Empty Monochromatic Simplices Oswin Aichholzer 1 , Ruy Fabila-Monroy - - PowerPoint PPT Presentation

empty monochromatic simplices
SMART_READER_LITE
LIVE PREVIEW

Empty Monochromatic Simplices Oswin Aichholzer 1 , Ruy Fabila-Monroy - - PowerPoint PPT Presentation

Institute for Software Technology Graz University of Technology P 23629N18 Empty Monochromatic Simplices Oswin Aichholzer 1 , Ruy Fabila-Monroy 2 , naloza 3 , Thomas Hackl 1 , Clemens Huemer 4 , David Flores-Pe and Jorge Urrutia 3 1


slide-1
SLIDE 1

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

1

Empty Monochromatic Simplices

Oswin Aichholzer1, Ruy Fabila-Monroy2, David Flores-Pe˜ naloza3, Thomas Hackl1, Clemens Huemer4, and Jorge Urrutia3

1 Institute for Software Technology, Graz University of Technology 2 Departamento de Matem´

aticas, Cinvestav, Mexico City, Mexico

3 Instituto de Matem´

aticas, UNAM, Mexico City, Mexico

4 Departament de Matem`

atica Aplicada IV, UPC, Barcelona, Spain This presentation is partly funded by EUROCORES - EuroGiga: ComPoSe.

slide-2
SLIDE 2

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
slide-3
SLIDE 3

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
  • ”Roadmap”
slide-4
SLIDE 4

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
  • ”Roadmap”
  • Large sized triangulations
slide-5
SLIDE 5

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
  • ”Roadmap”
  • Large sized triangulations
  • ”Technical” results
  • Pulling complexes
  • Generalized ”Order” and ”Discrepancy” lemmata
slide-6
SLIDE 6

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
  • ”Roadmap”
  • Large sized triangulations
  • ”Technical” results
  • Pulling complexes
  • Generalized ”Order” and ”Discrepancy” lemmata
  • Empty monochromatic simplices in k-colored point sets
slide-7
SLIDE 7

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

2

Overview

  • Introduction
  • ”Roadmap”
  • Large sized triangulations
  • ”Technical” results
  • Pulling complexes
  • Generalized ”Order” and ”Discrepancy” lemmata
  • Empty monochromatic simplices in k-colored point sets
  • Conclusion
slide-8
SLIDE 8

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

3

Introduction

  • sets S ⊂ Rd of n points in general position
slide-9
SLIDE 9

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

3

Introduction

  • sets S ⊂ R2 of n points in general position
  • How many empty triangles exist?
slide-10
SLIDE 10

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

3

Introduction

  • sets S ⊂ R2 of n points in general position
  • How many empty triangles exist?
  • Katchalski and Meir, 1988:

n−1

2

  • ≤ # empty triangles ≤ cn2
  • Dehnhardt, 1987:

B´ ar´ any and F¨ uredi, 1987: n2−5n+10 ≤ # empty triangles ≤ 2n2

ar´ any and Valtr, 2004: # empty triangles ≤ 1.6195 . . . n2 + o(n2)

  • Aichholzer, Fabila-Monroy, H., Huemer, Pilz,

and Vogtenhuber, 2012: n2− 32n

7 + 22 7 ≤ # empty triangles

slide-11
SLIDE 11

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty triangles exist?
slide-12
SLIDE 12

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty monochromatic triangles exist?
slide-13
SLIDE 13

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty monochromatic triangles exist?
  • Devillers, Hurtado, K´

arolyi, and Seara, 2003: k = 2: ≥ n

4

  • −2 compatible empty monochr. triangles
slide-14
SLIDE 14

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty monochromatic triangles exist?
  • Devillers, Hurtado, K´

arolyi, and Seara, 2003: k = 2: ≥ n

4

  • −2 compatible empty monochr. triangles

k = 3: ∃ sets with no empty monochr. triangles

slide-15
SLIDE 15

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty monochromatic triangles exist?
  • Devillers, Hurtado, K´

arolyi, and Seara, 2003: k = 2: ≥ n

4

  • −2 compatible empty monochr. triangles

k = 3: ∃ sets with no empty monochr. triangles

  • Aichholzer, Fabila-Monroy, Flores-Pe˜

naloza, H., Huemer, and Urrutia, 2008: k = 2: Ω

  • n

5

/

4

  • empty monochromatic triangles
slide-16
SLIDE 16

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

4

Introduction

  • k-colored sets S ⊂ R2 of n points in general position
  • How many empty monochromatic triangles exist?
  • Devillers, Hurtado, K´

arolyi, and Seara, 2003: k = 2: ≥ n

4

  • −2 compatible empty monochr. triangles

k = 3: ∃ sets with no empty monochr. triangles

  • Aichholzer, Fabila-Monroy, Flores-Pe˜

naloza, H., Huemer, and Urrutia, 2008: k = 2: Ω

  • n

5

/

4

  • empty monochromatic triangles
  • Pach and T´
  • th, 2008:

k = 2: Ω

  • n

4

/

3

  • empty monochromatic triangles
slide-17
SLIDE 17

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
slide-18
SLIDE 18

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
  • How many empty d-simplices exist?
slide-19
SLIDE 19

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
  • Definition of a d-simplex:
slide-20
SLIDE 20

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
  • Definition of a d-simplex:
  • Convex hull of S′ ⊆S:
  • Conv(S′): intersection of all convex sets containing S′
  • CH(S′): boundary of Conv(S′)
slide-21
SLIDE 21

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
  • Definition of a d-simplex:
  • Convex hull of S′ ⊆S:
  • Conv(S′): intersection of all convex sets containing S′
  • CH(S′): boundary of Conv(S′)
  • 0≤m≤d: ”

m-simplex is Conv(X) (X ⊆S, |X| = m+1)”

  • vertices of an m-simplex: v ∈X
  • faces of an m-simplex: Conv(X′), X′ ⊂X
slide-22
SLIDE 22

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

5

Introduction

  • sets S ⊂ Rd of n points in general position
  • Definition of a d-simplex:
  • Convex hull of S′ ⊆S:
  • Conv(S′): intersection of all convex sets containing S′
  • CH(S′): boundary of Conv(S′)
  • 0≤m≤d: ”

m-simplex is Conv(X) (X ⊆S, |X| = m+1)”

  • vertices of an m-simplex: v ∈X
  • faces of an m-simplex: Conv(X′), X′ ⊂X
  • Simplicial complex K:
  • ”K is a set of interior disjoint empty d-simplices”
slide-23
SLIDE 23

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

6

Introduction

  • sets S ⊂ Rd of n points in general position
  • How many empty d-simplices exist?
slide-24
SLIDE 24

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

6

Introduction

  • sets S ⊂ Rd of n points in general position
  • How many empty d-simplices exist?
  • Katchalski and Meir, 1988:

at least n−1

d

  • = Ω(nd) empty d-simplices

ar´ any and F¨ uredi, 1987: at most cd n

d

  • = O(nd) expected empty d-simplices

in a random set

slide-25
SLIDE 25

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

7

Introduction

  • k-colored sets S ⊂ Rd of n points in general position
  • How many empty monochromatic d-simplices exist?
slide-26
SLIDE 26

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

7

Introduction

  • k-colored sets S ⊂ Rd of n points in general position
  • How many empty monochromatic d-simplices exist?
  • Urrutia, 2003:

d = 3, k = 4: there always exists an empty monochromatic d-simplex (tetrahedron)

slide-27
SLIDE 27

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

7

Introduction

  • k-colored sets S ⊂ Rd of n points in general position
  • How many empty monochromatic d-simplices exist?
  • Urrutia, 2003:

d = 3, k = 4: there always exists an empty monochromatic d-simplex (tetrahedron)

  • by proving that every S can be triangulated with more

than 3n tetrahedra

slide-28
SLIDE 28

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

7

Introduction

  • k-colored sets S ⊂ Rd of n points in general position
  • How many empty monochromatic d-simplices exist?
  • Urrutia, 2003:

d = 3, k = 4: there always exists an empty monochromatic d-simplex (tetrahedron)

  • by proving that every S can be triangulated with more

than 3n tetrahedra

  • Problem: triangulate S with many d-simplices
slide-29
SLIDE 29

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

7

Introduction

  • k-colored sets S ⊂ Rd of n points in general position
  • How many empty monochromatic d-simplices exist?
  • Urrutia, 2003:

d = 3, k = 4: there always exists an empty monochromatic d-simplex (tetrahedron)

  • by proving that every S can be triangulated with more

than 3n tetrahedra

  • Problem: triangulate S with many d-simplices

(”minmax”)

slide-30
SLIDE 30

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

8

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
slide-31
SLIDE 31

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

8

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
  • Edelsbrunner, Preparata, and West, 1990:

R3: upper bound of

7 15n2 + O(n) tetrahedra

  • Brass, 2005:

R3: ∃ sets of points where every triangulation has O(n

5

/

3) tetrahedra

slide-32
SLIDE 32

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

8

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
  • Edelsbrunner, Preparata, and West, 1990:

R3: upper bound of

7 15n2 + O(n) tetrahedra

  • Brass, 2005:

R3: ∃ sets of points where every triangulation has O(n

5

/

3) tetrahedra

Rd: ∃ sets of points where every triangulation has O(n

1 d + d−1 d

⌈ d

2 ⌉) d-simplices

slide-33
SLIDE 33

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

9

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
  • Rothschild and Straus, 1985:

Rd: all triangulations have at least (n−d) d-simplices

slide-34
SLIDE 34

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

9

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
  • Rothschild and Straus, 1985:

Rd: all triangulations have at least (n−d) d-simplices

  • Urrutia, 2003:

R3: ∃ triangulation with more than 3n tetrahedra

slide-35
SLIDE 35

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

9

Introduction

  • sets S ⊂ Rd of n points in general position
  • Large sized triangulations:
  • Rothschild and Straus, 1985:

Rd: all triangulations have at least (n−d) d-simplices

  • Urrutia, 2003:

R3: ∃ triangulation with more than 3n tetrahedra

  • we generalize / improve to:
  • ∃ triangulation with at least (dn+Ω(log n)) d-simplices
slide-36
SLIDE 36

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Literature Theorem 1

”Lower Bound Theorem” Theorem 2

slide-37
SLIDE 37

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d T1, T2

slide-38
SLIDE 38

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Pulling complexes Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d Lemmata 6 – 13 Lemma 9 Lemma 10 T1, T2 T1 T5

slide-39
SLIDE 39

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Pulling complexes Order lemma Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d Lemmata 6 – 13 Lemma 9 Lemma 10 T1, T2 T1 T5 Lemmata 14, 15 L3 T2 L6

slide-40
SLIDE 40

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Pulling complexes Discrepancy lemma Order lemma Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d Lemmata 6 – 13 Lemma 9 Lemma 10 T1, T2 T1 T5 Lemmata 14, 15 L3 Lemmata 16, 18 – 23 T1 T2 L6 L8, L11–13 Corollary 17

slide-41
SLIDE 41

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Empty Monochromatic Simplices in k-Colored Point Sets

Pulling complexes Discrepancy lemma Order lemma Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d Lemmata 6 – 13 Lemma 9 Lemma 10 T1, T2 T1 T5 Lemmata 14, 15 L3 Lemmata 16, 18 – 23 T1 T2 L6 L8, L11–13 Corollary 25: d>2, k =d+1

∃ linear number of EMS

Theorem 29: d≥k ≥3

Ω(nd−k+1+2−d) EMS

Theorems 24, 27 – 29 Corollaries 25, 26 Corollary 17 T5 L15 L19, 21–23 C17 , L16

slide-42
SLIDE 42

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

10

Roadmap

Empty Monochromatic Simplices in k-Colored Point Sets Empty Monochromatic Simplices in 2-Colored Point Sets

Pulling complexes Discrepancy lemma Order lemma Literature Theorem 1

”Lower Bound Theorem” Theorem 2

Large sized triangulations Theorem 5: d>2

∃T , |T |≥dn+max

  • h, log2(n)

2d

  • −cd

Lemmata 3, 4 cd =d3+d2+d Lemmata 6 – 13 Lemma 9 Lemma 10 T1, T2 T1 T5 Lemmata 14, 15 L3 Lemmata 16, 18 – 23 T1 T2 L6 L8, L11–13 Corollary 25: d>2, k =d+1

∃ linear number of EMS

Theorem 29: d≥k ≥3

Ω(nd−k+1+2−d) EMS

Theorems 24, 27 – 29 Corollaries 25, 26 Theorems 31 – 33 Observation 30 Corollary 17 Theorem 33: d≥2, k =2

Ω(nd−2

/ 3) EMS

T5 L15 L14 L19, 21–23 C17 , L16 L18, 20

slide-43
SLIDE 43

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position

slide-44
SLIDE 44

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position

d>2, n>d(d+1)

slide-45
SLIDE 45

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd

d>2, n>d(d+1), cd = d3+d2+d . . . constant

slide-46
SLIDE 46

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd

d>2, n>d(d+1), cd = d3+d2+d . . . constant

slide-47
SLIDE 47

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd Proof:

By Theorem 1: CH(S) has at least dn− d(d+1)

2

edges

d>2, n>d(d+1), cd = d3+d2+d . . . constant

slide-48
SLIDE 48

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd Proof:

By Theorem 1: CH(S) has at least dn− d(d+1)

2

edges ⇒ ∃ point p ∈ S with degree at least 2d in CH(S)

as long as n>d(d+1)

d>2, n>d(d+1), cd = d3+d2+d . . . constant

slide-49
SLIDE 49

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd Proof:

By Theorem 1: CH(S) has at least dn− d(d+1)

2

edges ⇒ ∃ point p ∈ S with degree at least 2d in CH(S)

as long as n>d(d+1)

  • successively remove such points p from S until d(d+1) points left
  • arbitrary triangulation Td(d+1) of size at least d(d+1)−d = d2
  • insert points p in reversed order: ≥2d−(d−1) = d+1 d-simplices each

d>2, n>d(d+1), cd = d3+d2+d . . . constant

slide-50
SLIDE 50

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

11

Large sized triangulations: convex set

Lemma 3: ∀ S ⊂Rd of n points in convex position ∃ triangulation of size at least (d+1)n − cd Proof:

By Theorem 1: CH(S) has at least dn− d(d+1)

2

edges ⇒ ∃ point p ∈ S with degree at least 2d in CH(S)

as long as n>d(d+1)

d>2, n>d(d+1), cd = d3+d2+d . . . constant

  • successively remove such points p from S until d(d+1) points left
  • arbitrary triangulation Td(d+1) of size at least d(d+1)−d = d2
  • insert points p in reversed order: ≥2d−(d−1) = d+1 d-simplices each

⇒ triangulation of size at least d2+(d+1)(n−d(d+1)) = (d+1)n − cd

slide-51
SLIDE 51

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

slide-52
SLIDE 52

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

slide-53
SLIDE 53

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least dn+max

  • h, log2(n)

2d

  • − cd
slide-54
SLIDE 54

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

dn+max

  • h, log2(n)

2d

  • − cd
slide-55
SLIDE 55

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h > log2(n)/(2d) > d(d + 1)

dn+max

  • h, log2(n)

2d

  • − cd
slide-56
SLIDE 56

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h > log2(n)/(2d) > d(d + 1)
  • ∃ triangulation of P of size at least (d+1)h − cd

dn+max

  • h, log2(n)

2d

  • − cd
slide-57
SLIDE 57

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h > log2(n)/(2d) > d(d + 1)
  • ∃ triangulation of P of size at least (d+1)h − cd
  • insert the remaining n−h points ⇒ d additional d-simplices each

dn+max

  • h, log2(n)

2d

  • − cd
slide-58
SLIDE 58

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h > log2(n)/(2d) > d(d + 1)
  • ∃ triangulation of P of size at least (d+1)h − cd
  • insert the remaining n−h points ⇒ d additional d-simplices each
  • ⇒ resulting triangulation has size at least

dn+h−cd > dn+ log2(n)

2d

− cd

dn+max

  • h, log2(n)

2d

  • − cd
slide-59
SLIDE 59

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h ≤ log2(n)/(2d)

dn+max

  • h, log2(n)

2d

  • − cd
slide-60
SLIDE 60

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h ≤ log2(n)/(2d)
  • Erd˝
  • s-Szekeres: ∃ convex set Q ⊂S, |Q|> log2(n)

2

>d(d+1)

dn+max

  • h, log2(n)

2d

  • − cd
slide-61
SLIDE 61

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h ≤ log2(n)/(2d)
  • Erd˝
  • s-Szekeres: ∃ convex set Q ⊂S, |Q|> log2(n)

2

>d(d+1)

  • P ′=P\Q: ∃ triangulation of P ′∪ Q of size at least (d+1)|Q|−cd+|P ′|

dn+max

  • h, log2(n)

2d

  • − cd
slide-62
SLIDE 62

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h ≤ log2(n)/(2d)
  • Erd˝
  • s-Szekeres: ∃ convex set Q ⊂S, |Q|> log2(n)

2

>d(d+1)

  • P ′=P\Q: ∃ triangulation of P ′∪ Q of size at least (d+1)|Q|+|P ′|−cd
  • insert the remaining points ⇒ d additional d-simplices each

dn+max

  • h, log2(n)

2d

  • − cd
slide-63
SLIDE 63

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

12

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least Proof:

Two cases:

  • |P| = h ≤ log2(n)/(2d)
  • Erd˝
  • s-Szekeres: ∃ convex set Q ⊂S, |Q|> log2(n)

2

>d(d+1)

  • P ′=P\Q: ∃ triangulation of P ′∪ Q of size at least (d+1)|Q|+|P ′|−cd
  • insert the remaining points ⇒ d additional d-simplices each
  • ⇒ resulting triangulation has size at least

(d+1)|Q|+|P ′|−cd + d(n−|Q|−|P ′|) > dn+ log2(n)

2d

− cd

dn+max

  • h, log2(n)

2d

  • − cd
slide-64
SLIDE 64

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

13

Note on Theorem 5

  • The constant cd in Lemma 3 can be improved to
  • d3

2 + 13d2 12 + 7d 12 . . . equals 25 for d = 3

  • For d = 3 Theorem 5 improves to
  • 3n + max
  • h, log2 n

6

  • − 25
slide-65
SLIDE 65

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

13

Note on Theorem 5

[EWP] H. Edelsbrunner, F.P. Preparata, and D.B. West. Tetrahedrizing point sets in three dimensions. 1990.

  • The constant cd in Lemma 3 can be improved to
  • d3

2 + 13d2 12 + 7d 12 . . . equals 25 for d = 3

  • For d = 3 Theorem 5 improves to
  • 3n + max
  • h, log2 n

6

  • − 25
  • [EPW]: Every set of n points in general position in R3,

with h convex hull points, has a tetrahedrization of size at least 3(n − h) + 4h − 25 for h ≥ 13.

slide-66
SLIDE 66

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

14

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least dn + max

  • h, log2(n)

2d

  • − cd
slide-67
SLIDE 67

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

14

Large sized triangulations

Theorem 5: ∀ S ⊂Rd of n points in general position

d>2, n>4d2(d

+ 1), h . . . number of convex hull points

∃ triangulation of size at least dn + log2(n)

2d

− cd

slide-68
SLIDE 68

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
slide-69
SLIDE 69

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set
slide-70
SLIDE 70

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set
slide-71
SLIDE 71

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set

p

slide-72
SLIDE 72

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set

p

}

  • n−1

2

  • n−1

2

  • }

Π

slide-73
SLIDE 73

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set

Π′ Π′′ p

}

  • n−1

2

  • n−1

2

  • }

Π

slide-74
SLIDE 74

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set

Π′ Π′′ p

slide-75
SLIDE 75

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

15

Pulling complexes

  • d-simplicial complex K of S ⊂Rd such that
  • for a predefined subset X ⊂S, (1 ≤ |X| ≤ d−1)
  • each d-simplex contains X in its vertex set

Π′ Π′′ p

slide-76
SLIDE 76

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

16

Pulling complexes

Lemma 9:

  • ∀ S ⊂Rd (d>3) of n>4d2(d+1) points in general position
  • ∀ point p∈S

⇒ ∃ d-dimensional simplicial complex of size at least (d−1)n+ log2 n

2(d−1) −2cd−1

all whose d-simplices have p as a vertex

slide-77
SLIDE 77

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

17

Pulling complexes

slide-78
SLIDE 78

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

17

Pulling complexes

Π X r := |X|

(r−1)-dimensional

slide-79
SLIDE 79

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

17

Pulling complexes

Π Π′ X r := |X|

(r−1)-dimensional (d−(r−1))-dimensional

slide-80
SLIDE 80

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

17

Pulling complexes

Π Π′ X r := |X|

(r−1)-dimensional (d−(r−1))-dimensional

slide-81
SLIDE 81

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

17

Pulling complexes

Π Π′ X r := |X|

(r−1)-dimensional (d−(r−1))-dimensional

slide-82
SLIDE 82

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

18

Pulling complexes

Lemma 10:

  • ∀ S ⊂Rd (d>3) of n>4d2(d+1) points in general position
  • ∀X ⊂S and 1 ≤ |X| ≤ d−3

⇒ ∃ d-dimensional simplicial complex of size at least (d−|X|)n+

log2 n 2(d−|X|) −2cd−1

all whose d-simplices contain X in their vertex set

slide-83
SLIDE 83

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

19

”Generalized Order Lemma”

  • ”Generalized Order Lemma” (Lemma 15)
  • S ⊂ Rd set of n ≥ d+1 points in general position
  • d > 2, h := |CH(S) ∩ S|

⇒ ∃ d-dimensional simplicial complex with at least (d−1)n + (n−h)(2(1−d))+2h−cd d-simplices, each having at least one of their vertices in CH(S) ∩ S

slide-84
SLIDE 84

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

19

”Generalized Order Lemma”

  • ”Generalized Order Lemma” (Lemma 15)
  • S ⊂ Rd set of n ≥ d+1 points in general position
  • d > 2, h := |CH(S) ∩ S|

⇒ ∃ d-dimensional simplicial complex with at least (d−1)n + (n−h)(2(1−d))+2h−cd d-simplices, each having at least one of their vertices in CH(S) ∩ S (n−h)(2(1−d)) ⇔

  • .

. .

  • (n−h)

}

(d−1) times

slide-85
SLIDE 85

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

20

Discrepancy

  • k-colored set S ⊂ Rd of n points in general position
  • k . . . constant, d ≥ 2
  • (S1, . . . , Sk) . . . color classes of S
  • Smax . . . biggest color class
slide-86
SLIDE 86

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

20

Discrepancy

  • k-colored set S ⊂ Rd of n points in general position
  • k . . . constant, d ≥ 2
  • (S1, . . . , Sk) . . . color classes of S
  • Smax . . . biggest color class
  • discrepancy δ(S):
slide-87
SLIDE 87

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

20

Discrepancy

  • k-colored set S ⊂ Rd of n points in general position
  • k . . . constant, d ≥ 2
  • (S1, . . . , Sk) . . . color classes of S
  • Smax . . . biggest color class
  • discrepancy δ(S):
  • bichromatic (k=2):

δ(S) := |Smax|−|S\Smax|

slide-88
SLIDE 88

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

20

Discrepancy

  • k-colored set S ⊂ Rd of n points in general position
  • k . . . constant, d ≥ 2
  • (S1, . . . , Sk) . . . color classes of S
  • Smax . . . biggest color class
  • discrepancy δ(S):
  • δ(S) := (|Smax|−|Si|)
slide-89
SLIDE 89

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

20

Discrepancy

  • k-colored set S ⊂ Rd of n points in general position
  • k . . . constant, d ≥ 2
  • (S1, . . . , Sk) . . . color classes of S
  • Smax . . . biggest color class
  • discrepancy δ(S):
  • δ(S) := (|Smax|−|Si|)

= (k−1)|Smax|−|S\Smax| = k|Smax|−n

slide-90
SLIDE 90

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1)

slide-91
SLIDE 91

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

slide-92
SLIDE 92

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points
slide-93
SLIDE 93

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ apply Lemma 10 to Smax and X

  • ∃ d-dimensional simplicial complex, KX(Smax)
  • |KX(Smax)| ≥ (d−|X|)|Smax|+ log2 |Smax|

2(d−|X|) −2cd−1

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

slide-94
SLIDE 94

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ |KX(Smax)| ≥ (d−|X|)|Smax|+ log2 |Smax|

2(d−|X|) −2cd−1

slide-95
SLIDE 95

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ |KX(Smax)| ≥ (d−|X|)|Smax|+ log2 |Smax|

2(d−|X|) −2cd−1

slide-96
SLIDE 96

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ |KX(Smax)| ≥ (k−1)|Smax|+ log2 |Smax|

2(k−1)

−2cd−1

slide-97
SLIDE 97

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ |KX(Smax)| ≥ (k−1)|Smax|+ log2 |Smax|

2(k−1)

−2cd−1 of which at least

  • (k−1)|Smax|+ log2 |Smax|

2(k−1)

−2cd−1−|S \ Smax| are empty

slide-98
SLIDE 98

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ |KX(Smax)| ≥ (k−1)|Smax|+ log2 |Smax|

2(k−1)

−2cd−1 of which at least

  • (k−1)|Smax|+ log2 |Smax|

2(k−1)

−2cd−1−|S \ Smax| are empty (k−1)|Smax|−|S \ Smax| = δ(S)

slide-99
SLIDE 99

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ KX(Smax) : ≥ δ(S)+ log2 |Smax|

2(k−1)

−2cd−1 empty monochr. d-simplices

slide-100
SLIDE 100

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ KX(Smax) : ≥ δ(S)+ log2 |Smax|

2(k−1)

−2cd−1 empty monochr. d-simplices

  • |Smax|

d−k+1

  • many subsets X
slide-101
SLIDE 101

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ KX(Smax) : ≥ δ(S)+ log2 |Smax|

2(k−1)

−2cd−1 empty monochr. d-simplices

  • |Smax|

d−k+1

  • many subsets X
  • over-count each d-simplex at most
  • d+1

d−k+1

  • times
slide-102
SLIDE 102

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

21

”Generalized Discrepancy Lemma”

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Proof:
  • choose a set X ⊂Smax of d−k+1 points

→ KX(Smax) : ≥ δ(S)+ log2 |Smax|

2(k−1)

−2cd−1 empty monochr. d-simplices

  • |Smax|

d−k+1

  • many subsets X
  • over-count each d-simplex at most
  • d+1

d−k+1

  • times → (|Smax|

d−k+1)

(

d+1 d−k+1)

slide-103
SLIDE 103

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

22

Simple observation

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

slide-104
SLIDE 104

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

22

Simple observation

  • ”Generalized Discrepancy Lemma” (Lemma 19)
  • k-colored set S ⊂ Rd of n points in general position
  • d ≥ k > 3,

n > k·4d2(d+1) ⇒ S determines Ω

  • nd−k+1 · (δ(S) + log n)
  • empty

monochromatic d-simplices

  • Corollary 26:

S determines Ω

  • nd−k+1 log n
  • empty monochromatic

d-simplices

slide-105
SLIDE 105

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

23

More discrepancy lemmata

empty monochr. d-simplices Lemma 19 d ≥ k > 3 Ω

  • nd−k+1 · (δ(S) + log n)
  • Lemma 18

d = 2, k = 2 Lemma 20 d ≥ 3, k = 2 Lemma 21 d = 3, k = 3 Ω(nd−k+1 · δ(S)) Lemma 22 d > 4, k = 3 Lemma 23 d = 4, k = 3

slide-106
SLIDE 106

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

slide-107
SLIDE 107

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • O. Devillers, F. Hurtado, G. K´

aroly, and C. Seara. Chromatic variants of the Erd˝

  • s-Szekeres theorem on

points in convex position. 2003. ∃ arbitrarily large 3-colored point sets in R2 which do not contain an empty monochromatic triangle

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

slide-108
SLIDE 108

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • Recall Theorem 5:
  • ∀ S ⊂Rd of n points in general position
  • d>2, n>4d2(d

+ 1)

  • ∃ triangulation of size at least dn+ log2(n)

2d

− cd

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

slide-109
SLIDE 109

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

  • Apply Theorem 5 to Smax:

|Smax| ≥

  • n

d+1

  • ∃ triangulation of size at least

d|Smax|+ log2(|Smax|)

2d

− cd

slide-110
SLIDE 110

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

  • Apply Theorem 5 to Smax:

|Smax| ≥

  • n

d+1

  • ∃ triangulation of size at least

d|Smax|+ log2(|Smax|)

2d

− cd

  • at most d|Smax| points of remaining colors
slide-111
SLIDE 111

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

24

(d+1)-Colored Point Sets

  • (d+1)-colored set S ⊂ Rd of n points in general position
  • d > 2,

n ≥ (d+1)·4d(cd+1)

cd = d3+d2+d

  • Apply Theorem 5 to Smax:

|Smax| ≥

  • n

d+1

  • ∃ triangulation of size at least

d|Smax|+ log2(|Smax|)

2d

− cd

  • at most d|Smax| points of remaining colors

⇒ at least

log2(|Smax|) 2d

− cd ≥ 2d(cd+1)

2d

− cd = 1 empty d-simplices

slide-112
SLIDE 112

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

25

(d+1)-Colored Point Sets

  • Theorem 24:
  • Every (d+1)-colored set S ⊂ Rd (d>2)
  • f n ≥ (d+1)·4d(cd+1) points in general position

determines an empty monochromatic d-simplex.

slide-113
SLIDE 113

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

25

(d+1)-Colored Point Sets

  • Theorem 24:
  • Every (d+1)-colored set S ⊂ Rd (d>2)
  • f n ≥ (d+1)·4d(cd+1) points in general position

determines an empty monochromatic d-simplex.

slide-114
SLIDE 114

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

25

(d+1)-Colored Point Sets

  • Theorem 24:
  • Every (d+1)-colored set S ⊂ Rd (d>2)
  • f n ≥ (d+1)·4d(cd+1) points in general position

determines an empty monochromatic d-simplex.

  • Corollary 25:
  • Every (d+1)-colored set of n points in general position

in Rd (d>2) determines at least a linear number of empty monochromatic d-simplices.

slide-115
SLIDE 115

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

26

d-Colored Point Sets

  • Theorem 27:
  • ∀ d-colored sets S ⊂Rd of n points in general position
  • d>2, n≥f(d)

f(d) constant w.r.t. n

slide-116
SLIDE 116

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

26

d-Colored Point Sets

  • Theorem 27:
  • ∀ d-colored sets S ⊂Rd of n points in general position
  • d>2, n≥f(d)

f(d) constant w.r.t. n

  • for each color 1≤j ≤d, either:
slide-117
SLIDE 117

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

26

d-Colored Point Sets

  • Theorem 27:
  • ∀ d-colored sets S ⊂Rd of n points in general position
  • d>2, n≥f(d)

f(d) constant w.r.t. n

  • for each color 1≤j ≤d, either:
  • ∃ Ω
  • n1+2−d

empty monochromatic d-simplices

  • f color j
slide-118
SLIDE 118

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

26

d-Colored Point Sets

  • Theorem 27:
  • ∀ d-colored sets S ⊂Rd of n points in general position
  • d>2, n≥f(d)

f(d) constant w.r.t. n

  • for each color 1≤j ≤d, either:
  • ∃ Ω
  • n1+2−d

empty monochromatic d-simplices

  • f color j
  • r:
  • ∃ convex set C ⊂Rd, such that

|S∩C| = Θ(n) and δ(S∩C) = Ω(n2−d)

slide-119
SLIDE 119

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

27

d-Colored → k-Colored

  • Theorem 28:
  • ∀ k-colored sets S ⊂Rd of n points in general position
  • d≥k>2, n≥f(d, k)

f(d, k) constant w.r.t. n

  • for each color 1≤j ≤k, either:
  • ∃ Ω
  • nd−k+1+2−d

empty monochromatic d-simplices

  • f color j
  • r:
  • ∃ convex set C ⊂Rd, such that

|S∩C| = Θ(n) and δ(S∩C) = Ω(n2−d)

slide-120
SLIDE 120

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

28

Improvement

  • Theorem 29:
  • Every k-colored set in n points in general position in

Rd (d≥k≥3) determines Ω

  • nd−k+1+2−d

empty monochromatic d-simplices.

slide-121
SLIDE 121

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

28

Improvement

  • Theorem 29:
  • Every k-colored set in n points in general position in

Rd (d≥k≥3) determines Ω

  • nd−k+1+2−d

empty monochromatic d-simplices.

  • Proof:
slide-122
SLIDE 122

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

28

Improvement

  • Theorem 29:
  • Every k-colored set in n points in general position in

Rd (d≥k≥3) determines Ω

  • nd−k+1+2−d

empty monochromatic d-simplices.

  • Proof:
  • either: Ω
  • nd−k+1+2−d

EMS directly by Theorem 28

slide-123
SLIDE 123

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

28

Improvement

  • Theorem 29:
  • Every k-colored set in n points in general position in

Rd (d≥k≥3) determines Ω

  • nd−k+1+2−d

empty monochromatic d-simplices.

  • Proof:
  • either: Ω
  • nd−k+1+2−d

EMS directly by Theorem 28

  • or: by Theorem 28:
  • ∃ convex set C ⊂Rd: |S∩C|=Θ(n), δ(S∩C)=Ω(n2−d)
slide-124
SLIDE 124

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

28

Improvement

  • Theorem 29:
  • Every k-colored set in n points in general position in

Rd (d≥k≥3) determines Ω

  • nd−k+1+2−d

empty monochromatic d-simplices.

  • Proof:
  • either: Ω
  • nd−k+1+2−d

EMS directly by Theorem 28

  • or: by Theorem 28:
  • ∃ convex set C ⊂Rd: |S∩C|=Θ(n), δ(S∩C)=Ω(n2−d)
  • and by discrepancy lemmata:

  • nd−k+1 · δ(S)
  • EMS
slide-125
SLIDE 125

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

29

2-Colored Point Sets

  • similar to the case d ≥ k > 2
  • Theorem 33:
  • Every 2-colored set of n points in general position

in Rd (d≥2) determines Ω

  • nd−2

/

3

  • empty monochromatic d-simplices.
slide-126
SLIDE 126

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
slide-127
SLIDE 127

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
slide-128
SLIDE 128

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
  • # empty monochromatic d-simplicies if S is k-colored
slide-129
SLIDE 129

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
  • # empty monochromatic d-simplicies if S is k-colored

colors d ≥ 3 k = 2 Ω(nd−2/3)

slide-130
SLIDE 130

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
  • # empty monochromatic d-simplicies if S is k-colored

colors d ≥ 3 k = 2 Ω(nd−2/3) 3 ≤ k ≤ d Ω(nd−k+1+2−d)

slide-131
SLIDE 131

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
  • # empty monochromatic d-simplicies if S is k-colored

colors d ≥ 3 k = 2 Ω(nd−2/3) 3 ≤ k ≤ d Ω(nd−k+1+2−d) k = d + 1 Ω(n)

slide-132
SLIDE 132

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

30

Conclusion

  • ∀ sets S ⊂Rd of n points, h = |CH(S)∩S|
  • ∃ triangulation of size at least

dn+max

  • h, log2(n)

2d

  • − cd
  • # empty monochromatic d-simplicies if S is k-colored

colors d ≥ 3 k = 2 Ω(nd−2/3) 3 ≤ k ≤ d Ω(nd−k+1+2−d) k = d + 1 Ω(n) k ≥ d + 2 unknown

slide-133
SLIDE 133

P 23629–N18

Graz University of Technology Institute for Software Technology Thomas Hackl: Eurogiga Midterm Conference, July 9th – 13th, 2012

31

Thank you for your attention !