Databases Reference
In-Depth Information
FIGURE 4.9: Non-uniform distributed HGN approach with 7-21-7 composi-
tions for 35-element patterns with two possible values. (With kind permission
from John Wiley & Sons, Inc.: Mobile Intelligence, “An Online Scheme for
Threat Detection Within Mobile Ad Hoc Networks,”pp. 380-411, 2010, Khan,
A. I. and Muhamad Amin, A. H. and Raja Mahmood, R. A., Figure 17.13,
http://dx.doi.org/10.1002/9780470579398.ch17.)
The non-uniform distribution takes the environment into account. Some
parts of the network might have lower power resources, and thus their process-
ing capabilities are lower than other parts of the network. With this scenario
in mind, the effect of an unbalanced composition on the pattern recognition
accuracy of the distributed approach is analyzed.
The results of this simulation show that the non-uniform model offers an
almost equivalent level of accuracy to the HGN. Furthermore, it requires less
neurons in its composition. The number of neurons required for a single HGN
composition can be derived from Equation 4.1. The number of neurons re-
quired, n
for s subnets in a distributed HGN composition for a pat-
tern of size S = a is determined using the following equation:
x
(all,s)
2
2
2
s
x
(all,s)
a
1
+ 1
2
a
2
+ 1
2
a
s
+ 1
2
n
= v
+
+ . . . +
; for
a
i
i=1
s
2
a
i
+ 1
2
= v
(4.12)
i=1
Note that the squared term in Equation 4.12 is be substantially smaller
than the squared term in Equation 4.1 for the same sized problem, resulting
in fewer required neurons.
The mapping process in our simulation begins with the input of the pat-
terns. Each of the patterns, as shown in Table 4.4, are segmented and loaded
