Information Technology Reference
In-Depth Information
Fig. 10.3
Decomposition of an associative memory with stochastic weights into a superposition
of weight matrices, each one having its own attractors
8
<
9
=
0
@
1
A
0
1
12
a
A
i
C
1
0
A
i
l
12
1
13
13
W
3
D
@
A
;
12
0
A
i
C
1
a
A
i
12
23
:
23
;
13
a
A
i
C
1
1
13
23
a
A
i
0
23
8
<
9
=
0
@
1
A
0
1
a
A
i
C
1
13
0
A
i
12
l
12
1
13
W
4
D
@
A
;
0
A
i
C
1
23
a
A
i
12
12
1
23
:
;
a
A
i
C
1
13
a
A
i
C
1
23
1
13
1
23
0
8
<
0
1
9
=
0
1
0
A
i
C
1
12
a
A
i
l
1
12
13
13
@
A
W
5
D
@
A
;
a
A
i
C
1
12
0
A
i
23
1
12
23
:
;
a
A
i
13
a
A
i
13
23
23
0
8
<
0
1
9
=
0
1
0
A
i
C
1
12
a
A
i
13
l
1
12
13
@
A
W
6
D
@
A
;
a
A
i
C
1
12
0
A
i
23
1
12
1
23
:
;
a
A
i
C
1
23
a
A
i
13
13
1
23
0
8
<
0
1
9
=
0
1
12
a
A
i
C
1
0
A
i
l
1
12
1
13
13
@
A
W
7
D
@
A
;
a
A
i
C
1
12
0
A
i
23
1
12
23
:
;
a
A
i
C
1
13
a
A
i
1
13
23
23
0
8
<
9
=
0
@
1
A
0
1
0
A
i
C
1
12
a
A
i
C
1
l
1
12
1
13
13
W
8
D
@
A
;
a
A
i
C
1
12
0
A
i
C
1
23
1
12
1
23
:
;
a
A
i
C
1
13
a
A
i
C
1
1
13
1
23
0
23
