Information Technology Reference
In-Depth Information
8
If age of
d
>
t
h en detector is old
9
Replace
d
by a new random detector
10
else
11
Increase age of
d
12
d
←
d
+
η
.
dir
13
endIf
14
else
←
15
age of
d
0
∑
∑
( (
dd d
dd d
)
d
d
∈
Detectors
16
dir
( (
)
d
d
∈
Detectors
17
η
.
dir
18
endIf
19
endFor
20
endWhile
d
=
d
+
4.6.1
Detector Generation Using Evolutionary Algorithms
Dasgupta and Gonzalez (2002) used a GA to evolve a set of rules (detectors) to cover
the nonself space (Figure 4.7). h e self-space consisted of a set
S
, a subset of [0, 1]
n
;
accordingly, a data point was represented as a feature vector
x
(
x
1
, …,
x
n
) in [0, 1]
n
.
A detection rule was considered to be good if it did not cover any positive
sample (the self ) and it covered a large area of the nonself space. A detector was then
represented as a “detector rule” in the form
=
=
R
i
:
if
cond
i
then
nonself, for
i
1, …,
m
where cond
i
=
X
1
in [low
1
, high
i
1
] and … and
X
n
in [low
i
n
, high
i
n
],
Self-data
Choose
Replace closest
Generate
parent if fitness
initial
two parents
Figure 4.7
Generation of negative detection rules using a GA.
Search WWH ::
Custom Search