Categorical Semantics for Linear Logic Categorical semantics for - - PowerPoint PPT Presentation

categorical semantics for linear logic
SMART_READER_LITE
LIVE PREVIEW

Categorical Semantics for Linear Logic Categorical semantics for - - PowerPoint PPT Presentation

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical Semantics for Linear Logic Categorical semantics for linear logic Interaction Wolfgang Jeltsch between linear and non-linear Based on work by Nick Benton


slide-1
SLIDE 1

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Categorical Semantics for Linear Logic

Wolfgang Jeltsch

Based on work by Nick Benton (1994) TT¨ U K¨ uberneetika Instituut

Teooriaseminar

18 Juni 2013

slide-2
SLIDE 2

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

1 Linear logic 2 Categorical semantics for linear logic 3 Interaction between linear and non-linear logic 4 References

slide-3
SLIDE 3

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

1 Linear logic 2 Categorical semantics for linear logic 3 Interaction between linear and non-linear logic 4 References

slide-4
SLIDE 4

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Linear logic

  • useful for reasoning about resources
  • each proposition must be used exactly once in a proof
  • very different from the normal understanding of logic
  • classical and intuitionistic variant
  • in this talk, only intuitionistic linear logic
slide-5
SLIDE 5

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Linear logic formulas

  • language:

F ::= F ⊗ F | 1 | F & F | ⊤ | F ⊕ F | 0 | F ⊸ F | !F

  • meanings:

α ⊗ β α and β hold simultaneously 1 nothing holds α & β α and β hold (not necessarily simultaneously) ⊤ tautology α ⊕ β α or β holds 0 absurdity α ⊸ β if α holds in addition, then β holds !α α holds arbitrarily often

slide-6
SLIDE 6

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Linear logic example

  • atomic propositions:

e I have one euro. s/p/i I get a soup/a pancake/an icecream.

  • derived propositions:
  • For four euros, I get a soup and a pancake:

e ⊗ e ⊗ e ⊗ e ⊸ s ⊗ p

  • For two euros, I get a soup or a pancake (my choice):

e ⊗ e ⊸ s & p

  • For two euros, I get a pancake or an icecream

(cafeteria’s choice): e ⊗ e ⊸ p ⊕ i

  • I am the central bank:

!e

slide-7
SLIDE 7

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Linear λ-calculus

  • the Curry–Howard analog of intuitionistic linear logic
  • values have to be used exactly once:
  • a value can represent the current state of an object
  • changes to the state (destructive updates) expressible

as pure functions

  • some functions with destructive updates:
  • array update:

ι ⊗ α ⊗ Array ι α ⊸ Array ι α

  • opening a file:

FileName ⊗ World ⊸ File ⊗ World

  • writing to an opened file:

String ⊗ File ⊸ File

  • closing a file:

File ⊗ World ⊸ World

slide-8
SLIDE 8

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

1 Linear logic 2 Categorical semantics for linear logic 3 Interaction between linear and non-linear logic 4 References

slide-9
SLIDE 9

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Products and coproducts

  • intuitionistic (non-linear) logic:
  • finite products for ∧ and ⊤
  • finite coproducts for ∨ and ⊥
  • intuitionistic linear logic:
  • finite products for & and ⊤
  • finite coproducts for ⊕ and 0
  • seems strange that ∧/⊤ and &/⊤ are modeled by the same

constructions, although they denote quite different things

  • however, analogous statements hold for ∧/⊤ and &/⊤:

α ⊢ α ∧ α α ⊢ α & α α ∧ β ⊢ α α & β ⊢ α α ⊢ ⊤ α ⊢ ⊤

slide-10
SLIDE 10

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Symmetric monoidal structure

  • axioms of ⊗ and 1:
  • associativity of ⊗:

(α ⊗ β) ⊗ γ ⊢ α ⊗ (β ⊗ γ) α ⊗ (β ⊗ γ) ⊢ (α ⊗ β) ⊗ γ

  • commutativity of ⊗:

α ⊗ β ⊢ β ⊗ α

  • 1 as neutral element:

1 ⊗ α ⊢ α α ⊢ 1 ⊗ α

  • symmetric monoidal structure for ⊗ and 1
slide-11
SLIDE 11

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Adjunctions

  • non-linear logic:
  • cartesian closed structure for ∧ and →
  • −B defined as right-adjoint of − × B
  • corresponds to equivalence of

α ∧ β ⊢ γ and α ⊢ β → γ

  • linear logic:
  • symmetric monoidal closed structure for ⊗ and ⊸
  • B ⊸ − defined as right-adjoint of − ⊗ B
  • corresponds to equivalence of

α ⊗ β ⊢ γ and α ⊢ β ⊸ γ

slide-12
SLIDE 12

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Structure for !

  • symmetric lax monoidal functor structure:

!α ⊗ !β ⊢ !(α ⊗ β) 1 ⊢ !1

  • comonad structure:

!α ⊢ α !α ⊢ !!α

  • commutative comonoid structure:

!α ⊢ !α ⊗ !α !α ⊢ 1

  • some additional coherence conditions
slide-13
SLIDE 13

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

1 Linear logic 2 Categorical semantics for linear logic 3 Interaction between linear and non-linear logic 4 References

slide-14
SLIDE 14

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Linear and non-linear models

  • for now:
  • non-linear logic with only ∧, ⊤, and →
  • linear logic with only ⊗, 1, and ⊸
  • categorical models:

non-linear logic cartesian closed category: (C, ×, 1, →) linear logic symmetric monoidal closed category: (L, ⊗, I, ⊸)

  • beware:

proposition ⊤ ˆ = object 1 proposition 1 ˆ = object I

slide-15
SLIDE 15

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Interaction

  • symmetric lax monoidal adjunction (F, ϕ, ψ) ⊣ (G, υ, ν)

between (L, ⊗, I) and (C, ×, 1):

  • adjunction F ⊣ G between L and C:

F : C → L G : L → C

  • (F, ϕ, ψ) and (G, υ, ν) are symmetric lax monoidal functors

between (L, ⊗, I) and (C, ×, 1): ϕX,Y : FX ⊗ FY → F(X × Y ) ψ : I → F1 υA,B : GA × GB → G(A ⊗ B) ν : 1 → GI

  • unit and counit of F ⊣ G are monoidal transformations
slide-16
SLIDE 16

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Isomorphisms

Theorem

If (F, ϕ, ψ) ⊣ (G, υ, ν) is a lax monoidal adjunction, then ϕ and ψ are isomorphisms.

  • inverses:

ϕ−1

X,Y : F(X × Y ) → FX ⊗ FY

ϕ−1

X,Y = Φ−1(υFX,FY ◦ (ηX × ηY ))

ψ−1 : F1 → I ψ−1 = Φ−1(ν)

  • closer relationship between × and ⊗ as well as 1 and I:

FX ⊗ FY ∼ = F(X × Y ) I ∼ = F1

slide-17
SLIDE 17

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

Derived structure for !

  • adjunction F ⊣ G gives rise to a comonad (!, ε, δ):

! : L → L δ : FG → FGFG ! = FG δ = FηG

  • symmetric monoidal functor structures for F and G

give rise to a symmetric monoidal functor structure for !

  • commutative comonoid structure can be derived:

ξA : FGA → FGA ⊗ FGA ξA = ϕ−1

GA,GA ◦ F∆GA

χA : FGA → I χA = ψ−1 ◦ F!GA

  • further coherence conditions follow
slide-18
SLIDE 18

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

More structure

  • more structure can be required:
  • finite products in L for & and ⊤:

(L, &, ⊤)

  • finite coproducts in C for ∨ and ⊥:

(C, +, 0)

  • finite coproducts in L for ⊕ and 0:

(L, ⊕, 0)

  • no additional coherence conditions
  • interesting properties can still be derived
slide-19
SLIDE 19

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

More isomorphisms

  • right-adjoints preserve limits:

GA × GB ∼ = G(A & B) 1 ∼ = G⊤

  • consequence:

!A ⊗ !B ∼ = !(A & B) I ∼ = !⊤

  • left-adjoints preserve colimits:

FX ⊕ FY ∼ = F(X + Y ) 0 ∼ = F0

slide-20
SLIDE 20

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

1 Linear logic 2 Categorical semantics for linear logic 3 Interaction between linear and non-linear logic 4 References

slide-21
SLIDE 21

Categorical Semantics for Linear Logic Wolfgang Jeltsch Linear logic Categorical semantics for linear logic Interaction between linear and non-linear logic References

References

  • P. N. Benton.

A mixed linear and non-linear logic: Proofs, terms and models. Technical Report UCAM-CL-TR-352, University of Cambridge, Oct. 1994.