Upload
chithu1234
View
216
Download
0
Embed Size (px)
Citation preview
7/28/2019 C1-5 (1)
1/7
On Node Ranking of Graphs
Yung-Ling Lai, Sheng-Hsiung Wang and Wei-Chun Lin
Department of Computer Science and Information Engineering
National Chiayi University, Chiayi, Taiwan
{ yllai, s0942783, s0942810}@mail.ncyu.edu.tw
Abstract
For a connected graph ( , )G V E= , a node
ranking labeling { }: 1,2,...C V k is a mappingsuch that for every path between any two vertices
u, v with ( ) ( )C u C v= , there exists at leastone vertex w on the path with
( ) ( ) ( )C w C u C v> = . The value ( )C v iscalled the rank of the vertex v and G is said to
be k-rankable if there exists a node ranking
labeling of G with maximum rank k . The
node ranking problem is the problem of findingminimum k such that G is k-rankable. There
are two versions of node ranking problem, namely
offline and online. In off-line version, the node
ranking problem deals with a graph where all
vertices and edges are given in advance. In
on-line version, the vertices are given one by one
in an arbitrary order and only the edges of the
induced subgraph of given vertices are known
when the rank of each vertex has to be chosen.
The rank can not be changed after it is assigned.
This paper solved node ranking problem of corona
graphs and mesh in off-line version, and gave
on-line node ranking algorithm for sun graphs.
1 Introduction
Let ( , )G V E= be an undirected graph where V
denotes the set of vertices and E denotes the set
of edges. G is said to be k-rankable if there is a
mapping { }: 1, 2,...C V k such that every pathbetween any two vertices u , v with
( ) ( )C u C v= , there exists at least one vertex w on the path with ( ) ( ) ( )C w C u C v> = . Themapping C is called a node ranking labeling of
G and the value ( )C v is called the rankof thevertex v . The node ranking number ( )r G isthe smallest integer k such that G is
k-rankable. A node ranking labeling is an
optimal node ranking labeling of G if its
maximum rank is ( )r G . The node rankingproblem is the problem of finding the node
ranking number ( )r G of a graph G . Wesaid the node ranking problem is solved if a
formula for the node ranking number is obtained
or an algorithm of an optimal node ranking
labeling is given. The node ranking problem is
applicable in communication network
design[3][15]; finding the minimum height of a
node separator tree of a graph [3][8][15], which
are extensively used in VLSI layout [11][14]; and
developing algorithms for planning efficient
assembly of products in manufacturing systems [4]
etc. There are two versions of node ranking
problem, off-line and on-line versions. In
off-line version, the graph with all vertices and
edges are given in advance. In on-line version,
the vertices are given one by one in an arbitrary
order together with the edges adjacent to the
vertices that are already given. The rank has to
be assigned in real time and once the rank is
assigned to a vertex, it is not changeable. Similar
to the offline version, the on-line node ranking
number is denoted as ( )*r G , which is thesmallest integer k such that G is on-line
k-rankable.
Node ranking problem has been studied since
1980s. In the offline version, it is known that for
a path nP , 1n , ( ) 2log 1r nP n = + [5].For a cycle
nC , 3n , ( )r 2log 1nC n = +
[1]. For wheel graphs 1,nW with n vertices on
the cycle, 3n , 2log)( 21 +=+ nWnr [7]; thenode ranking number of the complete bipartite
graph ,m nK is ( ), 1r m nK m = + for m n [8];and star graphs )(
1,1 =
nnKS with n vertices,
2)( =nr
S for 3n [13]. An upper bound of
the node ranking number for an arbitrary tree with
n nodes was given by [3], they proposed an
( )2logO n n time optimal node ranking algorithm,which was further improved to ( )O n by [12].There are fewer results in the on-line version. [1]
gave tight bounds for paths and cycles as
( )*2 21 2 1log logr nPn n+ < < + for 1n ;and ( )*2 2log 1 2 log 1r nn C n+ < < for
3n , respectively. The on-line version of a
complete bipartite graph with m n was proved
to be ( ) { }* ,1 min 1,2 1r m nm K n m+ + + [8]and the on-line node ranking number of star graph
is given by [13] as 3)(* =nr
S for 3n .
There are several papers worked on on-line trees,
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
108
7/28/2019 C1-5 (1)
2/7
the best known on-line node ranking algorithm for
trees which has ( )2O n time complexity and( )O n space complexity was given by [6]. [9]
proposed an online ranking algorithm for trees
using CREW PRAM model with ( )3 2logO n n processors. [10] provided an optimal online
ranking algorithm for stars and an
( )
3O n time
algorithm for arbitrary trees.
This paper established offline node ranking
number for mesh and the corona of any two
graphs and provided an on-line algorithm with
tight bound of online node ranking number for sun
graphs.
2 Offline Results for Corona Graphs
Given graphs G and H with n and m vertices
respectively, the corona of G with respect to H
denoted as HG , is the graph with vertex set
{( ) ( )V G H V G n = distinct copies of ( )V H denoted as ( ) ( )1 2, ,V H V H ( )}nV H and theedge set ( )E G H {( )E G n= distinct copiesof ( )E H denoted as ( ) ( )1 2, ,E H E H
( )}nE H {( , ) : ( ), ( ),i i iu v u V G v V H }1 i n . Figure 1 shows an example of graph
3 4P P .
Figure1: An example of graph3 4P P .
Theorem 1: Let HG denote the corona of
G with respect to H. Then the node rankingnumber ( ) ( ) ( )
r r rG H G H = + .
Proof: First we show that ( )r G H
( ) ( )r rG H + . Let 1f and 2f be optimal
node ranking labeling of G and H
respectively. Let ( ) { }1 2, , , nV u u uG = and
( ) { }1 2, , , mV H v v v= , ( ) ( )V G H V G =
{ ,i jv which is the i-th copy of the j-th vertex of
H , }1 , 1i n j m . Define f as
( ) ( ) ( )1i i rf u f u H= + for ( )iu V G and
( ) ( ), 2i j jv vf f= for ( ),i j iv V H , then f is a
node ranking labeling of HG with maximum
rank ( ) ( )r rG H + , which implies
( ) ( ) ( )r r rG H G H + .
Next we show that ( ) ( ) ( )r r rG H G H + .Suppose to the contrary, that
( ) ( ) ( )r r rG H G H < + . Let g be an
optimal node ranking of G H .
Case1: 1 ,1i n j m , ( ) ( ),i ji vg u g> .Then ( ) ( )i rg u H> , ( )iu V G . Define a
labeling 'g of G as '( ) ( ) ( )i i r
g u g u H = ,
then 'g is a node ranking labeling of G with
maximum rank ( ) ( ) ( )r r rG H GH for all 1 ,1i n j m . Note
that for each exchange, since it starts from the
smallest label, the order of the ranks are not
changed in graphi
H , and for ( ) ( )i kh u h u= we
must have somej that ( ) ( ),i j kg v g u= , since g is
a node ranking labeling of graph HG , there
must be a w oni k
u u path with
( )( ) ( )i kh w h u h u> = , which implies h is a node
ranking labeling of HG
. Since themaximum label ofh is the same as the maximum
label of g , which lead us back to case 1 and
produce a contradiction.
Let ,n mS be the graph which contains a n-cycle
and each cycle vertex has m hairs. Since the
graph has the shape of sun, we named it as sun
graph. Figure 2 shows an example of graph
5,2S .
Figure 2: An example of graph5,2S .
Since the sun graph ,n mS may be viewed as
n mC K , following corollary comes directly from
Theorem 1 by knowing that 2( ) log 1r nC n = +
(from [1]) and ( ) 1r mK = .
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
109
7/28/2019 C1-5 (1)
3/7
Corollary 1: Let ,n mS be the sun graph of
order ( 1)n m + . Then the node ranking number
, 2( ) log 2r n mS n = + .
3 Offline Results for Mesh Graphs
In this section, we proposed an off-line ranking
algorithm for mesh graphsn n
P P . The
algorithm produced a node ranking labeling of
n nP P for odd n and 15n , for even n and
18n . Figure 3 shows an example of the graph
6 6P P .
Figure 3: An example of graph6 6
P P .
Algorithm Offline_Ranking_Mesh
Input : , 15n n if n is odd and 18n if n is
even. Each vertex is denoted as,i j
v for
1 ,i j n .
Output : An off-line rank assignment ofn n
P P .
Method:
[ ] [ ]array y x holds the label of vertex ,y xv If( n is odd ) Then
Step1: If x y even+ = Then [ ] [ ] 1array y x = .Step2: For 1x = to 2n
do
Dividing the graph by x y n+ = .
For 2x n= to n doDividing the graph by 2x y n+ = + .
Step3: For 2x = to 2n do
Dividing the graph by 1x y = .
For 2 1x n= + to 1n doDividing the graph by 1y x = .
// The graph is then divided into four isomorphic
subgraphs 1 2 3 4, , , as shown in Figure 4.
Figure 4: An example of graphs 1 2 3 4, , , .
Step4: In the 1 graph, 2 1i n= ; call Alpha
( i ); use the label for corresponding vertices in 2
to4
.
// i means the number of nodes on the edge of
subgraph1
.
Else If( n is even and 0 mod4n ) Then
Step1: If x y even+ = Then [ ] [ ] 1array y x = .Step2: For 4 3x n= + to 2 4n + do
Dividing the graph by 2 5x y n+ = + .
For 4 3x n= + to 2 1n + do
Dividing the graph by 1x y = .
For 2x n= to 2 1n + doDividing the graph by 1x y n+ = + .
For 2x n= to 3 4 2n do
Dividing the graph by 1y x = .
For 2 3x n= to 3 4 2n do
Dividing the graph by
03 2 3x y n u+ = = .
// The graph is divided into two isomorphic
subgraphs1 2, as shown in Figure 5.
Figure 5: An example of graphs1 2, .
Step3: In the1
graph, Dividing the graph by
12 5x y n u+ = + = . (Repeat the same processes
for the corresponding vertices in2
.)// The graph is divided into two subgraphs
1 2, asshown in Figure 6.
Figure 6: An example of graphs1 2, .
Step4: Dividing1 and 2 by 1x y = .and
02 1y x n v = = respectively.
// The1
graph is divided into two subgraphs
1 2, and the 2 graph is divided into two
subgraphs1 2, as shown in Figure 7.
Figure 7: An example of graphs 1 2 1 2, , , .
Step5: In the1
graph, 4i n= ; callAlpha ( i );In the
2 graph, ( )3 12 8j n= ; call
Alpha ( j );
In1
and2
, there are lines of vertices
denoted as 1 2, , , le e e , such that
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
110
7/28/2019 C1-5 (1)
4/7
( )1 0 12 , 1 2 1x y u c c u u+ = + where
( )3 4 4 3 12 8l n n= . Let 2j l= .
Dividing1
and2
byj
e and1j
e+
respectively.
Let2
u x y= + for the vertices ,y xv on line
je and
3u x y= + for the vertices ,y xv on line
1je
+.
// The1
graph is divided into two subgraphs
11 12, and the 2 graph is divided into twosubgraphs
21 22, as shown in Figure 8.
Figure 8: An example of graphs 11 12 21 22, , , .
Step6: In11
, Dividing the graph by
12 2 2 1y x n n v = + = .// Graph
11 is divided into two subgraphs 11 11,
as shown in Figure 9.
In12
, Dividing the graph by
( ) 22 2 4 16 1y x n n v = + + = .// Graph
12 is divided into two subgraphs
12 12, as shown in Figure 9.
In21
, ( )3 1 2 1p u u= ; ( )0 1 2q v= ;callBeta ( ,p q );
In22 , ( )0 3 2 1p u u= ; ( )0 3 2q v= ;
callBeta ( ,p q );
Figure 9: An example of graphs11 11 12 12, , , .
Step7: In11
, ( )0 1 2 2i u v= ; callAlpha ( i );In
11 , let ( )2 1 2 1p u u= ;
( )1 0 2 1q v v= ; callBeta ( ,p q );In 12 , ( )0 2 2 2i u u= ; callAlpha ( i );
In 12 , let ( )0 2 2 1p u u= ;( )2 0 2 1q v v= ; callBeta ( ,p q );
Else //n is even and 0 mod 4n .
Step1: If x y even+ = Then [ ] [ ] 1array y x = .Step2: For 4 3x n= + to 2 3n + do
Dividing the graph by 2 4x y n+ = + .
For 4 3x n= + to 2 1n + do
Dividing the graph by 1x y = .
For 2x n= to 2 1n + do
Dividing the graph by 1x y n+ = + .
For 2x n= to 3 4n doDividing the graph by 1y x = .
For 2 2x n= to 3 4n doDividing the graph by
03 2 2x y n u+ = = .
// The graph is divided into two isomorphicsubgraphs
1 2, which is the same as in Figure 5.Step3: In the
1 graph, Dividing the graph by
( ) 12 2 2 8x y n n u+ = + = . (Repeat the sameprocesses for the corresponding vertices in
2 .)
// The graph is divided into two subgraphs1 2, as
shown in Figure 6.
Step4: Dividing 1 and 2 by 1x y = and
02y x n v = = respectively.
// The graph1
is divided into two subgraphs
1 2, and the graph 2 is divided into twosubgraphs
1 2, as shown in Figure 7.Step5: In
1 , 4i n= ; callAlpha ( i );
In2
,
( )3 4 1 2i n =
; callAlpha ( i );
In1 and 2 , there are lines of vertices
1 2, , , le e e such that 1 2x y u c+ = + , where
( )0 11 2 1c u u , ( ) ( )2 8 2 4 3l n n= + + + .Let 2j l= , dividing 1 and 2 by
and 1je + respectively.
Let 2u x y= + for the vertices ,y xv on line
je and 3u x y= + for the vertices ,y xv on line
1je
+.
// Graphs 1 and 2 are divided into subgraphs
11 12, and 21 22, respectively, as shown inFigure 8.
Step6: Dividing the graph 11 by
( ) 12 2 2 8 2y x n n v = + + =
.// Graph 11 is divided into two subgraphs 11 11, as shown in Figure 9.
Dividing graph 12 by
( ) 22 2 2 16y x n n v = + + = .// Graph 12 is divided into two subgraphs 12 12, as shown in Figure 9.
In graph 21 , let ( )3 1 2 1p u u= ;( )0 1 2q v= ; callBeta ( ,p q );
In graph 22 , let ( )0 3 2 1p u u= ;( )0 3 2q v= ; callBeta ( ,p q );
Step7: In graph 11 , let ( )0 1 2 2i u v= ; callAlpha ( i );
In graph 11 , let ( )2 1 2 1p u u= ;
( )1 0 2 1q v v= ; callBeta ( ,p q );In 12 , let ( )0 2 2 2i u u= ; callAlpha ( i );In graph 12 , let ( )0 2 2 1p u u= ;
( )2 0 2 1q v v= ; callBeta ( ,p q );End of Algorithm Offline_Ranking_Mesh_Graph
Procedure Alpha ( int i )
If ( )3i Then {3p i= ; 1q i p= ;
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
111
7/28/2019 C1-5 (1)
5/7
// p means the number of nodes of the
dividing line P.
1 1p p p= + ;
Dividing the graph by1: 2 2 1I y x p c = + .
Dividing the graph by1: 2 2 1P x y p c+ = + + .
// The graph1 is divided into three
subgraphs 11, 12, 13 as shown in Figure 10.
Figure 10: Subgraphs 11, 12, 13 .
In graph11
, callAlpha (p );
If ( ) ( )( )( 11r r pP p + and( ) ( )( )11 2 1r r pP and( )( )
)11r pP Then {
In the 12 graph, dividing thegraph by 0 : 4a x y n+ = .
2q q= ;
}
Else If ( ) ( )( )( 11r r pP p + and( ) ( )( )11 2 1r r pP > and ( )( ))11r pP
Then {
Dividing graph 12 by
0 : 6a x y n+ = .
3q q= ;
}
Else {
Dividing graph12 by
0 : 8a x y n+ = .4q q= ;
}
// The graph 12 is divided into twosubgraphs 1 2, as shown in Figure 11.
Figure 11: An example of graphs1 2, .
In 1 , callBeta ( ,p q );1i i p= ; c + + ;
In 2 , callBeta ( ,p q );In
13 , callAlpha ( i );
Return ( )1r ; // ranking of 1 }
Else If ( )2i = Then Rank the vertices individuallywith 2, 3, 4.
Else Rank the vertex with 2. // 1i =
End of Procedure Alpha
Procedure Beta ( int p , int q )
If ( )3q Then {In
1 , there are dividing lines : 1 2, , , qa a a
as shown in Figure 12.
Figure 12: An example of dividing lines
1 2, , , qa a a .
If q is odd Then {
Dividing the graph by2q
a
.
// The dividing line divides the graph into
two the same graphs.
2q q= ;
}If q is even Then {
Dividing the graph by q 2 +1a .
2q q= ;
}
}
Else If ( )2q = Then call Gamma( p );Else Rank by optimal ranking as path // 1q =
callBeta( ,p q );
End of Procedure Beta
Procedure Gamma ( int p )
If ( )3p Then {
// graph 1 is divided into two subgraphs: 1 2, by1a .
In1
, there are dividing lines: 1 2, , , pb b b .
Figure 13 shows1 2, and 1 2, , , pb b b .
Figure 13: Graphs1 2, and
1 2, , ,
p
b b b .
If p is odd Then {
Dividing the graph byp 2
b
.
// The graph is divided into two identical
subgraphs.
2p p= ;
}
If p is even Then {
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
112
7/28/2019 C1-5 (1)
6/7
Dividing the graph by 2 1pb + .
2p p= ;
}
}
Else If ( )2p = Then Rank the nodes using 2 to 5.Else Rank the nodes individually with 2 and 3.
call Gamma (p );
End of Procedure Gamma
Theorem 2: The labeling given by Algorithm
Offline_Ranking_Mesh is a node ranking labeling of
n nP P for odd n and 15n , for even n and
18n .
Proof: In the algorithm, we always assign greater
rank for the dividing vertices, and on every path
between two nodes of the same rank there is at least
one dividing line, so the labels given by the
algorithm must follow the node ranking condition for
n nP P .
Theorem 3: The maximum rank given by Algorithm
Offline_Ranking_Mesh for odd n and 15n is
( ) ( ) ( )2 1 6 1 6 8 1 12n n n n n+ + + + + + 10 x + ; and for even n, 0 mod4n and 20n
is ( ) ( ){4 2 8 1 max 3 4 2 8 ,n n n n n+ + + + }2 4 8 5n n+ ; and for even n 0 mod4n and
18n is ( )4 2 8 1n n n+ +
( ){ ( )max 3 4 2 8 ,2 4 3 12n n n n+ + + + ( ) ( ) }3 12 8 3 24 9n n x+ + + + , where
0 if 0, 2
2 if 4
4 if 6,8
5 if 10
y
yx
y
y
=
==
= =
for ( )( )3 12 3 12y n n= + .
Theorem 3 gave us an upper bound for the node
ranking number of a meshn n
P P for 17n .
For 16n , Table 1 shows the node ranking number
ofn n
P P which is obtained by a brute force
computer testing, and is denoted by the formula that
represents the relation between n and ( )r n nP P .
Table 1: ( )r n nP P for 16n .n ( )r n nP P
1 7n 2 1n
8 10n 2n
11 2 1n +
12 2n
13 2 2n +
14 2 1n +
15 2 3n +
16 2 2n +
4 Online Results
In this section, we proposed an on-line ranking
algorithm for sun graphs,n m n mS C K= , and
analyze both time and space complexities of our
algorithm. Notice that there are ( 1)n m +
vertices in ,n mS . Let 1v to ( 1)n mv + represent
the order of the coming vertices.
Algorithm Online_Ranking_Sun_Graph
We use structure array Sun[] to record the
information of each node, and use value_record[]
as the counter of the index values appearance.
Input:
n: the number of vertices on the cycle of,n mS .
m: the number of vertices on the hairs of ,n mS .
iv for 1 ( 1)i n m + the vertices together with
the adjacent verticesj
v for 1 j i < .
Output:
An online node ranking labeling of ,n mS .
Method:For 1i = to ( )1m n+ do {
Ifi
v has no adjacent vertices, ( ) 2i
rank v =
Else ifi
v has only one neighborj
v for
j i< andj
v has two neighbors with
degree >1 then ( ) 1i
rank v =
Else {
1. Set the occurrence record to zero except rank 1(which is set to 2)
2. Applying online tree algorithm given in [5] toloop through each adjacent path, go as long as
possible, to find the max value j that has
occurred at least twice
3.
Find minimum available rank k that is greaterthanj
4. Set ( )i
rank v k =
}
}
End of Algorithm Online_Ranking_Sun_Graph
Theorem 4: Algorithm Online_Ranking_Sun has
( )( )3O mn time complexity and ( )O mn spacecomplexity.
Proof:
First we analyze the time complexity of the
algorithm. Since a Sun graph with ( )1m n+ verticeshas exactly ( )1m n+ edges, each time when insert anew node iv , the looping for finding max j usedonline tree algorithm which takes 2( )O i times , so
after ( )1m n+ vertices inserted, the whole processwill be at most ( )( )
22 21 2 1 nm+ + + +
( )( )3O mn= times. For the space complexity, weuse structure array with ( )1m n+ elements to storethe node plus one array to record the rank occurrence,
the total space complexity then is
( ) ( ) ( )O mn O mn O mn+ = .
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
113
7/28/2019 C1-5 (1)
7/7
Theorem 5: For a sun graph,n mS with 7n is
( )*2 , 2log 2 log 6r n mn S n m+ + .Proof:
Since the online node ranking number is at least as
much as the offline node ranking number, we know
that ( )*2 ,log 2 r n mn S+ . Let ,n mS be thegraph which contains a n-cycle for 1
iv i n and
each cycle vertex has m hairs denoted as, for 1 ,1i ju i n j m . If we use adversary
analysis, for the best we can do, saving rank 1 for the
vertices which are known as hair, assume that the
vertices of,n mS are coming as follows:
1 2 3 4, , ,v v v v coming in first, so they get labels
2,3,2,4.
For 3 3i n , all ,i ju for 1 j m comes before
2iv
+.
Then comes2, for 1n ju j m and
, for 1,2, 1 and 1i ju i n j m= .
At last,n
v then, for 1n ju j m .
In this coming order, beforen
v comes in, no vertex
can be recognized as end vertex, hence online tree
algorithm worked for this case which is the worst
case and our algorithm guarantee the maximum rank
is no more than 2log 6n m+ , hence we have
( )* , 2log 6r n mS n m + , which complete the proof.
5 Conclusion
There are several papers discussed the node
ranking problem. In this paper, we presented
off-line node ranking number for the corona graphs
and gave labeling algorithm for mesh. We also
gave on-line algorithm for sun graphs. Both timeand space complexities of our algorithms are
analyzed.
References
[1] E. Bruoth and M. Hornak, On-line rankingnumber for cycles and paths, Discussiones
Mathematicae, Graph Theory 19, pp.
175-197, 1999.
[2] J.S. Deogun, T. Kloks, D. Kratsch, and H.Muller, On vertex ranking for permutation
and other graphs, Lecture Notes in
Computer Science 775, pp. 747-758, 1994.[3] A. V. Iyer, Optimal node ranking of trees,
Information Processing Letters 28, pp.
225-229, 1988.
[4] Md. A. Kashem, X. Zhou, T. Nishizeki,Algorithm for generalized vertex-rankings
of partial k-trees, Theoretical Computer
Science 240, pp. 407-427, 2000.
[5] M. Katchalski, W. McCuaig, and S. Seager,Ordered colorings, Discrete Mathematics
142, pp. 141-154, 1995.
[6] Y.-L. Lai and Y.-M. Chen, An ImprovesOn-line Node Ranking Algorithm of Trees,
Proceedings of the 23rd
Workshop on
Combinatorial Mathematics and
Computation Theory, pp. 345-348, April
19-20, 2006.
[7] Y.-L. Lai and Y.-M. Chen, On NodeRanking of Graphs under Strong
Orientation, Proceedings of the 24th
Workshop on Combinatorial Mathematics
and Computation Theory, April 27-28, 2007.
[8] Y.-L. Lai and Y.-M. Chen, On NodeRanking of Complete r-partite Graph,
Journal of Combinatorial Mathematics and
Combinatorial Computing 67, 2008.
[9] C. Lee and J. S. Juan, Parallel Algorithm forOn-Line Ranking in Trees, Proc. Of the 22nd
Workshop on Combinatorial Mathematics
and Computational Theory, May 20-21, pp.
151-156, 2005.
[10]C. Lee and J. S. Juan, Jun. 2005, On-LineRanking Algorithms for Trees, Proc. of Int.
Conf. on Foundations of Computer Science,
Monte Carlo Resort, Las Vegas, Nevada,
USA,. June 27-30, pp. 46-51, 2005.
[11]C.E. Leiserson, Area-efficient graph layouts(for VLSI), Proceedings of the 21st Annual
IEEE Symposium on Foundations of
Computer Science, pp. 270-281, 1980.
[12]A. A. Schaffer, Optimal node ranking oftrees in linear time, Information Processing
Letters 33, pp. 91-96, 1989.
[13]G. Semanisin and R. Sotak, Kosice, A noteon on-line ranking number of graphs,Czechoslovak Mathematical Journal 56, pp.
591-599, 2006.
[14]A. Sen, On a graph partition problem withapplication to VLSI layout, Information
Processing Letters 43, pp. 87-94, 1992.
[15]P. de la Torre, R. Greenlaw, A. A. Schffer,Optimal edge ranking of trees in polynomial
time,Algorithmca 13, pp. 592-618, 1995.
T e 26t Wor s op on Comi natori a Mat emati cs an Computati on T eory
114