Next: Concluding Remarks
Up: Time-Stamped Graphs and Their
Previous: The Restricted Question
Bounds for
In this section we give some upper bounds for
. We start with the
following simple bound.
Proof.
Let

be the number of vertices of

. We construct

as follows.
Replace each arc

in

by two
undirected edges,

with time-stamp

and

with time-stamp

, where

is a new vertex.
Then

has

vertices and all time-stamps are distinct.
It is easy to see that

is the subdigraph of

induced by the vertices
of

. Hence

.
For trees, we can improve this bound.
We consider an undirected tree
with
vertices as a digraph.
For the tree with two vertices in Figure 5a,
Figure 5b gives a time-stamped graph whose
associated influence digraph contains Figure 5a as an induced
subgraph. Although it is not
optimal, it provides a basic structure to use in an arbitrary tree.
We call the
graph in Figure 5b a tadpole, with
its head and
its tail. The edge incident
to
, called the tail-edge,
has a medium time-stamp, whereas the edges incident to
have
a small time-stamp and a large time-stamp.
Figure 5:
A tree with two vertices and a time-stamped graph
 |
To apply this idea to an arbitrary tree, we root the tree at some vertex.
We now carry out the procedure described in Figure 5,
replacing each edge by a tadpole,
with an adjustment so that there are no repeated time-stamps,
working from the top down to guarantee that no two tails coincide
(see Figure 6 and Figure 7).
It is easy to check that the resulting graph
is a time-stamped graph whose associated influence digraph
contains
as an induced subdigraph.
This gives the following result.
Figure 6:
A rooted tree
 |
Figure 7:
A time-stamped graph with seven new vertices
whose associated influence digraph contains Figure 6 as an induced
subdigraph
 |
This is better than the bound given in Proposition 5.1.
However, this bound is not optimal (even for trees), since
.
To improve this bound, we note that in our construction,
we just have to avoid two tails coinciding. Every vertex is
a tail except the root. So we will not apply the procedure at one of the edges
incident to the root; this edge, called a special edge,
will have a medium time-stamp (see Figure 8).
Thus we have
This is optimal for a tree with two vertices. Can we do better in general?
We note that our construction is not unique. Consider a path on four
vertices. Figure 9 gives two additional constructions. Consider
a star of size 3. Figure 10 gives a different construction.
As we will prove next, this bound is optimal for stars. We conjecture that it
is optimal for trees.
Figure 8:
A time-stamped graph with six new vertices
whose associated influence digraph contains Figure 6 as an induced
subdigraph
 |
Figure 9:
Both the second and the third graph are time-stamped graphs whose
associated influence digraph has the first graph as an induced subdigraph. Both
are different from the construction given in the proof of
Theorem 5.3.
 |
Figure 10:
The second graph is a time-stamped graph whose associated influence
digraph has the first graph as an induced subdigraph. The construction is
different from the construction given in the proof of Theorem 5.3.
 |
Theorem 5.4
Let
be a star with
vertices. Then
.
Proof.
This follows directly from Theorem
5.3 and Proposition
4.6.
Conjecture 5.5
Let
be a tree with
vertices. Then
.
We can now give a better bound on
for graphs, based on the bound
for trees.
Proposition 5.6
Let
be a digraph with
vertices and
arcs.
Let
be the set of edges of
, i.e., the pairs of vertices joined by
antiparallel arcs. Suppose
is a connected graph.
Then
.
Proof.
Let

be a spanning tree of

. We apply the procedure given in
the proof of
Theorem
5.3 to obtain a time-stamped graph whose
associated influence digraph is the subdigraph of

induced by

.
Moreover, we choose the time-stamps on the
tail-edges of all the tadpoles to be 4, the time-stamps on the two edges
of each tadpole incident to its head to be 1 and 6 respectively,
and the time-stamp on the special edge to be 3.
Let

be

with all the arcs forming the edges in

removed.
For every arc in

, we apply the procedure given in the proof
of Proposition
5.1
with the additional constraint that one of the two time-stamps generated by such
an arc is 2 while the other is 5. It is routine to check that
the subgraph induced by

of the digraph associated with this time-stamped
graph is

. Every arc generated one new vertex, with
the exception that every edge (two arcs) in

other than the special edge
generated only one new vertex and the special edge (two arcs) generated
none. This gives a saving of

new vertices, so

, as required.
Another way of looking at Proposition 5.6 is that we have a special edge
that doesn't generate any new vertices, and we have trees rooted at vertices
of this special edge, whose edges generate one vertex for each edge (two
arcs) rather than two. We can use any clique in place of this special edge,
leading to the following result.
Theorem 5.7
Let
be a digraph with
vertices and
arcs.
Let
be the set of edges of
, i.e., the pairs of vertices joined by
antiparallel arcs. Suppose
is a connected graph with a clique of
size
. Then
.
Proof.
Let

be a clique of size

with vertices

in

. Let

be the trees in
a spanning forest of

rooted at

respectively. It is clear that such a spanning forest exists.
We apply the procedure given in the proof of Proposition
5.2
(not Theorem
5.3) to obtain a time-stamped graph

for
each

such that the subdigraph of

induced by

is

. Moreover,
we choose the time-stamps on the tail-edges of all the tadpoles to be 4, and
the time-stamps on the two edges of each tadpole incident to its head to be 1
and 6 respectively.
Each edge in

is given time-stamp 3.
Let

be the set of arcs not in

.
For every arc in

, we apply the procedure given in the proof of
Proposition
5.1, but
with the two time-stamps generated by such
an arc being 2 and 5. It is routine to check that
the subdigraph induced by

of the digraph associated with this time-stamped
graph is

. Since every arc generated one new vertex with
the exception that every edge (two arcs) in

generated only one new vertex and every edge (two arcs) in

generated none, we have a saving of

vertices, so

, as required.
A remark similar to Remark 4.9 can be made to the construction given
in the proof of Proposition 5.6 and Theorem 5.7.
In fact, Theorem 5.7 can be strengthened with only a slight
modification. We state this generalization without proof.
Next: Concluding Remarks
Up: Time-Stamped Graphs and Their
Previous: The Restricted Question
Eddie Cheng
2003-05-01