Some Results and Open Problems Giuseppe (Beppe) Liotta (University - - PowerPoint PPT Presentation

some results and open problems
SMART_READER_LITE
LIVE PREVIEW

Some Results and Open Problems Giuseppe (Beppe) Liotta (University - - PowerPoint PPT Presentation

Graph Drawing Beyond Planarity: Some Results and Open Problems Giuseppe (Beppe) Liotta (University of Perugia) Outline Graph Drawing (GD) beyond planarity Combinatorial relationships Optimization trade-offs and algorithms Open


slide-1
SLIDE 1

Graph Drawing Beyond Planarity: Some Results and Open Problems

Giuseppe (Beppe) Liotta (University of Perugia)

slide-2
SLIDE 2

Outline

  • Graph Drawing (GD) beyond planarity
  • Combinatorial relationships
  • Optimization trade-offs and algorithms
  • Open problems
slide-3
SLIDE 3

GD beyond planarity

slide-4
SLIDE 4

Relational Data Sets

slide-5
SLIDE 5

Graph Drawing

in

  • ut

5 1 6 2 3 4

drawing

the drawing must be readable

G = (V, E) V = {1, 2, 3, 4, 5, 6} E = { (1,3) (1,6) (2,3) (2,5) (2,4) (2,6) (3,5) (4,5) (4,6) }

Graph Drawing System

slide-6
SLIDE 6

Readability and Crossings

edge crossings significantly affect the readability (see, e.g., Sugiyama et al., Warshall, North et al., Batini et al., mid 80s) - confirmed by cognitive experimental studies (Purchase et al., 2000-2002) rich body of graph drawing techniques assume the input is a planar (planarized) graph and avoid edge crossings as much as possible

slide-7
SLIDE 7

The planarization handicap

for dense enough or constrained enough drawings, many edge crossing are unavoidable

FlyCircuit Database, NTHU

slide-8
SLIDE 8

Mutzel’s intuition about crossings

34 crossings: 24 crossings: minimum “skewness” (number of edges whose deletion makes it planar) minimum number of crossings

slide-9
SLIDE 9

Observations from eye tracking

  • No crossings: eye movements were smooth

and fast.

  • Large crossing angle: eye movements were

smooth, but a little slower.

  • Small crossing angle: eye movements were

very slow and no longer smooth (back-and- forth movements at crossing points).

Experiments of Eades,Hong,Huang

slide-10
SLIDE 10

Example

[Didimo, L., Romeo, “A Graph Drawing Application to Web Site Traffic Analysis”, JGAA 2011]

slide-11
SLIDE 11

Beyond planarity

the visual complexity not only depends on the number of crossings but also on the type of crossings challenge: compute drawings where some “bad” crossing configurations are forbidden (minimized)

slide-12
SLIDE 12

RAC h-PLANAR (h=3) h-QUASI-PLANAR (h=3)

Drawings with forbidden crossing configurations

SKEWNESS-h (h=1)

slide-13
SLIDE 13

Strong 1-visibility drawing Weak 1-visibility drawing

Drawings with forbidden crossing configurations

slide-14
SLIDE 14

Most explored research directions

Turán-type: find upper bounds on the edge density Recognition: how hard is it to test whether a graph admits a drawing with a forbidden configuration? Fáry-type: given a drawing (with jordan arcs), is there a straight-line drawing that preserves the given topology?

slide-15
SLIDE 15

New research directions

study the combinatorial relationships between different families of nearly planar graphs study trade-offs between crossing complexity and

  • ther aesthetic criteria
slide-16
SLIDE 16

Combinatorial relationships between nearly planar graphs

slide-17
SLIDE 17

1-planar WeB1 quasi-planar planar StB1 Caterp. K5 K6 K3,3 C4 K7 +K3,3 K9

[Evans et al., 2014 ]

K8

1-planarity, quasi-planarity and 1-visibility

slide-18
SLIDE 18

RAC Graphs 1-planar graphs Max dense RAC Graphs

RAC and 1-planarity

[Eades,L., 2013 ]

slide-19
SLIDE 19

Theorem A maximally dense RAC graph is 1-planar. Also, for every integer i such that i≥0 there exists a 1- planar graph with n=8+4i vertices and 4n-10 edges that is not a RAC graph. Finally, for every integer n > 85, there exists a RAC graph with n vertices that is not 1-planar. [Eades,L., 2013 ]

RAC graphs and 1-planarity

slide-20
SLIDE 20

Some details about the proof

slide-21
SLIDE 21

Preliminaries: edge coloring

  • red edges do not cross
  • each green edge crosses with a blue edge
  • red-blue (embedded planar) graph = red + blue edges
  • red-green (embedded planar) graph = red + green edges

G Grb Grg

slide-22
SLIDE 22

Preliminaries: Grb and Grg in a maximally dense RAC graph

G

Grb Grg

each internal face of Grb (Grg ) has at least two red edges [Didimo, Eades, L., 2011 ]

slide-23
SLIDE 23

Preliminaries: Grb and Grg in a maximal RAC graph

G

Grb Grg

Notation:

  • mr, mb, mg = number of red, blue, and green edges
  • frb = number of faces of the red-blue graph Grb

Assumption:

  • mg  mb
slide-24
SLIDE 24

Maximally dense RAC graphs are 1-planar

Approach:

  • suppose we can show that Grb and Grg are both

maximal planar graphs; then:

slide-25
SLIDE 25

Maximally dense RAC graphs are 1-planar

Approach:

  • suppose we can show that Grb and Grg are both

maximal planar graphs; then:

f1 f2

slide-26
SLIDE 26

Grb and Grg are maximal planar graphs (1)

  • the following is proven first:

Claim 1: the external face of Grb and Grg is a 3-cycle

  • then, we consider the internal faces of Grb that share at

least one edge with the external face (fence faces)

fence face

there are at least 1 and at most 3 fence faces

slide-27
SLIDE 27

Grb and Grg are maximal planar graphs (2)

  • ...and prove the following

Claim 2: If G is maximal, Grb has three fence faces and

each fence face is a 3-cycle

  • obs: at least two fence faces consist of red edges

    +  +   360°  < 90°

   90° and   90°

slide-28
SLIDE 28

Grb and Grb are maximal planar graphs (3)

since: (1) each internal face of Grb has at least 2 red edges; (2) the external face of Grb is a red 3-cycle; (3) at least two fence faces are red 3-cycles  2mr  2(frb – 3) + 3 + 3 +3 By Euler’s formula for planar graphs  mr + mb  n + frb – 2 mb  n – 4 mr  frb +2 since mr and frb are integers, we obtain

slide-29
SLIDE 29

Grb and Grb are maximal planar graphs (4)

G is a maximally dense RAC graph  mb + mr + mg = 4n – 10 mb  n - 4 since by assumption mg  mb and since both Grg and Grb are planar  Grg and Grb are both maximal planar graphs

Therefore a maximal RAC graph is 1-planar

mr + mg ≥ 3n-6

slide-30
SLIDE 30

RAC Graphs that are not 1-planar

There exists a graph G with less than 4n-10 such that G is a RAC graph but is not 1-planar

>=4

slide-31
SLIDE 31

RAC Graphs that are not 1-planar

There exists a graph G with less than 4n-10 such that G is a RAC graph but is not 1-planar

>=4 w u v z

slide-32
SLIDE 32

Not all 1-planar graphs with 4n-10 edges are maximal RAC

Go Gi

Go has n=8 vertices and 4n-10=22 edges; for i≥0, Gi has n=8+4i vertices and 4n-10 edges

Gi-1

slide-33
SLIDE 33

Not all 1-planar graphs with 4n-10 edges are maximal RAC

Go Gi

Go has n=8 vertices and 4n-10=22 edges; for i≥0, Gi has n=8+4i vertices and 4n-10 edges; they are 1-planar graphs we show that Gi cannot be realized as a RAC graph (by induction on i)

Gi-1

slide-34
SLIDE 34

Go is not RAC realizable

every vertex has degree 5 or 6. for every 3-cycle there is a K4 for every K4, there is a 4-cycle through the other vertices

slide-35
SLIDE 35

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-36
SLIDE 36

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-37
SLIDE 37

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-38
SLIDE 38

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-39
SLIDE 39

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-40
SLIDE 40

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-41
SLIDE 41

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-42
SLIDE 42

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-43
SLIDE 43

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-44
SLIDE 44

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-45
SLIDE 45

Go is not RAC realizable

if Go were RAC realizable, the external face of the realization would be a 3-cycle

slide-46
SLIDE 46

RAC Graphs 1-planar graphs Max dense RAC Graphs

…summarizing….

slide-47
SLIDE 47

Area Requirement Beyond Planarity

RAC h-planar skewness-h

slide-48
SLIDE 48

A result by Angelini et al.

RAC straight-line drawings of planar graphs may require quadratic area (Angelini et al., JGAA 2011)

slide-49
SLIDE 49

w

Area req. of h-planar drawings

3h u v z y

h-planar (constant h) straight-line drawings (and RAC straight-line drawings) of planar graphs may require quadratic area [Di Giacomo et al., 2012]

slide-50
SLIDE 50

Area req. of skewness-h drawings

G0 G1 Gh+1

O(n2) area, if planar

skewness-h (constant h) straight-line drawings of planar graphs may require quadratic area

slide-51
SLIDE 51

linear area upper bound

h-quasi-planar drawings 4-quasi-planar

slide-52
SLIDE 52

Bounded treewidth

G has treewidth ≤ k  G is a partial k-tree

3-tree

slide-53
SLIDE 53

Bounded treewidth

3-tree

G has treewidth ≤ k  G is a partial k-tree

slide-54
SLIDE 54

Bounded treewidth

3-tree

G has treewidth ≤ k  G is a partial k-tree

slide-55
SLIDE 55

Bounded treewidth

3-tree

G has treewidth ≤ k  G is a partial k-tree

slide-56
SLIDE 56

Bounded treewidth

partial 3-tree

G has treewidth ≤ k  G is a partial k-tree

slide-57
SLIDE 57

The good news

every n-vertex graph with bounded treewidth admits an h-quasi planar straight-line drawing in linear area such that the value of h does not depend

  • n n

[Di Giacomo, Didimo, L., Montecchiani, 2013]

slide-58
SLIDE 58

Di Giacomo et al., 2013

every h-colorable graph has a linear area s.l.drawing [Wood, CGTA, 2005]

Applying the result

slide-59
SLIDE 59

Ingredients

study the relationship between (c,t)-track layouts and h-quasi planar straight-line drawings new technique to compute a (2,t)-track layout of a partial k-tree

slide-60
SLIDE 60

Open problems

slide-61
SLIDE 61

Inclusion properties and RAC graphs

Recognizing those graphs that have a RAC drawing is NP-hard. Does this problem remain NP-hard for those graphs with n vertices and 4n-10 edges? Characterize those 1-planar graphs that have a RAC drawing

slide-62
SLIDE 62

Area-crossing complexity trade-offs

Do all planar graphs have a sub-quadratic area h- quasi planar straight-line drawing with constant h? Do partial k-trees admit a O(1)-quasi planar straight line drawing in linear area and constant aspect ratio? For, example, do outerplanar graphs admit a 3- quasi planar straight line drawing in linear area and constant aspect ratio?

slide-63
SLIDE 63

Other problem categories

Turan-type Recognition Fary-type RAC O(n) NP-hard (linear-time for 2-layer)

  • 1-planar

O(n) NP-hard (linear time for given rot. syst.)

  • charact. test,

drawing 3-quasi- planar O(n)

?? ??

skewness-1 O(n) polynomial

  • charact. test,

drawing