Databases Reference
In-Depth Information
Fig. 3.6.
Architecture of evolving polynomial network.
values to the power of their respective degrees. The bias value is raised to
power one only. Then they are passed through a multiplier unit represented
by
. Finally the output of all the neurons are summed up and passed
through a linear function to the output unit. Then these estimated values
are compared with the actual target output to generate the mean square
error. This error signal is passed to the Swarm Intelligence module to guide
different modules for appropriate training of the network.
3.4.3.
Parameters of evolving polynomial network
(
EPN
)
In our proposed EPN approach we evolve a set of polynomial equations to
classify the data set. The polynomial equation considered in our approach
can be expressed as:
n
p
x
r
y
=
C
0
+
C
i
(3.6)
i
=1
j
=1
