Information Technology Reference
In-Depth Information
Fig. 9.
Expected Distribution of MACA (n =30) withmultiple attractors (
m
= 1,2,3,4)
A
x
Class 2
y
Class 1
a '
d
c
b
b'
c'
d
'
0
n/2
0
0
HD
HD
Fig. 10.
Distribution of patterns in class 1 and class 2
EO
(
r, m
)=
N
MACA
(
r, m
)
/N
UB
(
r, m
)
(4)
where
N
UB
(
r, m
) shows the expected number of patterns with weight
r
in an
(
n − m
) dimensional vector subspace.
N
MACA
(
r, m
) = the expected number of
patterns with weight
r
in the zero basin of an
MACA
with 2
m
attractors. The
entire equation is specific for a particular
n
. It has been shown [3] that the bias
of the zero basin for low weight (
r
) states (
r<<n
) creeps up due to the three
neighborhood constraints of the
CA
[3].
