Endocrine-Immune Network and Its Application for Optimization - Intelligent Computing for Sustainable Energy and Environment - page 149

- Difference evaluation: The difference diff between one antibody and the mem-
oryantibodyisusedtomeasurethequalityoftheantibody.Ingeneral,we
can define diff by using Euclidian distance, Haming distance, etc.
- Hormonal regulation:Hormones are ecient bio-active substances, secreted
by endocrine cells and endocrine glands. They have an important role in ef-
fecting physiological function and adjusting the metabolism of tissue cells in
our body [15]. Simulating the behavior of hormone in endocrine system, H ,
exert a bidirectional regulation of immunity to increase the diversity of anti-
body population in EINET. The function of hormones may take two forms,
suppression and activation on antibodies. According to the diff of antibody,
hormone suppression generates low-level elimination probability and muta-
tion probability to avoid the good antibodies being affected; by means of the
high-level elimination probability and mutation probability, hormone acti-
vation can improve the bad antibody to be better.
Here, hormone updating function which determines the amounts of the hor-
mone can be expressed as follow:

H=Fun(diff) (1)
where, H [0 , 1] . In general, H updating function is an increasing function.
Then the H updating function can be defined by

Fun(diff ) [0 , 1] (2)

Fun(diff ) < 0 . 5 , diff < 0 . 5

(3a)

Fun(diff ) > 0 . 5 , diff > 0 . 5 (3b)

where, the constraints ensure that there is a balance point at 0.5 between
the H suppression and H activation operator.
- Antibody learning: For any antibody Ab , learning from the segment or char-
acteristic of memory antibody Ab m is a process of immune network's evolu-
tion, in which evolution itself is a type of learning. Once an antibody whose
anity is higher than the anity of memory antibody is found, that anti-
body will take the place of the original memory antibody as the new memory
antibody after the learning process.
- Enzymatic reaction: Enzymatic reaction is a local search algorithm that is
adopted in order to quickly improve the quality of the population of EINET.
Equivalently, enzymatic reaction is a greedy algorithm implemented to op-
timize the antibodies to relatively good. Since that there is no randomness
in the calculate process of the greedy algorithm, the convergence can be
reached by iteration.

As seen from the model main framework of Fig. 2 and the above description,
there are some differences between our proposed model and the existing AINs.
Unlike opt-aiNet, there is no clonal selection process in the EINET. Instead,
we mainly utilize the hormonal regulation related to the difference between the