Visualizing PML David Dumas University of Illinois at Chicago June - - PowerPoint PPT Presentation

visualizing pml
SMART_READER_LITE
LIVE PREVIEW

Visualizing PML David Dumas University of Illinois at Chicago June - - PowerPoint PPT Presentation

Visualizing PML David Dumas University of Illinois at Chicago June 30, 2016 The PML Visualization Project dumas.io/PML Joint work with Franois Guritaud (Univ. Lille) I will also demonstrate 3D graphics sofuware developed by Gilbert.


slide-1
SLIDE 1

Visualizing PML

David Dumas

University of Illinois at Chicago

June 30, 2016

slide-2
SLIDE 2

The PML Visualization Project

dumas.io/PML

Joint work with François Guéritaud (Univ. Lille)

I will also demonstrate 3D graphics sofuware developed by UIC undergraduate researchers Galen Ballew and Alexander Gilbert.

Mathematical Computing Laboratory

slide-3
SLIDE 3

What is PML?

The space of Projective Measured Laminations

A completion of the set C of simple closed curves on S Homeomorphic to SN−1, where N = dim(T) Piecewise linear structure, PL action of Mod(S)

{ , , }

slide-4
SLIDE 4

Linear analogy

The inclusions C ֒ → ML (discrete image) C ֒ → PML (dense image) are analogous to primitive(ZN) ֒ → RN (discrete image) primitive(ZN) ֒ → SN−1 (dense image)

slide-5
SLIDE 5

Linear visualization

slide-6
SLIDE 6

Linear visualization

slide-7
SLIDE 7

Linear visualization

slide-8
SLIDE 8

Linear visualization

slide-9
SLIDE 9

Linear visualization

slide-10
SLIDE 10

Linear visualization

slide-11
SLIDE 11

Linear visualization

slide-12
SLIDE 12

Linear visualization

slide-13
SLIDE 13

Linear visualization

slide-14
SLIDE 14

Linear visualization

slide-15
SLIDE 15

Linear visualization

slide-16
SLIDE 16

Linear visualization

slide-17
SLIDE 17

Not so fast

Can we visualize PML similarly? Several issues: Need to choose an identification ML ≃ RN. (Train tracks? Dehn-Thurston? Something else?) The “small” values of N = 6g − 6 + 2n are

N=2 for S0,4 and S1,1 N=4 for S0,5 and S1,2

slide-18
SLIDE 18

Stereographic projection

slide-19
SLIDE 19

Stereographic projection

slide-20
SLIDE 20

Stereographic projection

slide-21
SLIDE 21

Stereographic projection

slide-22
SLIDE 22

Stereographic projection

slide-23
SLIDE 23

Stereographic projection

slide-24
SLIDE 24

Thurston’s embedding

Fix X ∈ T(S), the base hyperbolic structure.

PML → T∗

XT(S)

[λ] → dX log(ℓλ)

Curve α ∈ C maps to a vector representing the sensitivity of its geodesic length to deformations of the hyperbolic structure X.

slide-25
SLIDE 25

Thurston’s drawing of PML

From “Minimal stretch maps between hyperbolic surfaces”, preprint, 1986.

slide-26
SLIDE 26

Punctured torus

S1,1

slide-27
SLIDE 27

Punctured torus

slide-28
SLIDE 28

Punctured torus

slide-29
SLIDE 29

Punctured torus

slide-30
SLIDE 30

Five-punctured sphere

S0,5

slide-31
SLIDE 31

pmls05-001

slide-32
SLIDE 32

Earthquake basis

R2

R2

slide-33
SLIDE 33

Rotating the pole

pmls05-010

slide-34
SLIDE 34

Closer?

pmls05-020

slide-35
SLIDE 35

Clifgord flow

pmls05-030

slide-36
SLIDE 36

Back to the linear analogy

It is “easy” to imagine Z4. What about its stereographic projection? And can this inform our understanding of the PML(S0,5) images?

slide-37
SLIDE 37

z4-011

slide-38
SLIDE 38

Rings

slide-39
SLIDE 39

Rings

slide-40
SLIDE 40

Rings

slide-41
SLIDE 41

pmls05-071

slide-42
SLIDE 42

Contact

David Dumas

david@dumas.io

slide-43
SLIDE 43

Visualizing PML

David Dumas

University of Illinois at Chicago

July 1, 2016

slide-44
SLIDE 44

The PML Visualization Project

dumas.io/PML

Joint work with François Guéritaud (Univ. Lille)

I will also demonstrate 3D graphics sofuware developed by UIC undergraduate researchers Galen Ballew and Alexander Gilbert.

Mathematical Computing Laboratory

slide-45
SLIDE 45

Five-punctured sphere

S0,5

slide-46
SLIDE 46

Earthquake basis

R2

R2

slide-47
SLIDE 47

Observations

Already apparent: Features related to short curves dominate Lots of “filaments”; all have corners Exploring variations and alternatives, we also found: Several choices for simple curve cutofgs give visually indistinguishable results “First person” perspective from the antipode is theoretically natural, but feels too limiting in pre-rendered animations

slide-48
SLIDE 48

pmls05-001

slide-49
SLIDE 49

Rotating the pole

pmls05-010

slide-50
SLIDE 50

Closer?

pmls05-020

slide-51
SLIDE 51

Clifgord flow

pmls05-030

slide-52
SLIDE 52

Rings

slide-53
SLIDE 53

Rings

slide-54
SLIDE 54

Rings

slide-55
SLIDE 55

pmls05-071

slide-56
SLIDE 56

Rotating the pole

pmls05-041

slide-57
SLIDE 57

Rotating the pole II

pmls05-061

slide-58
SLIDE 58

Twists

slide-59
SLIDE 59

Twists

slide-60
SLIDE 60

Twists

slide-61
SLIDE 61

Twists

slide-62
SLIDE 62

Twists

slide-63
SLIDE 63

Twists

slide-64
SLIDE 64

pmls05-081

slide-65
SLIDE 65

Twists

slide-66
SLIDE 66

Rings poster

PDF for full-size printing at: dumas.io/PML/

slide-67
SLIDE 67

Unity 3D Demo

By Galen Ballew and Alexander Gilbert, undergraduate researchers in UIC’s Mathematical Computing Laboratory.

slide-68
SLIDE 68

Toolchain

POV-Ray

Linux, Emacs, GNU Parallel, fgmpeg, ... Unity 3D, Oculus Rifu, WebGL, ...

slide-69
SLIDE 69

Toolchain

POV-Ray

Linux, Emacs, GNU Parallel, fgmpeg, ... Unity 3D, Oculus Rifu, WebGL, ...

slide-70
SLIDE 70

Toolchain

POV-Ray

Linux, Emacs, GNU Parallel, fgmpeg, ... Unity 3D, Oculus Rifu, WebGL, ...

slide-71
SLIDE 71

Toolchain

POV-Ray

Linux, Emacs, GNU Parallel, fgmpeg, ... Unity 3D, Oculus Rifu, WebGL, ...

slide-72
SLIDE 72

Toolchain

POV-Ray

Linux, Emacs, GNU Parallel, fgmpeg, ... Unity 3D, Oculus Rifu, WebGL, ...

slide-73
SLIDE 73

Process

1 Fuchsian representation 2 Cocycle basis 3 Enumerate simple closed curves 4 Covectors 5 Spheres 6 Ray-tracing 7 Encoding / post-processing

slide-74
SLIDE 74

Fuchsian representation

A description of the base hyperbolic structure X in a form that allows computation of lengths. Typical (e.g. S0,5): 2 × 2 matrix generators for the Fuchsian group Alternative (e.g. S1,1): Sufgiciently many traces of elements to determine the Fuchsian representation up to conjugacy

slide-75
SLIDE 75

Cocycle basis

A basis for TxT(S) represented in same form as the base hyperbolic structure. Write a family of representations ρt : π1S → SL2R as ρt(γ) = ( Id2×2 + t u(γ) + O(t2) ) ρ0(γ). Then u : π1S → Mat2×2R is a cocycle representing the tangent vector d

dtρt

  • t=0to T(S).
slide-76
SLIDE 76

Simple closed curves

Homotopy classes of closed curves are conjugacy classes in the group π1(S). Of these, we only want the simple ones. Procedure: Start with a few “seed” words (known to be simple) Generate more curves by applying mapping classes Repeat until a stopping condition attained, e.g.

Max word length Max hyperbolic length Max depth in Mod(S)

slide-77
SLIDE 77

Covectors

Hyperbolic translation length ℓ of an element A ∈ SL2R: ℓ = 2arccosh(1 2tr(A)) For each word w representing a simple curve α and for a basis

  • f cocycles ui:

Compute length of w at X and at X + ϵui Difgerence quotient approximates dlength(α) dui i.e. component i of the d(length) covector. Divide by length at X to get d(log(length))

slide-78
SLIDE 78

Spheres

In S0,5 case we now have a list of tuples (w, ℓ, dℓ du1 , dℓ du2 , dℓ du3 , dℓ du4 ) which in practice might look like:

acADaCbcd 22.5373 -0.6807 0.6506 -0.8551 0.3537

Stereographic projection of the 4-vector gives the center and a negative power of ℓ gives the radius. Generate a POV-Ray sphere primitive:

sphere { <-1.001967,-1.154298,0.477426>, 0.014278 }

slide-79
SLIDE 79

Ray-tracing and encoding

A POV-Ray scene file sets background, lighting, camera parameters and imports the list of spheres generated from the covectors. For animations: Iterate over a list of parameter values for stereographic projection, camera position, etc. to make a series of frame images. Compress/encode frame images to h.264/mp4 video with fgmpeg.

slide-80
SLIDE 80

Ray-tracing and encoding

Along the way, we made a fgmpeg frontend for encoding video from a series of frame images. Features: Read image file names from a “manifest” file Simplified option syntax

http://github.com/daviddumas/ddencode/

slide-81
SLIDE 81

PML rendering demo

Code at http://github.com/daviddumas/pmls05-demo/

slide-82
SLIDE 82

Glass cube

Laser engraving with technical assistance from Bathsheba Grossman

slide-83
SLIDE 83

3-punctured projective plane

N1,3 = Non-orientable surface with 1 crosscap and 3 punctures. Teichmüller space has dimension 3, so PML ≃ S2! Has one-sided and two-sided simple curves. Scharlemann: One-sided curves are isolated points of the image of C Two-sided curves are dense in a gasket, which is also the limit set of the one-sided curves Open problem: Compute Hausdorfg dimension of this gasket in PL coordinates or in the Thurston embedding.

slide-84
SLIDE 84

n13-010

slide-85
SLIDE 85

Thurston’s drawing of PML

From “Minimal stretch maps between hyperbolic surfaces”, preprint, 1986.

slide-86
SLIDE 86

Added in proof (afuer the lecture): There were questions about minimal but non-uniquely ergodic laminations. None of the pictures show these directly. Such laminations exist on S0,5 but I do not know whether they exist on N1,3. I suspect not.

slide-87
SLIDE 87

Contact

David Dumas

david@dumas.io