Information Technology Reference
In-Depth Information
2
g
The EKF has been implemented with the assumption that Q
ð
n
Þ
¼
I
n
;
n
r
and r
˃
2
.
The MRAN prediction algorithm Sundararajan et al. (
2002
), Yingwei et al.
(
1998
), with the EKF, here called MRANEKF algorithm, is shown in Fig.
9
and it
is summarized as follow:
(n)=
ð
ð
Þ;
ð
ÞÞ
1. For each observation
x
n
y
n
do: compute the overall network output:
ð
n
ÞÞ
¼ k
0
þ
P
i¼1
k
i
/
y
ð
n
Þ
¼f
ð
x
ð
k
x
ð
n
Þ
c
i
k
Þ
where K is the number of hidden
units;
2. Calculate the parameters required by the growth criterion:
kk
¼
ð
k
ð
Þ
ð
ð
ÞÞ
k
e
n
y
n
f
x
n
s
P
n
2
e
ð
j
Þ
e
rms
ð
Þ
¼
n
M
j¼n
ð
M
1
Þ
d
ð
n
Þ
¼
k
x
ð
n
Þ
c
r
ð
n
Þ
k
Fig. 9 Flow chart of the
MRANEKF algorithm
Search WWH ::
Custom Search