Mix Unitary Categories Robin Cockett, Cole Comfort, and Priyaa - - PowerPoint PPT Presentation

mix unitary categories
SMART_READER_LITE
LIVE PREVIEW

Mix Unitary Categories Robin Cockett, Cole Comfort, and Priyaa - - PowerPoint PPT Presentation

Mix Unitary Categories Robin Cockett, Cole Comfort, and Priyaa Srinivasan CT2018, Ponta Delgada, Azores 1/31 CP construction Motivation Mix Unitary Categories Dagger compact closed categories Dagger compact closed categories ( -KCC)


slide-1
SLIDE 1

Mix Unitary Categories

Robin Cockett, Cole Comfort, and Priyaa Srinivasan CT2018, Ponta Delgada, Azores

1/31

slide-2
SLIDE 2

2/31

Motivation Mix Unitary Categories CP∞ construction

Dagger compact closed categories

Dagger compact closed categories (†-KCC) provide a categorical framework for finite-dimensional quantum mechanics. The dagger (†) is a contravariant functor which is stationary on

  • bjects (A = A†) which is an involution (f †† = f ).

In a †-KCC, quantum processes are represented by completely positive maps. The CPM construction on a †-KCC chooses the completely positive maps from the category. FHilb, the category of finite-dimensional Hilbert Spaces and linear maps is the canonical example of a †-KCC. CPM[FHilb] is precisely the category of “quantum processes.”

2 / 31

slide-3
SLIDE 3

3/31

Motivation Mix Unitary Categories CP∞ construction

Finite versus infinite dimensions

For Hilbert Spaces (and additively enriched categories with negatives) compact closed ⇒ finite-dimensionality. Infinite-dimensional Hilbert spaces have a † but do not have “duals”: they are not compact closed. Infinite-dimensional systems occur in many quantum settings including quantum computation and quantum communication. There have therefore been various attempts to generalize the existing structures and constructions to infinite-dimensions.

3 / 31

slide-4
SLIDE 4

4/31

Motivation Mix Unitary Categories CP∞ construction

The CP∞ construction

CP∞ construction (Coecke and Heunen) generalizes the CPM construction to †-symmetric monoidal categories by reexpressing completely positive maps as follows:

(f †)∗ f

f f †

QUESTION: Is there a way to generalize the CPM construction to arbitrary dimensions while retaining duals and the dagger?

4 / 31

slide-5
SLIDE 5

5/31

Motivation Mix Unitary Categories CP∞ construction

Linearly distributive categories

∗-autonomous categories and linearly distributive categories generalize compact closed categories . . . Can quantum ideas be extended in this direction?1 They allow for infinite dimensions, have a nice graphical calculus, allow the expression of “duals” . . . but what about dagger? Recall a linearly distributive category (LDC) has two monoidal structures (⊗, ⊤, a⊗, uL

⊗, uR ⊗) and (⊕, ⊥, a⊕, uL ⊕, uR ⊕) linked by

natural transformations called the linear distributors: ∂L : A ⊗ (B ⊕ C) → (A ⊗ B) ⊕ C ∂R : (A ⊕ B) ⊗ C → A ⊕ (B ⊗ C)

1See Dusko Pavlovic “Relating Toy Models of Quantum Computation: Comprehension, Complementarity and Dagger Mix Autonomous Categories”

5 / 31

slide-6
SLIDE 6

6/31

Motivation Mix Unitary Categories CP∞ construction

Mix categories

A mix category is a LDC with a map m : ⊥ − → ⊤ in X such that mxA,B : A ⊗ B − → A ⊕ B :=

m

=

m

mx is called a mix map. The mix map is a natural transformation. It is an isomix category if m is an isomorphism. m being an isomorphism does not make the mx map an isomorphism.

6 / 31

slide-7
SLIDE 7

7/31

Motivation Mix Unitary Categories CP∞ construction

The Core of mix category

The core of a mix category, Core(X) ⊆ X, is the full subcategory determined by objects U ∈ X for which the natural transformation is also an isomorphism: U ⊗ ( )

mxU,( )

− − − − − → U ⊕ ( ) The core of a mix category is closed to ⊗ and ⊕. The core of an isomix category contains the monoidal units ⊤ and ⊥ and is a compact LDC (meaning tensor and par are essentially identical) .

7 / 31

slide-8
SLIDE 8

8/31

Motivation Mix Unitary Categories CP∞ construction

Roadmap

LDC

  • Define † -LDC
  • Define unitary isomorphisms in † -LDCs
  • Generalize CP∞ construction for † -LDCs

8 / 31

slide-9
SLIDE 9

9/31

Motivation Mix Unitary Categories CP∞ construction

The †?

The definition of † : Xop − → X as stationary on objects cannot be imported to LDCs because the dagger minimally has to flip the tensor products: (A ⊗ B)† = A† ⊕ B†. Why? If the dagger is identity-on-objects, then the linear distributor degenerates to an associator: (δR)† : (A ⊕ (B ⊗ C))† − → ((A ⊕ B) ⊗ C)† (δR)† : A† ⊕ (B† ⊗ C †) − → (A† ⊕ B†) ⊗ C †

9 / 31

slide-10
SLIDE 10

10/31

Motivation Mix Unitary Categories CP∞ construction

†-LDCs

A †-LDC is a LDC X with a dagger functor † : Xop − → X and the natural isomorphisms: tensor laxors: λ⊕ : A† ⊕ B† − → (A ⊗ B)† λ⊗ : A† ⊗ B† − → (A ⊕ B)† unit laxors: λ⊤ : ⊤ − → ⊥† λ⊥ : ⊥ − → ⊤† involutor: ι : A − → A†† which make † a contravariant (Frobenius) linear equivalence.

10 / 31

slide-11
SLIDE 11

11/31

Motivation Mix Unitary Categories CP∞ construction

Coherences for †-LDCs

Coherences for the interaction between the tensor laxors and the basic natural isomorphisms (6 coherences): A† ⊗ (B† ⊗ C †)

a⊗

  • 1⊗λ⊗
  • (A† ⊗ B†) ⊗ C †

λ⊗⊗1

  • (A† ⊗ (B ⊕ C)†)

λ⊗

  • (A ⊕ B)† ⊗ C †

λ⊗

  • (A ⊕ (B ⊕ C))†

(a−1

⊕ )† ((A ⊕ B) ⊕ C)†

11 / 31

slide-12
SLIDE 12

12/31

Motivation Mix Unitary Categories CP∞ construction

Coherences for †-LDCs (cont.)

Interaction between the unit laxors and the unitors (2 coherences): ⊤ ⊗ A†

λ⊤⊗1

  • ul

⊥† ⊗ A†

λ⊗

  • A†

(⊥ ⊕ A)†

  • (ul

⊕)†

⊥ ⊕ A†

λ⊥⊕1

  • ul

⊤† ⊕ A†

λ⊕

  • A†

(⊤ ⊗ A)†

  • (ul

⊗)†

Interaction between the involutor and the laxors (4 coherences): A ⊕ B

ι

  • i⊕i
  • ((A ⊕ B)†)†

λ†

  • (A†)† ⊕ (B†)†

λ⊕

(A† ⊗ B†)†

ι λ⊥

  • (⊥†)†

λ†

  • ⊤†

12 / 31

slide-13
SLIDE 13

13/31

Motivation Mix Unitary Categories CP∞ construction

Diagrammatic calculus for †-LDC

Extend the diagrammatic calculus of LDCs The action of dagger is represented using dagger boxes: † :

A B f

f A B A† B†

.

13 / 31

slide-14
SLIDE 14

14/31

Motivation Mix Unitary Categories CP∞ construction

Isomix †-LDCs

A mix †-LDC is a †-LDC with m : ⊥ − → ⊤ such that: ⊥

m

  • λ⊥

λ⊤

  • ⊤†

m† ⊥†

If m is an isomorphism, then X is an isomix †-LDC. Lemma: The following diagram commutes in a mix †-LDC: A† ⊗ B†

mx

  • λ⊗

A† ⊕ B†

λ⊕

  • (A ⊕ B)†

mx† (A ⊗ B)†

14 / 31

slide-15
SLIDE 15

15/31

Motivation Mix Unitary Categories CP∞ construction

Isomix †-LDCs

Lemma: Suppose X is a mix †-LDC and A ∈ Core(X) then A† ∈ Core(X). Proof: The natural transformation A† ⊗ X

mx

− − → A† ⊕ X is an isomorphism: A† ⊗ X

1⊗ι mx

  • nat. mx

A† ⊗ X ††

λ⊗ mx

  • Lemma above

(A ⊕ X †)†

mx†

  • A† ⊕ X

1⊕ι

A† ⊕ A††

λ⊕

(A ⊗ X †)†

commutes.

15 / 31

slide-16
SLIDE 16

16/31

Motivation Mix Unitary Categories CP∞ construction

Next step: Unitary structure

Define †-LDC Define unitary isomorphisms The usual definition of unitary maps (f † : B† → A† = f −1 : B − → A)

  • nly works when the † functor is stationary on objects.

16 / 31

slide-17
SLIDE 17

17/31

Motivation Mix Unitary Categories CP∞ construction

Unitary structure

An isomix †-LDC has unitary structure in case there is a small class of objects called unitary objects such that: Every unitary object, A ∈ U, is in the core; The dagger of a unitary object is unitary; Each unitary object A ∈ U comes equipped with an isomorphism, the unitary structure of A,

A A† : A

ϕA

− → A† such that

A† A†† A† A†† = = ι A A† A†† A A††

ϕA† = ((ϕA)−1)† (ϕAϕA†) = ι

17 / 31

slide-18
SLIDE 18

18/31

Motivation Mix Unitary Categories CP∞ construction

Unitary structure (cont.)

⊤, ⊥ are unitary objects with: ϕ⊥ = mλ⊤ ϕ⊤ = m−1λ⊥ If A and B are unitary objects then A ⊗ B and A ⊕ B are unitary objects such that: (ϕA ⊗ ϕB)λ⊗ = mx ϕA⊕B : A ⊗ B − → (A ⊗ B)† ϕA⊗Bλ−1

⊕ = mx(ϕA ⊕ ϕB) : A ⊗ B −

→ A† ⊕ B†

⊥ ⊤ = ⊥ ⊤ m

= ϕ⊥λ−1

⊤ = m

(ϕA ⊗ ϕB)λ⊗ = mx ϕA⊕B

18 / 31

slide-19
SLIDE 19

19/31

Motivation Mix Unitary Categories CP∞ construction

Mix Unitary Category (MUC)

An iso-mix †-LDC with unitary structure is a mix unitary category (MUC). The unitary objects of a MUC, X, determine a full subcategory, UCore(X) ⊆ X, called the unitary core. The unitary core is a unitary category. Remark: In order to obtain the right functorial properties a (general) MUC is an isomix †-category with a full and faithful structure preserving inclusion of a unitary category.

19 / 31

slide-20
SLIDE 20

20/31

Motivation Mix Unitary Categories CP∞ construction

Unitary isomorphisms

Suppose A and B are unitary objects. An isomorphism A f − → B is said to be a unitary isomorphism if the following diagram commutes:

A B B† A† f f

=

A A†

f ϕBf † = ϕA Lemma: In a MUC f † is a unitary map iff f is; f ⊗ g and f ⊕ g are unitary maps whenever f and g are. a⊗, a⊕, c⊗, c⊕, δL, m, and mx are unitary isomorphisms. λ⊗, λ⊕, λ⊤, λ⊥, and ι are unitary isomorphisms. ϕA is a unitary isomorphisms for for all unitary objects A. 20 / 31

slide-21
SLIDE 21

21/31

Motivation Mix Unitary Categories CP∞ construction

Five examples of MUCs

†-KCC These give a compact closed MUC with a stationary dagger and trivial unitary structure FFVecC “Framed” vector spaces (vector spaces with a chosen basis) is a compact closed MUC with non-trivial unitary structure. FinC C-modules over finiteness spaces is a ∗-autonomous category: maps are infinite dimensional matrices with composition controlled (by types) to avoid infinite

  • sums. The unitary subcategory is just Mat(C).

Bicomp(X) The bicompletion of a †-KCC, X is a mix †-∗-autonomous category with unitary objects in X. ChuY(I) The Chu construction on a symmetric monoidal closed category with conjugation, with dualizing

  • bject the unit I, gives a MUC.

21 / 31

slide-22
SLIDE 22

22/31

Motivation Mix Unitary Categories CP∞ construction

Next step: CP∞ construction on MUCs

Define †-LDC Define unitary isomorphisms Examples CP∞ construction on MUC

22 / 31

slide-23
SLIDE 23

23/31

Motivation Mix Unitary Categories CP∞ construction

Krauss maps

In a MUC, a map f : A − → U ⊕ B of X where U is a unitary object is called a Krauss map f : A →U B. U is called the ancillary system of f . In a MUC, quantum processes are represented using Krauss maps as follows:

f

:=

f f A U U† A† B B†

analogous to

A f f † A† U

in †-SMCs. A f − → U ⊕ B

mx−1

− − − → U ⊗ B

ϕ⊗1

− − → U† ⊗ B U† ⊗ B† λ⊗ − − → (U ⊕ B)† f † − → A†

23 / 31

slide-24
SLIDE 24

24/31

Motivation Mix Unitary Categories CP∞ construction

Combinator and test maps

Two Krauss maps f : A →U1 B and g : A →U2 B are equivalent, f ∼ g, if for all test maps h : B ⊗ X → V where V is an unitary

  • bject, the following equation holds:

h h f

=

h h g

Lemma: Let f : A →U1 B and f ′ : A →U2 B be Krauss maps such that U1

α

− → U2 is a unitary isomorphism with f ′ = (α ⊕ 1)f , then f ∼ f ′. In this case, f is said to be unitarily isomorphic to f ′.

24 / 31

slide-25
SLIDE 25

25/31

Motivation Mix Unitary Categories CP∞ construction

CP∞ construction

Given a MUC, X, define CP∞(X) to have: Objects: as of X Maps: CP∞(X)(A, B) := {f ∈ X(A, U⊕B)|U ∈ X and U is unitary}/ ∼ Composition:

f g

Identity: A

(uL

⊕)−1

− − − − → ⊥ ⊕ A ∈ X

25 / 31

slide-26
SLIDE 26

26/31

Motivation Mix Unitary Categories CP∞ construction

Tensor and Par

CP∞(X) inherits tensor and par from X: f ⊗g :=

f g

f ⊕g :=

f g

  • ⊤ := ⊤
  • ⊥ := ⊥

26 / 31

slide-27
SLIDE 27

27/31

Motivation Mix Unitary Categories CP∞ construction

Linear adjoints

Suppose X is a LDC and A, B ∈ X. Then, B is left linear adjoint to A (η, ε) : B ⊣ ⊣ A, if there exists η : ⊤ → B ⊕ A ε : A ⊗ B → ⊥ such that the following triangle equalities hold: B

(uL

⊗)−1

⊤ ⊗ B

η⊗1

(B ⊕ A) ⊗ B

∂R

  • B

B ⊕ ⊥

uR

  • B ⊕ (A ⊗ B)

1⊕ε

  • A

(uR

⊗)−1

A ⊗ ⊤

1⊗η

A ⊗ (B ⊕ A)

∂L

  • A

⊥ ⊕ A

uL

  • (A ⊗ B) ⊕ A

ε⊕1

  • η

ε

=

ε η

= When every object of a MUC has a linear adjoint, it is called a ∗- MUC.

27 / 31

slide-28
SLIDE 28

28/31

Motivation Mix Unitary Categories CP∞ construction

Unitary linear adjoints

A unitary linear adjoint (η, ε) : A ⊣ ⊣ u B is a linear adjoint, A ⊣ ⊣ B with A and B being unitary objects satisfying: ηA(ϕA ⊕ ϕB)c⊕ = λ⊤ε†λ−1

(ϕA ⊗ ϕB)λ⊗η†

A = c⊗εAλ⊥

ε

=

η

λ⊤ε†λ−1

⊕ = ηc⊕(ϕA ⊕ ϕB)

A MUC in which every unitary object has a unitary linear adjoint is called a MUdC.

28 / 31

slide-29
SLIDE 29

29/31

Motivation Mix Unitary Categories CP∞ construction

Dagger functor for CP∞(X)

Proposition: If X is a ∗-MUdC, then CP∞(X) is a ∗-MUdC. Sketch of proof: Suppose f : A − → U ⊕ B and (η, ε) : V ⊣ ⊣ u U † : CP∞(X)op − → CP∞(X);

f

f ε V † B† A†

Unitary structure and unitary linear adjoints are preserved due to the functoriality of Q.

29 / 31

slide-30
SLIDE 30

30/31

Motivation Mix Unitary Categories CP∞ construction

Summary

Generalized one important construction of categorical quantum mechanics to MUCs!

1 Mix Unitary Categories are †-LDCs with unitary subcategory. 2 There is a diagrammatic calculus for MUCs. 3 When unitary objects have unitary linear adjoint, then the

unitary core is a dagger compact closed category.

4 CP∞ on MUCs generalizes CP∞ construction on †-SMCs

(auxillary systems in the unitary core).

5 The construction is functorial and produces a *-MUdC when

every (unitary) object has a (unitary) linear adjoint.

30 / 31

slide-31
SLIDE 31

31/31

Motivation Mix Unitary Categories CP∞ construction

Bibliography

LDC: Robin Cockett, and Robert Seely. Weakly distributive

  • categories. Journal of Pure and Applied Algebra 114.2 (1997): 133-173.

The core of a mix category: Richard Blute, Robin Cockett, and Robert

  • Seely. Feedback for linearly distributive categories: traces and
  • fixpoints. Journal of Pure and Applied Algebra 154.1-3 (2000): 27-69.

Graphical calculus for LDCs: Richard Blute, Robin Cockett , Robert Seely, and Todd Trimble. Natural deduction and coherence for weakly distributive categories. Journal of Pure and Applied Algebra 113.3 (1996): 229-296. †-KCC and the CPM construction Peter Selinger. Dagger compact closed categories and completely positive maps. Electronic Notes in Theoretical computer science 170 (2007): 139-163. CP∞ construction on †-SMCs: Bob Coecke, and Chris Heunen. Pictures

  • f complete positivity in arbitrary dimension. Information and

Computation 250 (2016): 50-58.

31 / 31