Databases Reference
In-Depth Information
In the vote counting process, the SI module performs an index-matching
process to compare the index obtained from the test pattern with the indices
of patterns stored for each pattern class. The following pseudocode illustrates
this process:
Algorithm 3
SI Module Voting Scheme
1:
for
i = 1 to MaxTestPatternNo
do
2:
for
j = 1 to MaxSubnetNo
do
3:
for
k = 1 to MaxStoredPatternNo
do
4:
if
i.index ≡ k.index
then
5:
k.vote + +
6:
end if
7:
end for
8:
end for
9:
end for
The complexity of this process can be further analyzed using a Big-O anal-
ysis. We can deduce that the complexity of the vote-counting process is n-
polynomial, where n = 3. Given a vote-counting function f (v
cnt
), its com-
plexity in Big-O notation is as follows:
n
3
f (v
cnt
) = O
(7.3)
Where n represents a single executable instruction in the function. After
the numbers of votes are counted, the SI module performs a search function
to identify the pattern class that has the highest votes for the tested pattern.
This function will execute a linear search to find the maximum number of
votes.
7.2.2.2
Voting Scheme at the Coordinator Node
The voting scheme at the coordinator node is used to select the best accu-
racy parameter of a feature from a collection of available features that have
been used in earlier recognition schemes implemented on multiple DHGN net-
works. Each SI module will communicate the results of the recognition of
features as patterns to the coordinator node for further analysis. The coor-
dinator stores all of the accuracy parameters received from the SI modules.
Table 7.1 shows a sample of errors obtained from two SI module nodes for
each feature, on five different pattern classes.
Based on the values obtained from Table 7.1, we can conclude that Feature
1 is the best feature to represent pattern classes 1, 3, and 4 because its errors
are less than the errors of Feature 2. Pattern classes 2 and 5 are likely to
be represented by Feature 2. The voting function in the coordinator node
