Information Technology Reference
In-Depth Information
Fig. 2 High information content region
operating conditions are sampled according to their probability distribution.
Entropy is given by Eq. ( 1 )
X
c
i ¼ 1
Entropy
ð
S
Þ¼
p i log 2 p i
ð 1 Þ
where, S is training data, c is the number of classes, and pi is the proportion of S
classi
ed as class i. Given that the security boundary generally falls in the lower
probability region of the operating parameter state space, a database containing
samples within the boundary region has the maximum entropy, produced at reduced
computational cost. This is the central principle that is used to devise the ef
cient
training database generation approach proposed in this chapter.
The following section will delineate a technical approach that will be used in this
chapter to devise the ef
cient sampling method to generate the high information
Search WWH ::




Custom Search