Radoslav Fulek,IST Austria C-planarity & Approximating Maps - - PowerPoint PPT Presentation

radoslav fulek ist austria c planarity approximating maps
SMART_READER_LITE
LIVE PREVIEW

Radoslav Fulek,IST Austria C-planarity & Approximating Maps - - PowerPoint PPT Presentation

C-planarity of Embedded Cyclic c-Graphs Radoslav Fulek,IST Austria C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability


slide-1
SLIDE 1

C-planarity of Embedded Cyclic c-Graphs

Radoslav Fulek,IST Austria

slide-2
SLIDE 2

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995))

slide-3
SLIDE 3

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem.

slide-4
SLIDE 4

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

slide-5
SLIDE 5

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

slide-6
SLIDE 6

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

slide-7
SLIDE 7

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

slide-8
SLIDE 8

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

slide-9
SLIDE 9

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

slide-10
SLIDE 10

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

C-planarity with pipes (Cortese et al. (2005))

slide-11
SLIDE 11

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

V1 V2 V3 V4 V5

C-planarity with pipes (Cortese et al. (2005)) C-planarity with pipes is tractable for cycles.

slide-12
SLIDE 12

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

ϕ :

C-planarity with pipes

slide-13
SLIDE 13

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

ϕ :

C-planarity with pipes

slide-14
SLIDE 14

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

ϕ :

H

/ Approximating Maps by Emb. C-planarity with pipes

slide-15
SLIDE 15

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E).

ϕ :

H

/ Approximating Maps by Emb. Does there exist an embedding of G that is an ǫ-approximation

  • f ϕ : G → H for any ǫ > 0?

C-planarity with pipes

slide-16
SLIDE 16

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E). Approximating Maps by Embeddings Does there exist an embedding of G that is an ǫ-approximation

  • f ϕ : G → H for any ǫ > 0?

We consider only continuous maps ϕ that map a vertex to a vertex and an edge either to an edge or a vertex.

slide-17
SLIDE 17

C-planarity & Approximating Maps C-planarity (Feng, Cohen and Eades (1995)) Its tractability is a long standing open problem. In a fairly general special case, of flat clustered graphs, the input is a graph G = (V1 ⊎ . . . ⊎ Vk, E). Approximating Maps by Embeddings Does there exist an embedding of G that is an ǫ-approximation

  • f ϕ : G → H for any ǫ > 0?

We consider only continuous maps ϕ that map a vertex to a vertex and an edge either to an edge or a vertex. ϕ is approximable by an embedding if there exists an ǫ-approximation that is an embedding for every ǫ > 0.

slide-18
SLIDE 18

Warm up

slide-19
SLIDE 19

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge.

slide-20
SLIDE 20

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

slide-21
SLIDE 21

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? The algorithm just ouputs Yes, there exists.

slide-22
SLIDE 22

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? The algorithm just ouputs Yes, there exists.

G

ϕ :

H

slide-23
SLIDE 23

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? The algorithm just ouputs Yes, there exists.

G G′ ⊃ G

slide-24
SLIDE 24

Warm up Let G = (V1 ⊎ V2, E) be cycle, mapped by ϕ to an edge. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? The algorithm just ouputs Yes, there exists.

G G′ ⊃ G

slide-25
SLIDE 25

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

slide-26
SLIDE 26

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-27
SLIDE 27

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G K3,3

slide-28
SLIDE 28

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-29
SLIDE 29

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-30
SLIDE 30

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-31
SLIDE 31

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-32
SLIDE 32

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G?

G

slide-33
SLIDE 33

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 2 2323131313213231313131321 3

slide-34
SLIDE 34

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 2 2323131313213231313131321 3

slide-35
SLIDE 35

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 2323131313213231313131321 2 1 3 p

slide-36
SLIDE 36

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1

slide-37
SLIDE 37

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 1 1 1 1 1 1

G

slide-38
SLIDE 38

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 1 1 1 1 1 1

G

Algorithm: (Cortese et al. 2005) Sum labels and devide by three, wn =

  • e l(e)

3

. The instance is positive iff |wn| ≤ 1.

slide-39
SLIDE 39

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 Algorithm: (Cortese et al. 2005) Sum labels and devide by three, wn =

  • e l(e)

3

. The instance is positive iff |wn| ≤ 1. −1 1 1 −1 1 −1

G

3 = 0

slide-40
SLIDE 40

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 Algorithm: (Cortese et al. 2005) Sum labels and devide by three, wn =

  • e l(e)

3

. The instance is positive iff |wn| ≤ 1.

p G

slide-41
SLIDE 41

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 Algorithm: (Cortese et al. 2005) Sum labels and devide by three, wn =

  • e l(e)

3

. The instance is positive iff |wn| ≤ 1. 1 1 1

3 3 = 1

G

slide-42
SLIDE 42

Warm up Let G = (V1 ⊎ V2 ⊎ V3, E) be mapped to a cycle of length three. Can we decide in a polynomial time if ϕ is approximable by an embedding of G? 1 1 1 −1 −1 −1 Algorithm: (Cortese et al. 2005) Sum labels and devide by three, wn =

  • e l(e)

3

. The instance is positive iff |wn| ≤ 1.

G p

slide-43
SLIDE 43

C-planarity of Emedded Cyclic c-Graphs

slide-44
SLIDE 44

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2, E) be a planar graph mapped by ϕ to an edge.

slide-45
SLIDE 45

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2, E) be a planar graph mapped by ϕ to an edge. Theorem 3. Biedl (1998), Gutwenger et al. (2002) We can decide in polynomial time if ϕ is approximable by an embedding.

slide-46
SLIDE 46

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system.

slide-47
SLIDE 47

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. Theorem 2. (M.Skopenkov (2003)) The question of whether ϕ is approximable by an embedding with a given rotation system can be reduced to solving a system of linear equations over Z/2Z with O(n2) variables and equations.

slide-48
SLIDE 48

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. Theorem 2. (M.Skopenkov (2003)) The question of whether ϕ is approximable by an embedding with a given rotation system can be reduced to solving a system of linear equations over Z/2Z with O(n2) variables and equations. The technique of Skopenkov extends ideas of Minc (1997) considereding the case when G is a path.

slide-49
SLIDE 49

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. Theorem 2. (M.Skopenkov (2003)) The question of whether ϕ is approximable by an embedding with a given rotation system can be reduced to solving a system of linear equations over Z/2Z with O(n2) variables and equations. The technique of Skopenkov extends ideas of Minc (1997) considereding the case when G is a path. Sieklucki (1969) characterized all the graphs G for which any map ϕ : G → R is approximable in R2.

slide-50
SLIDE 50

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system.

3 3 = 1

1 1 1 −1 −1 −1

G

slide-51
SLIDE 51

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. 1 1 1 −1 −1 −1 p

G

slide-52
SLIDE 52

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. 1 1 1 −1 −1 −1 p

G

p p p

slide-53
SLIDE 53

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. 1 1 1 −1 −1 −1

G

slide-54
SLIDE 54

C-planarity of Emedded Cyclic c-Graphs Let G = (V1 ⊎ V2 ⊎ . . . ⊎ Vk, E) be a planar graph mapped by ϕ to a k-cycle. Theorem 1. (F.) We can decide in quadratic time if ϕ is approximable by an embedding with a given rotation system. Algorithm: Check the necessary winding number condition. Hence, after some preprocessing we assume that (i) winding number condition holds; (ii)Vi’s form independent sets; (iii) G is connected; and (iv) faces are “simple”. Solve a perfect matching problem.

slide-55
SLIDE 55

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

slide-56
SLIDE 56

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p f f p

+1,+1,+1,-1,-1,-1 +1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

f

+1,+1,+1

p

slide-57
SLIDE 57

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p f f p

+1,+1,+1,-1,-1,-1 +1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

f

+1,+1,+1

p

We assume that |wn(f)| > 0 for exactly one inner face of G.

slide-58
SLIDE 58

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p f f p

+1,+1,+1,-1,-1,-1 +1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

f

+1,+1,+1

p

We assume that |wn(f)| > 0 for exactly one inner face of G. Otherwise, the problem was solved by Angelini et al. (2013).

slide-59
SLIDE 59

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p

+1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

v u f

slide-60
SLIDE 60

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p

+1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

v u f u p f v f

slide-61
SLIDE 61

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p

+1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

v u f u p f v f

slide-62
SLIDE 62

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p

+1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

v u f u p f v f

slide-63
SLIDE 63

C-planarity of Emedded Cyclic c-Graphs Simple faces: (i) Labels change sign at most four times in the facial walk; and (ii) If there are four changes, the runs of +1’s and -1’s are of the same length.

p f p

+1,-1,+1,+1,+1 +1,+1,+1,-1,-1,-1, +1,+1,+1,-1,-1,-1

v u f u p f v f p f

slide-64
SLIDE 64

C-planarity of Emedded Cyclic c-Graphs The assignment of “concave angles” corresponds to a perfect matching in a face-vertex incidence graph I = (F ∪ V0, E0).

slide-65
SLIDE 65

C-planarity of Emedded Cyclic c-Graphs The assignment of “concave angles” corresponds to a perfect matching in a face-vertex incidence graph I = (F ∪ V0, E0).

p f p f f p f p

Augmenting G based of the perfect matching.

slide-66
SLIDE 66

C-planarity of Emedded Cyclic c-Graphs The assignment of “concave angles” corresponds to a perfect matching in a face-vertex incidence graph I = (F ∪ V0, E0).

p f p f f p f p

Augmenting G based of the perfect matching. The crux is to prove that we do not obtain a directed cycle.

slide-67
SLIDE 67

C-planarity of Emedded Cyclic c-Graphs The assignment of “concave angles” corresponds to a perfect matching in a face-vertex incidence graph I = (F ∪ V0, E0).

p f p f f p f p

Augmenting G based of the perfect matching. The crux is to prove that we do not obtain a directed cycle. Main observation: An area-minimal directed cycle − → C contains a cycle C′ such that |wn(C′)| = 1 in the cl. of its interior.

slide-68
SLIDE 68

C-planarity of Emedded Cyclic c-Graphs The assignment of “concave angles” corresponds to a perfect matching in a face-vertex incidence graph I = (F ∪ V0, E0).

p f p f f p f p

Augmenting G based of the perfect matching. The crux is to prove that we do not obtain a directed cycle. Main observation: An area-minimal directed cycle − → C contains a cycle C′ such that |wn(C′)| = 1 in the cl. of its interior. 0 = wn(− → C ) =

f⊂int(− → C ) wn(f) = wn(C′) + 0 = 0

slide-69
SLIDE 69

C-planarity of Emedded Cyclic c-Graphs Assuming that we can decide if a map is approximable in a polynomial time. Can we get FPT of c-planarity w.r.t. to the number of faces of the input graph?

slide-70
SLIDE 70

C-planarity of Emedded Cyclic c-Graphs Assuming that we can decide if a map is approximable in a polynomial time. Can we get FPT of c-planarity w.r.t. to the number of faces of the input graph? We conjecture that the result of M.Skopenkov extends to any map ϕ without a need to fix the rotation system.

slide-71
SLIDE 71

C-planarity of Emedded Cyclic c-Graphs Assuming that we can decide if a map is approximable in a polynomial time. Can we get FPT of c-planarity w.r.t. to the number of faces of the input graph? We conjecture that the result of M.Skopenkov extends to any map ϕ without a need to fix the rotation system. That would imply FTP w.r.t. the number of faces of both the input graph and the target graph.