Information Technology Reference
In-Depth Information
⎛
1
b
>
r
⎜
⎜
⎜
⎝
i
l
−
i
l
−
b
l
−
b
l
−
b
l
−
b
⎛
⎞
1
1
1
⎛
⎞
P
=
⎟
⎟
⎠
⎞
⎛
⎞
.
∑
∑
(3)
m
,
b
⎜
⎜
⎝
⎜
⎜
⎝
⎟
⎟
⎠
⎜
⎝
⎟
⎠
⎜
⎝
⎟
⎠
=
b
≤
r
i
2
2
l
−
b
i
2
i
=
r
−
b
i
=
r
−
b
In Table 2, some values of
P
,
are given.
b
P
with different
r
and
b
Table 2.
Some values of
,
b
P
,
P
,
l
r
b
l
r
b
b
b
16
14
0
0.0021
32
28
0
9.6506e-006
16
14
2
0.0065
32
28
2
2.9738e-005
16
14
4
0.0193
32
28
4
8.9996e-005
16
14
6
0.0547
32
28
6
0.0003
16
14
8
0.1445
32
28
8
0.0008
16
14
10
0.3438
32
28
10
0.0022
16
14
12
0.6875
32
28
12
0.0059
Given a detector set
R
=
{
d
,
d
,
L
d
}
, and the numbers of their “blank” bits are
1
2
N
R
{
b
,
b
,
L
,
b
}
, the failure probability
P
achieved by these
N
detectors is
R
1
2
R
N
R
=
P
=
(
−
P
)
.
(4)
f
m
,
b
i
i
1
As mentioned in section 2, we assume that all detectors are independent here.
4 Experiments
For convenience, the traditional negative selection algorithm is denoted by t-NSA,
and the heuristic algorithm given in section 3 is denoted by h-NSA.
In this paper, the following experiments are conducted to evaluate the performance
of the heuristic detector generation algorithm proposed in this paper. Every experi-
ments runs 10 times independently.
In section 4.1, experiments are conducted to estimate the average matching number
for generating one detector. In both t-NSA and h-NSA, all candidate detectors are
generated at random and some of them are removed because of matching one or more
self strings. In these two algorithms, the basic operator is the matching operator be-
tween the self string and the candidate detector (or the candidate detector template).
Therefore, the average matching number for generating one detector can reflect their
time costs experimentally.
In section 4.2, comparisons on
for fixed
are done. At the same time, the
N
P
R
f
actual
P
are given.
