
128 Optical Fiber Communication Systems with MATLAB
®
and Simulink
®
Models
where
p
i
(x|θ
i
) represents the pdf of each distribution in the mixture and each pdf has a weight
w
i
such that w
i
i
=
=
1
1
. This weight indicates the probability of each pdf. When adopted
f
or
o
ptical
c
ommunications,
t
he
E
M
a
lgorithm
i
s
im
plemented
w
ith
a m
ixture
o
f
m
ultiple
G
aussian
d
istributions.
A c
ritical
s
tage
i
n
t
he
M
GD
m
ethod,
w
hich
a
ffects
t
he
a
ccuracy
o
f
B
ER
c
alculations,
i
s
t
o
e
stimate
t
he
n
umber
o
f
G
aussian
d
istributions
f
or
u
se
i
n
t
he
m
ixture.
4.7.5.1.1
Se
lection of Number of Gaussian Distributions in MGD Method
The
n
umber
o
f
G
aussian
d
istributions
t
o
b
e
u
sed
i
s