Information Technology Reference
In-Depth Information
4
Experimental Parameters
All experiments reported are derived from simulations on a grid 100x100, giving
a total of 10,000 possible types. This is of the same order of magnitude as those
experiments reported in [4] in which types were represented by a binary string
of length 13, resulting in 2
13
=8192 possible types. All experiments used a lower
threshold of 1000, the maximum upper threshold is stated in each experiment,
and took values
U
10
,
000
,
100
,
000
,
200
,
000. Connectivity is determined by the
radius of a cell
r
— the minimum radius allowed in heterogeneous experiments
is 10 which has been shown in previous work to be the percolation value, i.e.
the minimum radius at which a network is able to spring into existence. The
maximum radius is limited to 15 in heterogenous experiments, and is fixed at
15 in homogeneous experiments. Again, this value has been shown in previous
work to give interesting network behaviour.
r
= 15 allows a maximum of 708
potential partners; this compares to the maximum number of partners in [4] of
378, obtained by using a threshold of 9. At the lower radius limit of 10, there
a 316 potential partners. Due to lack of space, all experiment results obtained
cannot be shown here — typical results are presented to illustrate trends, and
more detailed results are expected to be presented in a forthcoming publication.
∈
1800
1800
Homogeneous
Heterogeneous
1600
1600
1400
1400
200,000
200,000
1200
1200
100,000
1000
1000
100,000
800
800
600
600
10,000
400
400
10,000
200
200
0
0
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
0
2000
4000
6000
8000
10000
Fig. 1.
Number of cells against time plotted for homogeneous and heterogeneous net-
works for values of
U ∈
10
K,
100
K
and 200
K
. All experiments are run using the same
seed value.
5
Experimental Results
In this section, we present results obtained from running simulations of the
model over 10,000 iterations. The growth model is clearly dynamic, therefore,
the network obtained at iteration 10,000 is merely a snap-shot of the network at
some moment in time, and it is unrealistic to assume that all networks will be
in the same state at the same moment in time. Therefore, where it makes sense,
results presented are averages over a number of runs of the simulation, other-
wise, they give a snap-shot of a particular individual run, but can be considered
representative of the general trend.
