Endocrine-Immune Network and Its Application for Optimization - Intelligent Computing for Sustainable Energy and Environment - page 147

as antigen, and the feasible solutions are regarded as candidate antibodies ( Ab ).
Anity aff(.) between the antibody and the antigen is used to evaluate the com-
bination of each candidate antibody with the particular antigen. There will be
found one best antibody with a relatively high anity of each generation that
can be viewed as memory antibody Ab m . In particular, the memory antibody of
each generation will be learned by other general antibodies, and the difference
diff(.) between the general antibody and memory antibody evaluate the antibody
is good or bad. That is to say, an antibody similar to the memory antibody is
a relatively good antibody, and, conversely, a relatively bad antibody is quite
different from memory. Meanwhile, the antibody cluster around the memory
antibody is a set of immune networks . Net =

Therefore,
topology structure of EINET is a similar star network structure which center
on the memory antibody and consider the antibody difference as linkers con-
nected into a network, as shown in Fig.1. Finally, the highest anity memory
antibody( Ab opt )will be the optimal solution.
On the other hand, inspired from the hormonal mechanisms in artificial en-
docrine system[13,14], the process of Hormonal Regulation and Enzymatic Reac-
tion is provided to improve the immune network. In fact, the biological immune
system and the endocrine system are integrated into one single system of in-
formation communication, and they interact and cooperate with each other to
organize an intelligent regulatory network. The model integrated the advantages
of two functional characteristics of hormone, including the cooperation and an-
tagonism among hormone and the enzymatic reaction. The former mechanism
has a bidirectional regulation effects on immune system, which will result in an
activation or suppression on antibodies, that is, help increase diversity of im-
mune network with whole optimization. And the enzymatic reaction is a local
search algorithm that can speed up the evolution of antibody and then speed up
the convergence of the solution. When combined, these two mechanisms would
lead to significant possibilities for improvements of the model.

{ Ab 1 ,Ab 2 ,Ab 3 ...Ab n }

2.2 Details of the General Framework

In the proposed EINET for optimization, five operators, including anity eval-
uation, difference evaluation, hormonal regulation, antibody learning and enzy-
matic reaction, are designed to improve and enhance the adaptability of immune
network and the extreme research target is to upgrade the performance of the
proposed model in complex optimization problem. The framework of EINET is
shown in Fig.2. These operators are repeated until the termination conditions
are satisfied.
The main five operators of the new model is explained in detail as follows:

- Anity evaluation: Calculate anity aff between antigen and antibodies
including memory antibodies. The anity function is given based on the ob-
jective function of an actual optimization problem. In addition, the antibody
with the highest anity is set as memory cells.