

◮ A strategy for Eloise in the game GIF(ϕ, (A, s)) is a function that maps

every possible position that she controls and that may occur in some play of the game to a single next position of the play.

◮ A strategy function f of Eloise is consistent (with her knowledge) if the

following holds: If (∃x/W ϕ, (A, t)) and (∃x/W ϕ, (A, t′)) are positions in the game such that t(y) = t′(y) for every y ∈ Var \ W , then f ((∃x/W ϕ, (A, t)) = f ((∃x/W ϕ, (A, t′)).

slide-52
SLIDE 52

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

25/ 46 Strategies

◮ A strategy for Eloise in the game GIF(ϕ, (A, s)) is a function that maps

every possible position that she controls and that may occur in some play of the game to a single next position of the play.

◮ A strategy function f of Eloise is consistent (with her knowledge) if the

following holds: If (∃x/W ϕ, (A, t)) and (∃x/W ϕ, (A, t′)) are positions in the game such that t(y) = t′(y) for every y ∈ Var \ W , then f ((∃x/W ϕ, (A, t)) = f ((∃x/W ϕ, (A, t′)).

◮ A stategy of Eloise is a winning strategy if she wins every play of the game

in which she plays according to her strategy.

slide-53
SLIDE 53

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

26/ 46 Winner of the game

◮ Eloise wins the game GIF(ϕ, (A, s)) if she has a winning consistent strategy

for that game.

◮ A formula ϕ is true in A, s under GTS-IF (denoted by A, s |

=GTS−IF ϕ) if Eloise wins the game GIF(ϕ, (A, s)).

slide-54
SLIDE 54

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

26/ 46 Winner of the game

◮ Eloise wins the game GIF(ϕ, (A, s)) if she has a winning consistent strategy

for that game.

◮ A formula ϕ is true in A, s under GTS-IF (denoted by A, s |

=GTS−IF ϕ) if Eloise wins the game GIF(ϕ, (A, s)).

◮ Hintikka claimed that there does not exists a compostional (Taski-style)

semantics for IF-logic.

slide-55
SLIDE 55

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

26/ 46 Winner of the game

◮ Eloise wins the game GIF(ϕ, (A, s)) if she has a winning consistent strategy

for that game.

◮ A formula ϕ is true in A, s under GTS-IF (denoted by A, s |

=GTS−IF ϕ) if Eloise wins the game GIF(ϕ, (A, s)).

◮ Hintikka claimed that there does not exists a compostional (Taski-style)

semantics for IF-logic.

◮ Hodges (1997) presented a Tarski-style semantics for IF-logic that instead

  • f assignments use sets of assignments as satisfying elements (team

semantics).

slide-56
SLIDE 56

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

27/ 46 Toy examples

◮ A, s |

=GTS−IF ∀x∃y/{x}(x = y)

◮ A, s |

=GTS−IF ∀x∃y/{x}(¬x = y)

slide-57
SLIDE 57

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

27/ 46 Toy examples

◮ A, s |

=GTS−IF ∀x∃y/{x}(x = y) is true only if A is a singleton set.

◮ A, s |

=GTS−IF ∀x∃y/{x}(¬x = y) is never true.

slide-58
SLIDE 58

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

28/ 46

PART 3

Team semantics for FO and IF

slide-59
SLIDE 59

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

29/ 46 First-order logic

Grammar of first-order logic FO in negation normal form: ϕ ::= x = y | ¬(x = y) | R(x) | ¬R(x) | (ϕ ∨ ϕ) | (ϕ ∧ ϕ) | ∃xϕ(x) | ∀xϕ(x) A team of an FO-structure A is any set X of assignments s : VAR → A with a common domain VAR of FO variables.

slide-60
SLIDE 60

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

29/ 46 First-order logic

Grammar of first-order logic FO in negation normal form: ϕ ::= x = y | ¬(x = y) | R(x) | ¬R(x) | (ϕ ∨ ϕ) | (ϕ ∧ ϕ) | ∃xϕ(x) | ∀xϕ(x) A team of an FO-structure A is any set X of assignments s : VAR → A with a common domain VAR of FO variables. We want to define team semantics for FO s.t. we have the following property (flattness): If ϕ is an FO-formula, A a first-order structure, and X a set of assignments: A | =X ϕ ⇐ ⇒ ∀s ∈ X : A, s | = ϕ.

slide-61
SLIDE 61

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A, s | = R(x) ⇔ s(x) ∈ RA A, s | = ¬R(x) ⇔ s(x) ∈ RA A, s | = ϕ ∧ ψ ⇔ A, s | = ϕ and A, s | = ψ A, s | = ϕ ∨ ψ ⇔ A, s | = ϕ or A, s | = ψ A, s | = ∀xϕ ⇔ A, s(x → a) | = ϕ for all a ∈ A A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A

slide-62
SLIDE 62

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A, s | = ϕ ∧ ψ ⇔ A, s | = ϕ and A, s | = ψ A, s | = ϕ ∨ ψ ⇔ A, s | = ϕ or A, s | = ψ A, s | = ∀xϕ ⇔ A, s(x → a) | = ϕ for all a ∈ A A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A

slide-63
SLIDE 63

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ϕ ∧ ψ ⇔ A | =X ϕ and A | =X ψ A, s | = ϕ ∨ ψ ⇔ A, s | = ϕ or A, s | = ψ A, s | = ∀xϕ ⇔ A, s(x → a) | = ϕ for all a ∈ A A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A

slide-64
SLIDE 64

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ϕ ∧ ψ ⇔ A | =X ϕ and A | =X ψ A | =X ϕ ∨ ψ ⇔ A | =Y ϕ and A | =Z ψ for some Y ∪ Z = X A, s | = ∀xϕ ⇔ A, s(x → a) | = ϕ for all a ∈ A A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A

slide-65
SLIDE 65

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ϕ ∧ ψ ⇔ A | =X ϕ and A | =X ψ A | =X ϕ ∨ ψ ⇔ A | =Y ϕ and A | =Z ψ for some Y ∪ Z = X A | =X ∀xϕ ⇔ A | =X[A/x] ϕ A, s | = ∃xϕ ⇔ A, s(x → a) | = ϕ for some a ∈ A Where X[A/x] := {s(x → a) | s ∈ X, a ∈ A}

slide-66
SLIDE 66

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ϕ ∧ ψ ⇔ A | =X ϕ and A | =X ψ A | =X ϕ ∨ ψ ⇔ A | =Y ϕ and A | =Z ψ for some Y ∪ Z = X A | =X ∀xϕ ⇔ A | =X[A/x] ϕ A | =X ∃xϕ ⇔ A | =X[F/x] ϕ for some F : X → A Where X[A/x] := {s(x → a) | s ∈ X, a ∈ A} and X[F/x] := {s(x → F(x)) | s ∈ X}

slide-67
SLIDE 67

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

30/ 46 Team semantics for first-order logic

Recall that a team is a set of first-order assignments with a common domain. A | =X R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ¬R(x) ⇔ ∀s ∈ X : s(x) ∈ RA A | =X ϕ ∧ ψ ⇔ A | =X ϕ and A | =X ψ A | =X ϕ ∨ ψ ⇔ A | =Y ϕ and A | =Z ψ for some Y ∪ Z = X A | =X ∀xϕ ⇔ A | =X[A/x] ϕ A | =X ∃xϕ ⇔ A | =X[F/x] ϕ for some F : X → A Where X[A/x] := {s(x → a) | s ∈ X, a ∈ A} and X[F/x] := {s(x → F(x)) | s ∈ X} For every FO-formula ϕ the following holds: A | =X ϕ ⇐ ⇒ ∀s ∈ X : A, s | = ϕ.

slide-68
SLIDE 68

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

31/ 46 Team semantics for independence-friendly logic

◮ Otherwise the same as for FO, but an additional rule for ∃x/W is needed.

slide-69
SLIDE 69

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

31/ 46 Team semantics for independence-friendly logic

◮ Otherwise the same as for FO, but an additional rule for ∃x/W is needed. ◮ A function F : X → A is W -independent if for every s, t ∈ X the implication

(∀x ∈ dom(X) \ W : s(x) = t(x)) ⇒ F(s) = F(t) holds (above dom(X) denotes the domain of the assignments of X).

slide-70
SLIDE 70

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

31/ 46 Team semantics for independence-friendly logic

◮ Otherwise the same as for FO, but an additional rule for ∃x/W is needed. ◮ A function F : X → A is W -independent if for every s, t ∈ X the implication

(∀x ∈ dom(X) \ W : s(x) = t(x)) ⇒ F(s) = F(t) holds (above dom(X) denotes the domain of the assignments of X).

◮ Now we have the following rule for ∃x/W :

A | =X ∃x/W ϕ ⇔ A | =X[F/x] ϕ for some W -independent F : X → A

slide-71
SLIDE 71

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

32/ 46 Properties of IF-logic

◮ Downward closure holds: (A |

=X ϕ and Y ⊆ X) ⇒ A | =Y ϕ.

slide-72
SLIDE 72

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

32/ 46 Properties of IF-logic

◮ Downward closure holds: (A |

=X ϕ and Y ⊆ X) ⇒ A | =Y ϕ.

◮ IF is strictly more expressive than FO. ◮ Violates flattness: A |

=X ϕ iff ∀s ∈ X : A, s | = ϕ may not hold.

slide-73
SLIDE 73

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

32/ 46 Properties of IF-logic

◮ Downward closure holds: (A |

=X ϕ and Y ⊆ X) ⇒ A | =Y ϕ.

◮ IF is strictly more expressive than FO. ◮ Violates flattness: A |

=X ϕ iff ∀s ∈ X : A, s | = ϕ may not hold.

◮ Violates locality: Truth of an IF-formula may depend on the interpretations

  • f variables that do not occur in the formula!
slide-74
SLIDE 74

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

33/ 46 Toy examples

◮ ∀x∃y(∃z/{x}x = z ∧ ∃z′/{x, y}¬x = z′) ◮ ∃y/{x}x = y

slide-75
SLIDE 75

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

33/ 46 Toy examples

◮ ∀x∃y(∃z/{x}x = z ∧ ∃z′/{x, y}¬x = z′) is true if the model is infinite. ◮ ∃y/{x}x = y

slide-76
SLIDE 76

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

33/ 46 Toy examples

◮ ∀x∃y(∃z/{x}x = z ∧ ∃z′/{x, y}¬x = z′) is true if the model is infinite. ◮ ∃y/{x}x = y in a team X with domain {x, y, z} depends on the values of z

in X, although z does not occur in the formula (signalling).

slide-77
SLIDE 77

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

34/ 46

PART 4

Team semantics and database theory and dependence logic

slide-78
SLIDE 78

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

35/ 46 Team semantics via database theoretic spectacles

◮ A team is a set of assignments that have a common domain of variables. ◮ A team is a database table.

◮ Variables correspond to attributes. ◮ Assignments correspond to records.

◮ Dependency notions of database theory give rise to novel atomic formulae.

◮ Functional dependence gives rise to dependence atoms dep(x1, . . . , xn). ◮ Inclusion dependence gives rise to inclusion atoms x ⊆ y. ◮ Embedded multivalued dependency gives rise to independence atoms y⊥x z.

slide-79
SLIDE 79

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

36/ 46 Dependence logic (V¨

a¨ an¨ anen 2007)

In FO, formulas are formed using connectives ∨, ∧, ¬, and quantifiers ∃ and ∀.

Definition

Dependence logic D extends the syntax of FO by dependence atoms dep(x1, . . . , xn, y) .

slide-80
SLIDE 80

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

36/ 46 Dependence logic (V¨

a¨ an¨ anen 2007)

In FO, formulas are formed using connectives ∨, ∧, ¬, and quantifiers ∃ and ∀.

Definition

Dependence logic D extends the syntax of FO by dependence atoms dep(x1, . . . , xn, y) . The reading of dep(x1, . . . , xn, y) is that the value for y is determined by the values of the variables x1 . . . , xn.

slide-81
SLIDE 81

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

37/ 46 Interpretation of dependence atoms

Let A be a model and X a team with co-domain A and domain V s.t. {x1, ..., xn, y} ⊆ V . A | =X dep(x1, ..., xn, y), if and only if, for all s, s′ ∈ X:

  • 1≤i≤n

s(xi) = s′(xi) = ⇒ s(y) = s′(y).

slide-82
SLIDE 82

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

38/ 46 Examples of teams

We may think of the variables xi as attributes of a database such as x0 = SALARY and x2 = ID NUMBER. x0 . . . xn s0 a0,m . . . an,m . . . sm a0,m . . . an,m Then dependence atom dep(x2, x0) expresses the functional dependence ID NUMBER → SALARY.

slide-83
SLIDE 83

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

39/ 46 Toy example

◮ A |

=X ∀x(dep(x) ∨ dep(x) ∨ dep(x)) holds if and only if |A| ≤ 3.

◮ There is a simple formula expressing that |A| is even.

slide-84
SLIDE 84

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

40/ 46 Dependence logic vs. Independence-Friendly logic

◮ Idea in IF: in ∃x/W ϕ the value for x is picked independently on the

values of the variables in W .

◮ Idea in D: dep(x1, . . . , xn, y) states that the value for y depends only on

variables x1 . . . , xn

slide-85
SLIDE 85

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

40/ 46 Dependence logic vs. Independence-Friendly logic

◮ Idea in IF: in ∃x/W ϕ the value for x is picked independently on the

values of the variables in W .

◮ Idea in D: dep(x1, . . . , xn, y) states that the value for y depends only on

variables x1 . . . , xn

◮ D is an adaptation of IF in which we shift from declaring independences in

quantification of variables to declaring dependences by atomic statements.

◮ For D locality holds!

slide-86
SLIDE 86

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

40/ 46 Dependence logic vs. Independence-Friendly logic

◮ Idea in IF: in ∃x/W ϕ the value for x is picked independently on the

values of the variables in W .

◮ Idea in D: dep(x1, . . . , xn, y) states that the value for y depends only on

variables x1 . . . , xn

◮ D is an adaptation of IF in which we shift from declaring independences in

quantification of variables to declaring dependences by atomic statements.

◮ For D locality holds! ◮ It comes with no surprise that the expressive powers of the logics coincide.

slide-87
SLIDE 87

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

41/ 46

PART 5

Expressive power of D and IF

slide-88
SLIDE 88

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

42/ 46 Expressive power of IF and D

Theorem

For every formula ϕ ∈ IF that uses only variables {x1, . . . , xn} there exists a formula ϕ∗ ∈ D such that for every model A and team X of A, where Dom(X) = {x1, . . . xn}, it holds that A | =X ϕ ⇔ A | =X ϕ∗.

Theorem

For every formula ϕ ∈ D that uses only variables {x1, . . . , xn} there exists a formula ϕ∗ ∈ IF such that for every model A and team X of A, where Dom(X) = {x1, . . . xn}, it holds that A | =X ϕ ⇔ A | =X ϕ∗.

slide-89
SLIDE 89

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

43/ 46 Existential second-order logic

Formulas of existential second-order logic (ESO) over vocabulary τ are of the form ∃R1 . . . ∃Rnϕ, where Ri:s are relation variables (of some arity) and ϕ is a formula of first-order logic over the vocabulary τ ∪ {R1, . . . Rn}.

slide-90
SLIDE 90

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

43/ 46 Existential second-order logic

Formulas of existential second-order logic (ESO) over vocabulary τ are of the form ∃R1 . . . ∃Rnϕ, where Ri:s are relation variables (of some arity) and ϕ is a formula of first-order logic over the vocabulary τ ∪ {R1, . . . Rn}. Semantics for ESO is defined analogously to FO, with the additional rule: A, s | = ∃Rϕ ⇔ A, s(R → B) | = ϕ for some B ⊆ An, where R is a relation variable of arity n.

slide-91
SLIDE 91

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

44/ 46 Expressive Power

For sentences the expressive power of D and IF coincide with the expressive power of ESO (Ederton 1970, Walkoe 1970; the result for partially ordered quantifiers).

slide-92
SLIDE 92

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

44/ 46 Expressive Power

For sentences the expressive power of D and IF coincide with the expressive power of ESO (Ederton 1970, Walkoe 1970; the result for partially ordered quantifiers). Dependence logic (and thus also IF-logic) defines all downward closed ESO properties of teams.

Theorem (Kontinen, V¨ a¨ an¨ anen 2009)

For every sentence ψ ∈ ESO[τ ∪ {R}], in which R appears only negatively, there is φ(y1, . . . , yk) ∈ D[τ] s.t. for all A and X = ∅ with domain {y1, . . . , yk} A | =X φ ⇐ ⇒ (A, R := X(y)) | = ψ.

slide-93
SLIDE 93

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

45/ 46 Related works

The framework of team semantics has been introduced to many areas of logic and variety of results have been obtained.

◮ In first-order setting: dependence logic, inclusion logic, independence logic,

probabilistic variants, etc.

◮ Modal and propostional variants: modal dependence logic, propositional

dependence logics etc.

◮ Expressive power, axiomatizations, definability, frame definabily etc. have

been studied.

◮ Connections to dependency theory of database theory. ◮ Connections to notions of independence in statistics.

slide-94
SLIDE 94

Logics of independence and dependence Jonni Virtema What is logic? Tarski Semantics GTS for FO GTS for IF Team semantics for FO Team semantics for IF Teams as databases Dependence logic IF vs. D Relation to ESO Related works

46/ 46

THANKS!