SLIDE 1 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
joint work with G¨ unter Rote Rafel Jaume
SLIDE 2
Basic definition and tools Recursive regularity and finest regular coarsening Applications Properties of recursively regular subdivisions Overview
SLIDE 3 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
SLIDE 4
Let A = {p1, ...pn} be a set of points in Rd
SLIDE 5
Let A = {p1, ...pn} be a set of points in Rd A polyhedral subdivision S of A is
SLIDE 6
Let A = {p1, ...pn} be a set of points in Rd A polyhedral subdivision S of A is a polytopal complex
SLIDE 7
Let A = {p1, ...pn} be a set of points in Rd A polyhedral subdivision S of A is a polytopal complex covering conv(A)
SLIDE 8
Let A = {p1, ...pn} be a set of points in Rd A polyhedral subdivision S of A is a polytopal complex covering conv(A) whose set of vertices is contained in A
SLIDE 9
Let A = {p1, ...pn} be a set of points in Rd A polyhedral subdivision S of A is Full-dimensional polytopes of the complex are called cells a polytopal complex covering conv(A) whose set of vertices is contained in A
SLIDE 10 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
SLIDE 11
S
SLIDE 12
A coarsening of a subdivision S is a polyhedral subdivision S′ such that every face f ∈ S is contained in some face κ(f) ∈ S′ S
SLIDE 13 A coarsening of a subdivision S is a polyhedral subdivision S′ such that every face f ∈ S is contained in some face κ(f) ∈ S′ S
′
SLIDE 14 A coarsening of a subdivision S is a polyhedral subdivision S′ such that every face f ∈ S is contained in some face κ(f) ∈ S′ S
′
Reciprocally, we say that S is a refinement of S′
- r that S is finer than S′
SLIDE 15 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
SLIDE 16
S is regular if it is the orthogonal projection of a (d + 1)-dimensional polytope P Rd
SLIDE 17
Rd+1
SLIDE 18
Rd+1 pi ωi pi
ωi
SLIDE 19 Rd+1 P = conv( pi
ωi
SLIDE 20
Rd+1
SLIDE 21
A non-regular subdivision
SLIDE 22
A non-regular subdivision
SLIDE 23
A regular coarsening
SLIDE 24 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
SLIDE 25
Whether S is regular can be checked by a LP
SLIDE 26
wall C D
SLIDE 27
pb1 pb2 pb3 q affine basis of C point in D \ C C D
SLIDE 28 pb1 pb2 pb3 q C D pb1
ωb1
ωb3
ωb2
ωq
SLIDE 29
pb1 pb2 pb3 q C D
SLIDE 30
... 1 pb1 ... pbd+1
... 1 1 pb1 ... pbd+1 q wb1 ... wbd+1 wq
pb1 pb2 pb3 q C D Local folding condition
SLIDE 31
pc1 pc2 pcd+1 C
SLIDE 32
pc1 pc2 pcd+1 q1 C
SLIDE 33
pc1 pc2 pcd+1 q1 C
SLIDE 34
pc1 pc2 pcd+1 C q2
SLIDE 35 pc1 pc2 pcd+1 Coplanarity conditions
... 1 1 pc1 ... pcd+1 qi wc1 ... wcd+1 wi
C q2
SLIDE 36
... 1 pb1 ... pbd+1
... 1 1 pb1 ... pbd+1 q wb1 ... wbd+1 wq
Coplanarity conditions Local folding conditions
... 1 1 pc1 ... pcd+1 qi wc1 ... wcd+1 wi
SLIDE 37
... 1 pb1 ... pbd+1
... 1 1 pb1 ... pbd+1 q wb1 ... wbd+1 wq
Coplanarity conditions Local folding conditions s1 · ω > 0 sm · ω > 0 e1 · ω = 0 el · ω = 0 . . . . . . ω = ω1 . . . ωn
S
... 1 1 pc1 ... pcd+1 qi wc1 ... wcd+1 wi
SLIDE 38
... 1 pb1 ... pbd+1
... 1 1 pb1 ... pbd+1 q wb1 ... wbd+1 wq
Coplanarity conditions Local folding conditions s1 · ω > 0 sm · ω > 0 e1 · ω = 0 el · ω = 0 . . . . . . ω = ω1 . . . ωn Lemma (De Loera, Rambau, Santos) S is regular ⇐ ⇒ S is compatible
S
... 1 1 pc1 ... pcd+1 qi wc1 ... wcd+1 wi
SLIDE 39
The finest regular coarsening is well defined Theorem
SLIDE 40
C D
Local folding condition >
Proof The finest regular coarsening is well defined Theorem
SLIDE 41
C D
Local folding condition Coplanarity condition > =
Proof The finest regular coarsening is well defined Theorem
SLIDE 42
Proof The finest regular coarsening is well defined Theorem s1 · ω > 0 sm · ω > 0 e1 · ω = 0 el · ω = 0
S
. . . . . .
SLIDE 43
Proof The finest regular coarsening is well defined Theorem s1 · ω > 0 sm · ω > 0 e1 · ω = 0 el · ω = 0
S
s1 · ω ≥ 0 sm · ω ≥ 0 e1 · ω = 0 el · ω = 0
S′
. . . . . . . . . . . .
SLIDE 44
Proof The finest regular coarsening is well defined Theorem s1 · ω > 0 sm · ω > 0 e1 · ω = 0 el · ω = 0
S
s1 · ω ≥ 0 sm · ω ≥ 0 e1 · ω = 0 el · ω = 0
S′
. . . . . . . . . . . . Is a closed cone. A point in its relatives interior satisfies in an strict way the maximum number of inequalities.
SLIDE 45 Recursive regularity and the finest regular coarsening
- f a polyhedral subdivision
SLIDE 46
A polyhedral subdivision S is recursively regular if: S is regular or There exists a regular coarsening S′ of S with coarsening function κ such that κ−1(C) is recursively regular, for each cell C ∈ S′
SLIDE 47
A polyhedral subdivision S is recursively regular if: S is regular or There exists a regular coarsening S′ of S with coarsening function κ such that κ−1(C) is recursively regular, for each cell C ∈ S′
SLIDE 48
A polyhedral subdivision S is recursively regular if: S is regular or There exists a regular coarsening S′ of S with coarsening function κ such that κ−1(C) is recursively regular, for each cell C ∈ S′
SLIDE 49
A polyhedral subdivision S is recursively regular if: S is regular or There exists a regular coarsening S′ of S with coarsening function κ such that κ−1(C) is recursively regular, for each cell C ∈ S′ Has it applications? How different from regularity is? Is flip-connected? How can we identify a recursively regular subdivision ?
SLIDE 50
Regularity tree
SLIDE 51
Regularity tree
SLIDE 52
Regularity tree
SLIDE 53
Regularity tree FRC
SLIDE 54
Regularity tree FRC
SLIDE 55
Regularity tree FRC
SLIDE 56
Regularity tree FRC FRC
SLIDE 57
Regularity tree FRC FRC
SLIDE 58
Regularity tree FRC FRC
SLIDE 59 S is recursively regular
- its regularity tree has regular leaves
Proposition
SLIDE 60 S is recursively regular
- its regularity tree has regular leaves
Proposition Corollary It can be efficiently decided if a subdivision is recursively regular
SLIDE 61 S is recursively regular
- its regularity tree has regular leaves
Proposition Proof Regularity tree with regular leaves ⇒ recursively regular (def)
SLIDE 62 S is recursively regular
- its regularity tree has regular leaves
Proposition Proof Regularity tree with regular leaves ⇒ recursively regular (def) Recursively regular ⇒ regular leaves:
SLIDE 63 S is recursively regular
- its regularity tree has regular leaves
Proposition Proof Regularity tree with regular leaves ⇒ recursively regular (def) Recursively regular ⇒ regular leaves: There exists a “regularity tree” certifying that S is recursively
- regular. The first coarsening must be coarser than the FRC.
Then, each cell of the FRC is recursively regular, since regularity is preserved by intersection with convex sets.
SLIDE 64
Regular Cyclic Regular subdivisions are acyclic Acyclicity theorem (Edelsbrunner) Subdivisions
1 2 3 4
Cyclic
SLIDE 65 Regular Cyclic
Regular subdivisions are acyclic Acyclicity theorem (Edelsbrunner) Subdivisions Recursively regular subdivisions are a superset of the regular subdivisions and a subset of the acyclic ones Proposition
1 2 3 4
Cyclic
SLIDE 66
Recursively regular subdivisions are a superset of the regular subdivisions and a subset of the acyclic ones Proposition Proof Assume that it exists a cycle in a recursively regular subdivision.
SLIDE 67
Recursively regular subdivisions are a superset of the regular subdivisions and a subset of the acyclic ones Proposition Proof We can assume that the cycle belongs to more than one cell of the FRC and apply recursion otherwise. Assume that it exists a cycle in a recursively regular subdivision.
SLIDE 68
Recursively regular subdivisions are a superset of the regular subdivisions and a subset of the acyclic ones Proposition Proof We can assume that the cycle belongs to more than one cell of the FRC and apply recursion otherwise. Then, the FRC cannot be regular, since a cycle cannot be destroyed by means of merging cells. Assume that it exists a cycle in a recursively regular subdivision.
SLIDE 69
There is a 5-dimensional point set with at least 12 recursively regular triangulations pairwise disconnected and disconnected from any regular triangulation in the graph of flips Proposition
SLIDE 70
Applications Floodlights Homothecies Graph embeddings Spider webs
SLIDE 71
Floodlights polyhedral fan with n cells
SLIDE 72
Floodlights polyhedral fan with n cells n points
SLIDE 73 Floodlights polyhedral fan with n cells n points
SLIDE 74 Floodlights polyhedral fan with n cells n points
covering assignment
SLIDE 75
Covering fans A polyhedral fan is universally covering if for any point set there exists a covering assignment
SLIDE 76
Covering fans A polyhedral fan is universally covering if for any point set there exists a covering assignment Regular fans are universally covering Theorem (Galperin & Galperin, 1981)
SLIDE 77
Covering fans A polyhedral fan is universally covering if for any point set there exists a covering assignment Regular fans are universally covering Theorem (Galperin & Galperin, 1981) Recursively regular fans are universally covering Cyclic fans are not universally covering Theorem
SLIDE 78 Regular Cyclic
Fans
SLIDE 79 Regular Cyclic
Universally covering Fans
SLIDE 80 Regular Cyclic
Universally covering Fans
SLIDE 81 Regular Cyclic
Universally covering Fans
SLIDE 82 Regular Cyclic
Universally covering Fans
SLIDE 83
Overlapping condition
SLIDE 84
Overlapping condition
SLIDE 85
Overlapping condition
SLIDE 86
V = {1, ..., n} Directional graph embeddings − → E ⊂ V × V
SLIDE 87
V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Directional graph embeddings − → E ⊂ V × V
SLIDE 88
N : − → E − → Rd V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Directional graph embeddings − → E ⊂ V × V
SLIDE 89
N : − → E − → Rd V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Is there a σ : V ↔ P N(i, j) · pi ≥ N(i, j) · pj ∀(i, j) ∈ − → E ? proper embedding Directional graph embeddings − → E ⊂ V × V
SLIDE 90
N : − → E − → Rd V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Is there a σ : V ↔ P N(i, j) · pi ≥ N(i, j) · pj ∀(i, j) ∈ − → E ? A path is always embeddable proper embedding Directional graph embeddings − → E ⊂ V × V
SLIDE 91
N : − → E − → Rd V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Is there a σ : V ↔ P N(i, j) · pi ≥ N(i, j) · pj ∀(i, j) ∈ − → E ? A path is always embeddable A cycle is not always embeddable proper embedding Directional graph embeddings − → E ⊂ V × V
SLIDE 92
N : − → E − → Rd V = {1, ..., n} P = {p1, ...pn} ⊂ Rd Is there a σ : V ↔ P N(i, j) · pi ≥ N(i, j) · pj ∀(i, j) ∈ − → E ? A path is always embeddable A cycle is not always embeddable A tree? proper embedding Directional graph embeddings − → E ⊂ V × V
SLIDE 93
Proposition If the graph is the normal graph (or a projection of a subgraph) of a fan and there is a covering assignment forP, there is a proper graph embedding for P. If there is no assignment satisfying OC for... , there is no proper embedding
SLIDE 94
Proposition If the graph is the normal graph (or a projection of a subgraph) of a fan and there is a covering assignment forP, there is a proper graph embedding for P. If there is no assignment satisfying OC for... , there is no proper embedding Corollary The graph of a polytope is always embeddable.
SLIDE 95
Homothecies 1,3 1,5 0,9 1,1 0,6 1,2
SLIDE 96
Homothecies 1,3 1,5 0,9 1,1 0,6 1,2
SLIDE 97
Homothecies
SLIDE 98
Homothecies
SLIDE 99
Homothecies
SLIDE 100
Homothecies
SLIDE 101
Homothecies
SLIDE 102
Homothecies Proposition If there is a covering assignment for the fan and the points corresponding to the transformations, there is a transformation assignment If there is no assignment satisfying OC for a fan and a pointset, there is a set of transformations for which there is no transformation assignment
SLIDE 103
Spiderwebs
SLIDE 104 Spiderwebs (Positive) equilibrium stress ↔ (convex) lifting (for non-crossing planar frameworks) Maxwell-Cremona correspondence Theorem (Connelly) If a tensegrity framework is infinitesimally rigid, then it has an equilibrium stress that is nonzero
Theorem (Crapo, Whiteley) If a spider web has a positive equilibriuem stress, then it is rigid
SLIDE 105
Spiderwebs
SLIDE 106
Spiderwebs
SLIDE 107
Spiderwebs
SLIDE 108
Spiderwebs
SLIDE 109
Spiderwebs
SLIDE 110
Spiderwebs A recursively regular spider web is rigid. A recursively regular non-regular spider web is infinitesimally flexible. Theorem
SLIDE 111
Conclusions The finest regular coarsening is well defined
SLIDE 112
Conclusions The finest regular coarsening is well defined Recursively regular subdivisions are Superset of regular subdivisions, subset of acyclic Easy to identify Not connected by flips
SLIDE 113
Conclusions The finest regular coarsening is well defined Recursively regular subdivisions are Superset of regular subdivisions, subset of acyclic Easy to identify Not connected by flips They are related to the problems Covering by floodlights Covering by homothecies Directional graph embedding Spider webs
SLIDE 114
Conclusions The finest regular coarsening is well defined Recursively regular subdivisions are Superset of regular subdivisions, subset of acyclic Easy to identify Not connected by flips They are related to the problems Covering by floodlights Covering by homothecies Directional graph embedding Universality, counterexamples Spider webs
SLIDE 115
Conclusions The finest regular coarsening is well defined Recursively regular subdivisions are Superset of regular subdivisions, subset of acyclic Easy to identify Not connected by flips They are related to the problems Covering by floodlights Covering by homothecies Directional graph embedding Universality, counterexamples Spider webs Open: Decision problem, universality limits
SLIDE 116
Conclusions The finest regular coarsening is well defined Recursively regular subdivisions are Superset of regular subdivisions, subset of acyclic Easy to identify Not connected by flips They are related to the problems Covering by floodlights Covering by homothecies Directional graph embedding Universality, counterexamples Thank you! Spider webs Open: Decision problem, universality limits