Unifying Cubical Models of Univalent Type Theory Evan Cavallo - - PowerPoint PPT Presentation

unifying cubical models of univalent type theory
SMART_READER_LITE
LIVE PREVIEW

Unifying Cubical Models of Univalent Type Theory Evan Cavallo - - PowerPoint PPT Presentation

Unifying Cubical Models of Univalent Type Theory Evan Cavallo Anders Mrtberg Carnegie Mellon University Stockholm University Andrew W Swan University of Amsterdam CSL 2020 JAN 16 0 Univalent Type Theory Dependent type theory CSL


slide-1
SLIDE 1

CSL 2020 · JAN 16

Unifying Cubical Models

  • f Univalent Type Theory

Anders Mörtberg

Stockholm University

Evan Cavallo

Carnegie Mellon University

Andrew W Swan

University of Amsterdam

slide-2
SLIDE 2

CSL 2020 · JAN 16 1

▧ Dependent type theory

Univalent Type Theory

slide-3
SLIDE 3

CSL 2020 · JAN 16 1

▧ Dependent type theory

Univalent Type Theory

funcon/implicaon/∀ product/∃ inducve types equality universe(s) of types

slide-4
SLIDE 4

CSL 2020 · JAN 16 1

▧ Dependent type theory

Univalent Type Theory

funcon/implicaon/∀ product/∃ inducve types equality universe(s) of types

slide-5
SLIDE 5

CSL 2020 · JAN 16 2

▧ Identy

+

◈ Least reflexive relaon (⇒ symmetric, transive, etc.) ◈ “ Underdetermined”

Univalent Type Theory

slide-6
SLIDE 6

CSL 2020 · JAN 16 3

▧ Equivalence

Univalent Type Theory

▧ Univalence Axiom (Voevodsky)

slide-7
SLIDE 7

CSL 2020 · JAN 16 4

▧ Idenes are not unique

Univalent Type Theory

slide-8
SLIDE 8

CSL 2020 · JAN 16 4

▧ Idenes are not unique

Univalent Type Theory

slide-9
SLIDE 9

CSL 2020 · JAN 16 4

▧ Idenes are not unique

Univalent Type Theory

▧ More: add higher inducve types

◈ Quoents for proof-relevant identy ◈ Language for synthec homotopy theory

slide-10
SLIDE 10

CSL 2020 · JAN 16 5

◈ Classical seng for homotopy theory

Models of Univalent Type Theory

▧ Simplicial set model (Kapulkin & Lumsdaine ’12/’18, aer Voevodsky)

◈ Essenally non-construcve (Bezem, Coquand, & Parmann ’15) ◈ First construcve model of univalence

▧ Cubical set model (Bezem, Coquand, & Huber ’13)

◈ Problems with higher inducve types resolved in Cohen, Coquand, Huber, & Mörtberg ’15 and Angiuli, Favonia, & Harper ’18 models

slide-11
SLIDE 11

CSL 2020 · JAN 16 6

Cubical Set Models

▧ Interpret contexts as cubical sets

{ } { } { }

◈ family of sets indexed by interval variable contexts

slide-12
SLIDE 12

CSL 2020 · JAN 16 7

Cubical Set Models

{ } { }

▧ Interpret contexts as cubical sets

◈ family of sets indexed by interval variable contexts ◈

slide-13
SLIDE 13

CSL 2020 · JAN 16 8

Cubical Set Models

▧ Interpret contexts as cubical sets

◈ family of sets indexed by interval variable contexts ◈

▧ Interpret types as fibraons

slide-14
SLIDE 14

CSL 2020 · JAN 16 9

Fibraons

▧ Part 1 (coercion): then “if ”

slide-15
SLIDE 15

CSL 2020 · JAN 16 9

Fibraons

▧ Part 1 (coercion): then “if ”

slide-16
SLIDE 16

CSL 2020 · JAN 16 9

Fibraons

▧ Part 1 (coercion): then “if ”

slide-17
SLIDE 17

CSL 2020 · JAN 16 9

Fibraons

▧ Part 1 (coercion): then “if ”

slide-18
SLIDE 18

CSL 2020 · JAN 16 10

Fibraons

▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-19
SLIDE 19

CSL 2020 · JAN 16 10

Fibraons

▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-20
SLIDE 20

CSL 2020 · JAN 16 10

Fibraons

▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-21
SLIDE 21

CSL 2020 · JAN 16 10

Fibraons

▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-22
SLIDE 22

CSL 2020 · JAN 16 10

Fibraons

▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-23
SLIDE 23

CSL 2020 · JAN 16 11

Fibraons

▧ A fibraon is a family supporng these operaons ▧ Part 1 (coercion): ▧ Part 2 (composion): a cube in A can be adjusted then “if ”

slide-24
SLIDE 24

CSL 2020 · JAN 16 12

Two approaches

Cohen, Coquand, Huber, & Mörtberg ’15 ▧ Result: fibraons closed under type formers ▧

slide-25
SLIDE 25

CSL 2020 · JAN 16 13

Two approaches

Angiuli, Favonia, & Harper ’18 ▧ Result: fibraons closed under type formers ▧

slide-26
SLIDE 26

CSL 2020 · JAN 16 14

Two approaches

CCHM ▧

AFH ▧

◈ ◈ ◈

Is there a unifying construcon that generalizes these?

slide-27
SLIDE 27

CSL 2020 · JAN 16 15

Unifying construcon

Q: A:

slide-28
SLIDE 28

CSL 2020 · JAN 16 15

Unifying construcon

Q: A:

slide-29
SLIDE 29

CSL 2020 · JAN 16 15

Unifying construcon

Q: A:

slide-30
SLIDE 30

CSL 2020 · JAN 16 15

Unifying construcon

Q: A:

slide-31
SLIDE 31

CSL 2020 · JAN 16 15

Unifying construcon

Q: A: IDEA (CMS):

slide-32
SLIDE 32

CSL 2020 · JAN 16 16

Unifying construcon

Fibraons are closed under type formers ▧ Fibraons parcipate in a model structure ▧

slide-33
SLIDE 33

CSL 2020 · JAN 16 17

Unifying construcon

Parameterized by category C with 𝕁 and Φ (+ axioms) ▧ AFH = CMS(

) , ,

𝕁

CCHM = CMS(

) , ,

𝕁

Also new models, e.g. cartesian w/ only faces in Φ ▧

slide-34
SLIDE 34

CSL 2020 · JAN 16 18

Unifying construcon

◈ assume C interprets ordinary type theory

Formulated following Orton & Pis ’16 (for CCHM),

Angiuli, Brunerie, Coquand, Favonia, Harper, & Licata ’18 (for AFH)

◈ describe axioms and construcon in internal language ◈ enables straighorward formalizaon (ours in Agda)

slide-35
SLIDE 35

CSL 2020 · JAN 16 19

Unifying construcon

Model structure ▧

◈ seng for homotopy theory ◈ following Saler ’17 (for CCHM) ◈ use Swan ’18 to translate coercion r → s

{

C (cofibraons): generated by Φ W (weak equivalences): equivalences F (fibraons): fibraons ◈ Our (C, W, F) has F maximal such that families in F have coercion 0 → r

slide-36
SLIDE 36

CSL 2020 · JAN 16 20

Future work

Original cubical model: Bezem, Coquand, & Huber ’13 ▧

◈ Substructural: no diagonal maps between cubes

{ } { }

◈ Definions of fibraon structure for types rely on the absence of diagonals

How do cubical models relate to other models? ▧