Selling Category Theory to the Masses - LMPAsic/wordpress/wp-content/... · 2013-03-09 · Selling Category Theory to the Masses Bob Coecke – Quantum Group - Computer Science -

Selling Category Theory to the MassesBob Coecke – Quantum Group - Computer Science - Oxford University

=

f

f =

f f

f

ALICE

BOB

=

ALICE

BOB

f

=not

likeBobAlice

does

Alice not like

not

Bob

Samson Abramsky & BC (2004) A categorical semantics for quantum protocols. LiCS’04. arXiv:quant-ph/0402130 BC & Eric O. Paquette (2011) Categories for the practicing physicist. In: New Structures forPhysics. arXiv:0905.3010 BC (2010) Quantum picturalism. Contemporary Physics 51, 59–83. arXiv:0908.1787BC & Ross Duncan (2011) Interacting quantum observables. New Journal of Physics 13, 043016. arXiv:0906.4725 BC, Mehrnoosh Sadrzadeh & Stephen Clark (2011) Mathematical foundations for a compositional dis-tributional model of meaning. Linguistic Analysis - Lambek Festschrift. arXiv:1003.4394 Lucas Dixon, R. Dun-can, Aleks Kissinger, Alex Merry & Matvey Soloviev. sites.google.com/site/quantomatic/

. . . a tale of food, spiders and GoogleBob Coecke – Quantum Group - Computer Science - Oxford University

=

f

f =

f f

f

ALICE

BOB

=

ALICE

BOB

f

=not

likeBobAlice

does

Alice not like

not

Bob

Samson Abramsky & BC (2004) A categorical semantics for quantum protocols. LiCS’04. arXiv:quant-ph/0402130 BC & Eric O. Paquette (2011) Categories for the practicing physicist. In: New Structures forPhysics. arXiv:0905.3010 BC (2010) Quantum picturalism. Contemporary Physics 51, 59–83. arXiv:0908.1787BC & Ross Duncan (2011) Interacting quantum observables. New Journal of Physics 13, 043016. arXiv:0906.4725 BC, Mehrnoosh Sadrzadeh & Stephen Clark (2011) Mathematical foundations for a compositional dis-tributional model of meaning. Linguistic Analysis - Lambek Festschrift. arXiv:1003.4394 Lucas Dixon, R. Dun-can, Aleks Kissinger, Alex Merry & Matvey Soloviev. sites.google.com/site/quantomatic/

Task: selling Category Theory to the masses!


. . . is it just tedious abstract nonsense?


. . . is it just tedious abstract nonsense?

No! Categories are everywhere!

1. Let A be a raw potato.

1. Let A be a raw potato.A admits many states e.g. dirty, clean, skinned, ...


2. We want to process A into cooked potato B.B admits many states e.g. boiled, fried, deep fried,baked with skin, baked without skin, ...


2. We want to process A into cooked potato B.B admits many states e.g. boiled, fried, deep fried,baked with skin, baked without skin, ... Let

Af

-B Af ′

-B Af ′′

-B

be boiling, frying, baking.


2. We want to process A into cooked potato B.B admits many states e.g. boiled, fried, deep fried,baked with skin, baked without skin, ... Let

Af

-B Af ′

-B Af ′′

-B

be boiling, frying, baking. States are processes

I := unspecifiedψ

-A.

3. LetA

g ◦ f-C

be the composite process of first boiling Af

-B andthen salting B

g-C.

3. LetA

g ◦ f-C

be the composite process of first boiling Af

-B andthen salting B

g-C. Let

X1X -X

be doing nothing. We have 1Y ◦ ξ = ξ ◦ 1X = ξ.

4. Let A⊗D be potato A and carrot D and let


A⊗D f⊗h-B ⊗ E

be boiling potato while frying carrot.


A⊗D f⊗h-B ⊗ E

be boiling potato while frying carrot. Let

C ⊗ F x-M

be mashing spice-cook-potato and spice-cook-carrot.

5. Total process:

A⊗D f⊗h-B⊗E g⊗k

-C⊗F x-M=A⊗D x◦(g⊗k)◦(f⊗h)

-M.

5. Total process:



-M.

6. Recipe = composition structure on processes.

5. Total process:



-M.


7. Laws governing recipes:

(1B ⊗ g) ◦ (f ⊗ 1C) = (f ⊗ 1D) ◦ (1A ⊗ g)

5. Total process:



-M.



(1B ⊗ g) ◦ (f ⊗ 1C) = (f ⊗ 1D) ◦ (1A ⊗ g)i.e.

boil potato then fry carrot = fry carrot then boil potato

5. Total process:



-M.



(1B ⊗ g) ◦ (f ⊗ 1C) = (f ⊗ 1D) ◦ (1A ⊗ g)i.e.

boil potato then fry carrot = fry carrot then boil potato

⇒ Symmetric Monoidal Category

— Why does a tiger have stripes and a lion doesn’t? —

— Why does a tiger have stripes and a lion doesn’t? —

prey ⊗ predator ⊗ environment

dead prey ⊗ eating predatorhunt

?

BOXES AND WIRES

Roger Penrose (1971) Applications of negative dimensional tensors. In: Com-binatorial Mathematics and its Applications, 221–244. Academic Press.

Andre Joyal and Ross Street (1991) The Geometry of tensor calculus I. Ad-vances in Mathematics 88, 55–112.

Bob Coecke and Eric Oliver Paquette (2011) Categories for the practicingphysicist. In: New Structures for Physics, B. Coecke (ed), Springer-Verlag.arXiv:0905.3010

— wire and box language —

foutput wire(s)

input wire(s)Box =:


foutput wire(s)

input wire(s)Box =:

Interpretation: wire := system ; box := process


foutput wire(s)

input wire(s)Box =:

Interpretation: wire := system ; box := process

one system: n subsystems: no system:

︸︷︷︸1

. . .︸︷︷︸

n︸︷︷︸

0

— wire and box games —


sequential or causal or connected composition:

g ◦ f ≡g

f


sequential or causal or connected composition:

g ◦ f ≡g

f

parallel or acausal or disconnected composition:

f ⊗ g ≡ f fg

— merely a new notation? —


(g ◦ f )⊗ (k ◦ h) = (g ⊗ k) ◦ (f ⊗ h)

=f h

g k

f h

g k


(g ◦ f )⊗ (k ◦ h) = (g ⊗ k) ◦ (f ⊗ h)

=f h

g k

f h

g k

peel potato and then fry it,while,

clean carrot and then boil it=

peel potato while clean carrot,and then,

fry potato while boil carrot

MINIMAL QUANTUM PROCESS LANGUAGE

Samson Abramsky & BC (2004) A categorical semantics for quantum proto-cols. In: IEEE-LiCS’04. quant-ph/0402130

BC (2005) Kindergarten quantum mechanics. quant-ph/0510032

BC (2010) Quantum picturalism. Contemporary Physics. arXiv:0908.1787

— genesis —

[von Neumann 1932] Formalized quantum mechanicsin “Mathematische Grundlagen der Quantenmechanik”

— genesis —


[von Neumann to Birkhoff 1935] “I would like tomake a confession which may seem immoral: I do notbelieve absolutely in Hilbert space no more.” (sic)

— genesis —



[Birkhoff and von Neumann 1936] The Logic ofQuantum Mechanics in Annals of Mathematics.

— genesis —




[1936 – 2000] many followed them, ... and FAILED.

— genesis —




[1936 – 2000] many followed them, ... and FAILED.

— the mathematics of it —


Hilber space stuff: continuum, field structure of com-plex numbers, vector space over it, inner-product, etc.



WHY?



WHY?

von Neumann: only used it since it was ‘available’.

— the physics of it —


von Neumann crafted Birkhoff-von Neumann Quan-tum ‘Logic’ to capture the concept of superposition.



Schrodinger (1935): the stuff which is the true soul ofquantum theory is how quantum systems compose.



Schrodinger (1935): the stuff which is the true soul ofquantum theory is how quantum systems compose.

Quantum Computer Scientists: Schrodinger is right!

— the game plan —


Task 0. Solve:tensor product structure

the other stuff= ???




i.e. axiomatize “⊗” without reference to spaces.





Task 1. Investigate which assumptions (i.e. which struc-ture) on ⊗ is needed to deduce physical phenomena.





Task 1. Investigate which assumptions (i.e. which struc-ture) on ⊗ is needed to deduce physical phenomena.

Task 2. Investigate wether such an “interaction struc-ture” appear elsewhere in “our classical reality”.

Outcome 1a: “Sheer ratio of results to assumptions”

Outcome 1a: “Sheer ratio of results to assumptions”confirms that we are probing something very essential.

Hans Halvorson (2010) Editorial to: Deep Beauty: Understanding the Quan-tum World through Mathematical Innovation, Cambridge University Press.


Hans Halvorson (2010) Editorial to: Deep Beauty: Understanding the Quan-tum World through Mathematical Innovation, Cambridge University Press.


Outcome 1b: Exposing this structure has already helpedto solve open problems elsewhere. (e.g. 2× ICALP’10)

E.g.: Ross Duncan & Simon Perdrix (2010) Rewriting measurement-basedquantum computations with generalised flow. ICALP’10.



Outcome 1c: Framework is a simple intuitive (butrigorous) diagrammatic language, meanwhile adoptedby others e.g. Lucien Hardy in arXiv:1005.5164:

“... we join the quantum picturalism revolution [1]”

[1] BC (2010) Quantum picturalism. Contemporary Physics 51, 59–83.



Outcome 1c: Framework is a simple intuitive (butrigorous) diagrammatic language, meanwhile adoptedby others e.g. Lucien Hardy in arXiv:1005.5164:

“... we join the quantum picturalism revolution [1]”


Outcome 2a:Behaviors of matter (Abramsky-C; LiCS’04, quant-ph/0402130) :

=

f

f =

f f

f

ALICE

BOB

=

ALICE

BOB

f

Meaning in language (Clark-C-Sadrzadeh; Linguistic Analysis, arXiv:1003.4394) :

=

not

likeBobAlice

does

Alice not like

not

Bob

meaning vectors of words

pregroup grammar

Knowledge updating (C-Spekkens; Synthese, arXiv:1102.2368) :

conditionalindependence

=P(C|AB)

A A

=

A

=A

B

A

B=

B(BA) 1-

A

C 1- C 1-

C

P(AB|C) P(A|C) P(B|C) P(C|A) P(C|B)P(C|B) P(C|A)

BOXES AND WIRES II

— quantitative metric —

f : A→ B

f

A

B

— quantitative metric —

f† : B → A

f

B

A

— asserting (pure) entanglement —

quantum

classical=

==

— asserting (pure) entanglement —

quantum

classical=

==

⇒ introduce ‘parallel wire’ between systems:

subject to: only topology matters!

— quantum-like —

E.g.

=

Transpose:

ff

=Conjugate:

ff

=

classical data flow?

f

=

fff


f

=

f


f

=

f


f

ALICE

BOB

=

ALICE

BOB

f

⇒ quantum teleportation

— symbolically: dagger compact categories —

Thm. [Kelly-Laplaza ’80; Selinger ’05] An equa-tional statement between expressions in dagger com-pact categorical language holds if and only if it isderivable in the graphical notation via homotopy.

Thm. [Hasegawa-Hofmann-Plotkin; Selinger ’08]An equational statement between expressions in dag-ger compact categorical language holds if and onlyif it is derivable in the dagger compact category of fi-nite dimensional Hilbert spaces, linear maps, tensorproduct and adjoints.

— kindergarten quantum mechanics: the experiment —

Contest in problem solving between:

• Children using quantum picturalism

• Physics teachers using ordinary QM

I expect the children to win!


A SLIGHTLY DIFFERENT LANGUAGEFOR NATURAL LANGUAGE MEANING

BC, Mehrnoosh Sadrzadeh & Stephen Clark (2010) Mathematical foundationsfor a compositional distributional model of meaning. arXiv:1003.4394

BC (2012) The logic of quantum mechanics – Take II. arXiv:1204.3458

— the from-words-to-a-sentence process —


Consider meanings of words, e.g. as vectors (cf. Google):

word 1 word 2 word n...


What is the meaning the sentence made up of these?

word 1 word 2 word n...


I.e. how do we/machines produce meanings of sentences?

word 1 word 2 word n...?


I.e. how do we/machines produce meanings of sentences?

word 1 word 2 word n...grammar

Gerald Gazdar (1996) Paradigm merger in natural language processing. In:Computing tomorrow: future research directions in computer science, eds.,I. Wand and R. Milner, Cambridge University Press.


Information flow within a verb:

verb

object subject


Information flow within a verb:

verb

object subject

Again we have:

=

— grammar as pregroups – Lambek ’99 —

A Al A A

A Al

r

A Ar

=

A

A

A

A=A

A A

A

r

r

=A

A

A

A=A

A A

Al l

ll

r

r


For noun type n, verb type is −1(n) · s · (n)−1, so:



n · −1(n) · s · (n)−1 · n = s



n · −1(n) · s · (n)−1 · n = s

Diagrammatic typing:

n ns(n) (n)-1 -1



n · −1(n) · s · (n)−1 · n = s

Diagrammatic meaning:

verbn n

flow flow

—−−−→Alice⊗−−→does⊗−→not⊗−−→like⊗−−→Bob —


Alice not like Bob


not

grammar

does


Alice like Bob


grammar

not


Alice like Bob


grammar

not


Alice like Bob


grammar

not= not

like

BobAlice


Alice like Bob


grammar

not= not

like

BobAlice = not

likeBobAlice

Using: =

likelike

=

likelike


Alice like Bob


grammar

not= not

like

BobAlice = not

likeBobAlice

= ~g(~f (~x, ~y)

)

— experiment: word disambiguation —E.g. what is “saw”’ in: “Alice saw Bob with a saw”.

Edward Grefenstette & Mehrnoosh Sadrzadeh (2011) Experimental supportfor a categorical compositional distributional model of meaning. Acceptedfor: Empirical Methods in Natural Language Processing (EMNLP’11).

UNIVERSAL QUANTUM REASONING

— example outputs —New results on resource requirements, complexity oftranslations in MBQC (Duncan-Perdrix ICALP’10):

Example 18. The ubiquitous CNOT operation can be computed by the patternP = X3

4Z24Z2

1M03 M0

2 E13E23E34N3N4 [5]. This yields the diagram,

DP =

H

H

H

!, {3}

!, {2}

!, {2}

!, {3}!, {2}

,

where each qubit is represented by a vertical “path” from top to bottom, withqubit 1 the leftmost, and qubit 4 is the rightmost.

By virtue of the soundness of R and Proposition 10, if DP can be rewrittento a circuit-like diagram without any conditional operations, then the rewritesequence constitutes a proof that the pattern computes the same operation asthe derived circuit.

Example 19. Returning to the CNOT pattern of Example 18, there is a rewritesequence, the key steps of which are shown below, which reduces the DP tothe unconditional circuit-like pattern for CNOT introduced in Example 7. Thisproves two things: firstly that P indeed computes the CNOT unitary, and thatthe pattern P is deterministic.

H

H

H

!, {3}

!, {2}

!, {2}

!, {3}!, {2}

!!H

H

H

!, {3}

!, {2}

!, {2}!, {2} !, {3}

!! H

H

H

!, {3}!, {3}

!, {2}

!, {2}

!, {2}

!!!, {2}

!, {2}!, {2}

!!!, {2}!, {2}

!, {2} !, {2}!!

One can clearly see in this example how the non-determinism introduced bymeasurements is corrected by conditional operations later in the pattern. Thepossibility of performing such corrections depends on the geometry of the pat-tern, the entanglement graph implicitly defined by the pattern.

Definition 20. Let P be a pattern; the geometry of P is an open graph !(P) =(G, I,O) whose vertices are the qubits of P and where i ! j i! Eij occurs in thecommand sequence of P.

Definition 21. Given a geometry " = ((V,E), I, O) we can define a diagramD! = ((VD, ED), ID, OD) as follows:

Similar stuff for TMBQC (Clare Horsman NJP’11):

— automated theory exploration —

Lucas Dixon (Google), Ross Duncan (Brussels), Aleks Kissinger, Alex Merry (Oxf) andMatvey Soloviev (Camb) — http://sites.google.com/site/quantomatic/

WIRES?

SPIDERS!

— spiders —

‘spiders’ =

m︷︸︸︷....

....

︸︷︷︸n

such that, for k > 0:

m+m′−k︷︸︸︷

........

....

....

....

︸︷︷︸n+n′−k

=

....

....

Coecke, Pavlovic & Vicary (2006, 2008) quant-ph/0608035, 0810.0812

— spiders —

‘(co-)mult.’ =

m︷︸︸︷....

....

︸︷︷︸n


m+m′−k︷︸︸︷

........

....

....

....

︸︷︷︸n+n′−k

=

....

....

— spiders —

‘cups/caps’ =

m︷︸︸︷....

....

︸︷︷︸n


m+m′−k︷︸︸︷

........

....

....

....

︸︷︷︸n+n′−k

=

....

....

Coecke, Pavlovic & Vicary (2006, 2008) quant-ph/0608035, 0810.0812

— complementary spiders —

— complementary spiders —Thm.

BC & Ross Duncan (2008) Interacting quantum observables. ICALP’08 &New Journal of Physics 13, 043016. arXiv:0906.4725

Selling Category Theory to the Masses - LMPAsic/wordpress/wp-content/... · 2013-03-09 · Selling Category Theory to the Masses Bob Coecke – Quantum Group - Computer Science -

Documents