3

1
Ejemplo de uso
3
2
Vectores prototipo
p
shape
texture
weight
=
p
2
1
1
1
=
Prototype Banana Prototype Apple
Shape: {1 : round ; -1 : eliptical}
Texture: {1 : smooth ; -1 : rough}
Weight: {1 : > 1 lb. ; -1 : < 1 lb.}
Measurement
Vector
p
1
1
1
1
=
3
3
Perceptron
3
4
Caso de dos entradas
a hardlims n ( ) hardlims
1 2
p 2 ( ) + ( ) = =
w
1 1 ,
1 = w
1 2 ,
2 =
Wp b + 0 =
1 2
p 2 ( ) + 0 =
Limite de desicin
3
5
a hardlims
w
1 1 ,
w
1 2 ,
w
1 3 ,
p
1
p
2
p
3
b +
\ .
|
|
|
|
| |
=
El limite de decision deberia de
separar los vectores prototipo.
p
1
0 =
1 0 0
p
1
p
2
p
3
0 + 0 =
The weight vector should be
orthogonal to the decision
boundary, and should point in the
direction of the vector which
should produce an output of 1.
The bias determines the position
of the boundary
3
6
Probando la red
a har dlims
1 0 0
1
1
1
0 +
\ .
|
|
|
| |
1 banana ( ) = =
Platano:
Manzana:
a hardlims
1 0 0
1
1
1
0 +
\ .
|
|
|
| |
1 apple ( ) = =
Platano :
a har dlims
1 0 0
1
1
1
0 +
\ .
|
|
|
| |
1 banana ( ) = =
3
7
Red de Hamming
3
8
Feedforward Layer
For Banana/Apple Recognition
W
1
p
1
T
p
2
T
1 1 1
1 1 1
= =
b
1
R
R
3
3
= =
a
1
W
1
p b
1
+
p
1
T
p
2
T
p
3
3
+
p
1
T
p 3 +
p
2
T
p 3 +
= = =
S 2 =
3
9
Recurrent Layer
W
2
1 c
c 1
= c
1
S 1
------------ <
a
2
t 1 + ( ) posl in
1 c
c 1
a
2
t ( )
\ .
|
| |
posl in
a
1
2
t ( ) ca
2
2
t ( )
a
2
2
t ( ) ca
1
2
t ( )
\ .
|
|
|
| |
= =
3
10
Hamming Operation
p
1
1
1
=
Input (Rough Banana)
a
1
1 1 1
1 1 1
1
1
1
3
3
+
1 3 + ( )
1 3 + ( )
4
2
= = =
First Layer
3
11
Hamming Operation
a
2
1 ( ) posl in W
2
a
2
0 ( ) ( )
posl in
1 0. 5
0. 5 1
4
2 \ .
|
| |
posli n
3
0 \ .
|
| |
3
0
=

= =
a
2
2 ( ) posl in W
2
a
2
1 ( ) ( )
posl in
1 0. 5
0. 5 1
3
0 \ .
|
| |
posl in
3
1. 5 \ .
|
| |
3
0
=

= =
Second Layer
3
12
Hopfield Network
3
13
Apple/Banana Problem
W
1.2 0 0
0 0.2 0
0 0 0.2
b ,
0
0.9
0.9
= =
a
1
t 1 + ( ) sa tlins 1. 2a
1
t ( ) ( ) =
a
2
t 1 + ( ) satlins 0. 2a
2
t ( ) 0. 9 + ( ) =
a
3
t 1 + ( ) satlins 0. 2a
3
t ( ) 0. 9 ( ) =
a 0 ( )
1
1
1
= a 1 ( )
1
0.7
1
= a 2 ( )
1
1
1
= a 3 ( )
1
1
1
=
Test: Rough Banana
(Banana)
3
14
Summary
Perceptron
Feedforward Network
Linear Decision Boundary
One Neuron for Each Decision
Hamming Network
Competitive Network
First Layer Pattern Matching (Inner Product)
Second Layer Competition (Winner-Take-All)
# Neurons = # Prototype Patterns
Hopfield Network
Dynamic Associative Memory Network
Network Output Converges to a Prototype Pattern
# Neurons = # Elements in each Prototype Pattern