Information Technology Reference
In-Depth Information
Cooke's model. Also, the concept of a network a
nity threshold (NAT) is intro-
duced as a mechanism to control the density of the connections among antibodies
of the IN; therefore, only the strongest connections between them are considered.
In their model, a
nity measure values are normalized to be in [0, 1]; then, if the
a
nity measure is less than the NAT, the original similarity measure is considered;
otherwise, the lowest similarity value between two antibodies (i.e., 1) is assigned.
h us, a stimulation level (
sl
) is defi ned as
N
N
n
∑
∑
∑
sl x
() 1
(1
m x
(
,
x
)
m x
(
,
x
)
m x
(
,
y
)
(5.15)
i
i
j
i
j
i
j
j
1
j
1
j
1
>
=
+
=
if
sl
(
x
i
)
x
i
.
h e fi rst two terms in Equation 5.15 represent the interaction of a type
i
anti-
body with other antibodies. h e fi rst and second terms represent the inhibitory and
excitatory signals from other antibodies, respectively. h e last term represents the
eff ect of antibody stimulation by foreign antigens. Here the NAT value is dynami-
cally adjusted.
h e AINE model was later modifi ed to alleviate some di
culties such as popu-
lation control and the calculation of the NAT. h e modifi ed model was called
resource-limited AIN (RAIN), and the concept of artifi cial recognition ball (ARB),
which is a representation of a number of identical B cells instead of a single B cell,
was introduced. In this model, there is a resource pool (B cells) with centralized
control and the ARBs compete for allocating such resources. Unlike AINE, in
RAIN, those ARBs having zero resources are removed from the network, and the
NAT is time-independent and derived from antigen dataset. h e RAIN algorithm
is presented in Figure 5.6.
θ
, then
x
i
x
i
k
(
sl
), else
x
i
Self-stabilizing artifi cial immune system
(
SSAIS
). Neal (2002) proposed SAIS, which
is based on RAIN for continuous analysis of time-varying data. Also, SSAIS
does not consider B cell suppression while calculating the stimulation level.
Meta-stable memory IN
. Neal (2003) proposed a modifi ed version of SSAIS for
data analysis, clustering, and immune memory. In this model, each ARB
stimulation is done by foreign antigens and those neighbors, which are in a
Euclidean space. Here, the cloning process is employed only during the pri-
mary response, which is mediated by the NAT, but it does not consider muta-
tion operator. In this model, ARBs having resources less than the defi ned
mortality threshold are removed from the network.
5.2.2.2 Fractal Immune Network
h is version (Bentley and Timmis, 2004) uses the concept of ARB and coined a
new term “fractal recognition space” (FRS). Here, interactions among self-elements
Search WWH ::
Custom Search