1 An octal degree graph representation for the rectangular dissections...

Preview:

Citation preview

1

An octal degree graph representation for the rectangular dissections

有田友和(桜美林大学)、本橋友江(関東学院大)

土田賢省(東洋大)、夜久竹夫(日大)

2004年12月18日 WAAP 125

日本大学文理学部 Title corrected

2

資料

IASTED   SEA  02 (Cambridge, 2002) 

HJ2003 (Hangary-Japan Sympos. Discrete Math., Tokyo)

IASTED AI04 (Insbruck, 2004)

3

対象 均一型矩形分割

不均一型矩形分割

操作(壁指向) 変形(壁移動、セル・行・列の追加・削除・移動、など)

特徴抽出(表の行合計計算、表構造の正誤など) 応用

図表(文書 ) 建物・フロアプラン( OR) 地形図

1 1 11 1 11 1

1 2 22 2 2

2

4

1.Introduction Motivation Aim Known Representation

Method Purpose

5

Motivation Tables

Homogeneous Heterogeneous rectangular rectangulardissections dissection

Editing operations often cause unexpected results.

6

Motivation (continues) Column Insertion at Right to Cell 1

Unexpected result Expected result

C3

B2

A1

B

C3

2

1 A

C3

B2

A1

Word

7

Motivation

Excel does not allow this operation

300100

BA

70 200

300100

100

30

BA

8

Aim Representation method for

rectangular dissection processing system.

Formalization of rectangular dissection editing operations.

9

Known Representation Methods

Quad-Tree Representation [J.L.Bentley,1975] for Search Algorithm

Rectangular Dual Graph Representation [Kozminski,K and Kinnen,E,1984] for

Plant Layout

10

Known Representation Methods

Quad-Tree Representation

NE NW

SWSE

NE NW SW SE

11

Known Representation Methods Rectangular Dual Graph

Representation

Horizontal edgeVertical edge

12

3 4

56

W

N

E

S

643

12

Known Representation Methods

Example 1 Quad-tree Rep. has a weak power of expression.

13

Known Representation Methods

Example 2 Rectangular Dual Graph Rep. may require higher complexity in editing operation.

14

Known results (cont.)

Theorem. Decision problem of a graph

to be a (homogeneous) grid graph →O(n) (1990)

15

Relate Results

Data Structures : Tessellation graphs (Motohashi et. al. FOSE02-Matsuyama)

Viewer (Kirishima et. al., LA02Summer, IASTED SE 02-Cambridge-USA)

Equivalent condition of graphs to be tessellation graphs (Kirishima et. al.,  IASTED AI 03-Insbruck)

16

Purpose To propose a graph

representation method for tables in consideration of editing and drawing and to investigate mathematical properties.

To introduce typical algorithms on the graphs and evaluate their complexity.

To introduce a graph grammar

17

2. Attribute Graphs for tables

18

Example

19

Def 2.1

Table T (2,3)-table Partition P {(1,1),(2,1)},{(1,2)}, . . . Grid g=(grow,gcolumn)

north wall of c nw(c)=1, sw(c)=2, east wall of c ew(c)=6, ww(c)=2.

0 2 4 60

21

cell c

20

Def 2.2 Tabular Diagram D=(T,P,g)

perimeter cellswith width=0 or height=0

We consider tabular diagramswith perimeter cells.

0 0 2 4 6 60

0

1

2

2

21

Def 2.3 (Tessellation Graph)A tabular diagram D=(T,P,g) is

represented by an attribute graph G=(V,E,L,λ,A,α)

1st step put a node to a cell

D G

d c

vd

vc

22

Graph representation 2nd step set edges and edge

labels nw(c) =nw(d)

sw(c) =sw(d)

ew(c) =ew(d)

ww(c) =ww(d)

D G

c d

c d

c

d

c

d

North wall edge South Wall Edge

East Wall Edge

West Wall Edge

23

Graph representation

3rd step set an attribute

D G

vc(1,2,6,2)

nw(c)=1 sw(c)=2ew(c)=6 ww(c)=2

c

2 6

12

24

Tessellation GraphExample

D GD

The tessellation Graphfor D □

d

vd

vc

d c

25

Properties D : a tabular diagram of an (n,m)-table

GD : a tessellation graph for D. 1 node in GD corresponds to 1 cell

in D The degree of a node in GD ≦ 8.

Proposition 2.22|ED| = 6(2n-4) + 6(2m-4) + 8k + 16 (k : the number of non perimeter cells)

26

 3D :rectangularpiped dissection: 24 degree quasi grid graphs

Horizontal 4 links (black) x nsew 4= 16 links

Vertical 4 links (pink) x up-down 2 = 8 links 

Total 24 links

27

Similar results to Proposition 2.2 hold for the 16 degree quasi grid graphs ; degree ≦ 16

for the 24 quasi grid graphs ; degree ≦ 24.

28

Proposition 2.2 GD:a rectangular piped dissection graph 

for D of an (m,n,l)-cube.k : the number of inner cells in GD.For #ED, 2x#ED = 

20(n-2)(m-2)x2 + 20(m-2)(l-2)x2 + 20(l-2)(n-2)x2+ 16(n-2)x4 + 16(m-2)x4 + 16(l-2)x4+ 12 x 8+ 24k

29

Proposition (cf. Kundu)∃RD : A rectangular dissection with no wall crossing s.t. RD is not a T*-plan but is represented in a tessellation diagram

30

Proof.A spiral graph with perimeter cells and Its corresponding tessellation graphs

31

3. Algorithms

32

RemarkAttributes for location of inner cells : not necessary

← Location information cited in perimeter cells

33

3.1 ALGORITHM Unify-Cells(1) Input/Output Spec. Example

Input OutputGD=(VD,ED,L,λD,A,αD)

GE=(VE,EE,L,λE,A,αE)

vc, vd

D E

d

c c

34

ccdcdc

(2) Illustration of Unify-Cells Mechanism

GD GE

(3) Theorem (Unify-Cells)Algorithm Unify-Cells runs in constant time.

35

3.2 ALGORITHM Move-East-Wall

(1) INPUT/OUTPUT Spec. Exampleδ

D E

Moved Wall

c

Input OutputGD=(VD,ED,L,λD,A,αD)

GE=(VE,EE,L,λE,A,αE)

vcδ≧0 : movement value

cc cc

36

cc cc

Traverse nodes upward through east wall edges .

Update east wall of each node, through east wall edges .

Update west wall of nodes in the right side column.

(2) Illustration of Move-East-Wall mechanisms

37

3.3 ALGORITHM Insert-Column

D E

Inserted Column

Input OutputGD=(VD,ED,L,λD,A,αD)

GE=(VE,EE,L,λE,A,αE)

vc

cc cc

38

Traverse nodes upward through west wall edges .

Add nodes and change edges.

Apply Move- East-Wall.

Insert-Column (cont)

cc cc

39

complexity

e.w.movement

O(n)≦2n1cell-1node≦8Tessella

tionGraph

O(nm)≦nm1cell-1node

≦2n+2m

Rec.DualGraph

O(nm)≦nm1block-1node

≦ 5QuadTree

cell visit

cell-node rel.

node degree

model

3.4 2D Algorithms Comparison

40

Algorithm Comparison (cont.)

O(1)O(n+m)Tessellation Graph

O(nm)O(nm)Rec. Dual Graph

O(nm)O(nm)Quad Tree

Unify cellsinsert column

model

41

3.5 Theorem 3 [10, 2003]∃CS graph grammars to

generates the tessellation graphs,and the grid graphs, respectively.

Cf. There is no NLC graph grammar that generates the grid graphs (Janssen & Rozenberg).

42

APPENDIX Multi Dim. Dissections

[2004, SCAI04]

Multi layer 2D: Multi layer rectangular dissections( Stratum diagram)→16 degree quasi grid graphs

3D : Rectangular parallel piped dissections(Architect)→24 degree quasi grid graphs

Multi Dimension : General case→Wall linked quasi grid graphs

43

A1. Multi layer rectangular dissection: 16 degree quasi grid graph

Horizontal 2 links x nsew = 8 links

Nsew 4 links x updown 2 = 8 links

Total 16 links

44

A2. 3D :rectangularpiped dissection

: 24 degree quasi grid graphs

Horizontal 4 links (black) x nsew 4= 16 links

Vertical 4 links (pink) x up-down 2 = 8 links 

Total 24 links

45

Similar results to Proposition 2.2 hold for the 16 degree quasi grid graphs ; degree ≦ 16

for the 24 quasi grid graphs ; degree ≦ 24.

46

Related works• Browsers

– SVG and XML browsers [9, 2002]

• Quasi grid graphs SVG & XML

47

4. Conclusion

48

Summary Quasi grid graphs called tessellation

graphs for the rectangular dissections The node degrees ≦ 8.

Typical algorithms are shown, and their complexity approved.

A CS graph grammar that characterize the tessellation graphs is introduced.

Attribute graphs for 3D rectangular piped dissection; The node degrees ≦24

49

Accession to the Aim Representation Method for Table Processing system is proposed.

Table editing operations are formalized.

50

Future Works 1 Processing Systems on XML & SVGParsable graph grammars for tessellation graph

51

Future Works 2

~地形図への応用~多層の地形構造のためのデータ構造

地表プラス水面(2層)地層図

立体型の地形構造のためのデータ構造立体地層図

ブラウザ51

Recommended