Professional Documents
Culture Documents
24 AssociativeLearning1
24 AssociativeLearning1
24 AssociativeLearning1
1
• We discuss a collection of simple rules that allow
unsupervised learning.
• These rules give networks the ability to learn
associations between patterns that occur together
frequently. → pattern recognition and recall.
• How associations can be represented by a network?
• How a network can learn new associations?
• An association is any link between a system’s input
and output such that when a pattern A (stimulus) is
presented to the system it will respond with pattern B
(response).
2
Hebb’s Postulate
3
Simple Associative Network
1 stim ulus
p =
0 no stim ulus
1 response
a =
0 no r esponse
6
Unsupervised Hebb Rule
w ij q = w ij q – 1 + a i q p j q
α dictates how many times a stimulus and response must
occur together before an association is made.
Vector Form:
W q = W q – 1 + a q pT q
7
Banana Recognition Example
0
Initial Weights: w = 1 w 0 = 0
Training Sequence: p0 1 = 0 p 1 = 1 p0 2 = 1 p2 = 1
w 1 = w 0 + a 1p 1 = 0 + 0 1 = 0
8
Example
Second Iteration (sight works):
0 0
a 2 = ha r dlim w p 2 + w 1 p 2 – 0.5
= ha r dlim 1 ×1 + 0 ×1 – 0.5 = 1 (banana)
w 2 = w 1 + a 2p 2 = 0 + 1 1 = 1
w 3 = w 2 + a 3p 3 = 1 + 1 1 = 2
10
Hebb Rule with Decay
W q = W q – 1 + a q pT q – W q – 1
W q = 1 – W q – 1 + a q pT q
0 0
a 3 = hardlim w p 3 + w 2 p 3 – 0.5
= hardlim 1 0 + 1 1 – 0.5 = 1 (banana)
w 3 = w 2 + a 3p 3 – 0.1w 3 = 1 + 1 1 – 0.11 = 1.9
w w
13
q q
Problem of Hebb with Decay
• Associations will decay away if stimuli are not
occasionally presented.
If ai = 0, then
w
wij q = 1 – w ij q – 1
wi j q = 0.9 wi j q – 1
q
Therefore the weight decays by 10% at each iteration
14
where there is no stimulus.
Instar (Recognition Network)
We considered associations between scalar inputs and
outputs. Now we examine a neuron that has a vector
input.
15
Instar is similar with ADALINE, perceptron and linear Associator.
Instar Operation
a = hardlim Wp + b = hardlim 1wT p + b
or
1w
Tp = 1w p cos q – b
For normalized vectors, the largest inner product occurs when the
angle between the weight vector and the input vector is zero
the input vector is equal to the weight vector.
1w
w q = w q – 1 + a q p q – a qw q – 1
ij ij i j i ij
or
w ij q = wij q – 1 + a i q p j q – wi j q – 1
Vector Form:
i w q = iw q – 1 + a i q p q – iw q – 1
18
Graphical Representation
For the case where the instar is active (ai = 1):
iw q = iw q – 1 + p q – iw q – 1
or
iw q = 1 – iw q – 1 + p q
nnd15gis 20
Example
0 1 oran ge det ect ed visually
p =
0 oran ge not det ect ed
shape pj=±1→║p║=√3 →
p = tex ture
w eight b=-2 > -║p║2
21
Training
T The network should not respond to any
W 0 = 1w 0 = 0 0 0 combination of fruit measurements, so the
measurement weights will start with values of 0.
1 1
0 0 Assumption: the
p 1 = 0 p 1 = –1
p 2 = 1 p 2 = –1
visual system only
–1 –1 operates correctly on
even time steps.
First Iteration (=1):
0 0
a 1 = hardlim w p 1 + Wp 1 – 2
1
a 1 = h ardlim 3 0 + 0 0 0 – 1 – 2 = 0 ( no response)
– 1
0 1 0 0
1 w 1 = 1 w 0 + a 1 p 1 – 1w 0 = 0 + 0 –1 – 0 = 0
0 –1 0 0 22
Further Training
0 1 0 1
1w 2 = 1 w 1 + a 2 p 2 – 1 w 1 = 0 + 1 – 1 – 0 = – 1
0 – 1 0 –1
1 1 1 1
1 w 3 = 1 w 2 + a 3 p 3 – 1w 2 = –1 + 1 –1 – –1 = –1
–1 –1 –1 –1
nnd15is
● If we define X(q) as the set of all i such that ai(q) =1, this rule
will be as instar rule.
24
Outstar (Recall Network)
• The instar network with a vector input and a
scalar output can perform pattern recognition
by associating a particular vector stimulus with
a response.
• The outstar has a scalar input and a vector
output. It can perform pattern recall by
associating a stimulus with a vector response.
25
The symmetric saturating function is chosen because this
network is used to recall a vector containing values -1 or +1. 26
Outstar Operation
Then, when p = 1
wij q = wi j q – 1 + a i q p j q – p j q w ij q – 1
wi j q = wi j q – 1 + a i q – w ij q – 1 p j q
28
The outstar rule has properties complementary to
the instar rule. Learning occurs whenever pj is
nonzero (instead of ai). When learning occurs,
column wj moves toward the output vector.
Vector Form:
w j q = w j q – 1 + a q – w j q – 1 p j q
29
Example - Pineapple Recall
30
Definitions
0 0
a = satlins W p + Wp
100
0
W = 010
001
shape –1
0
p = tex ture p pi neap ple
= –1
weight 1
w j ( q) = w j ( q – 1) + ( a( q) – wj ( q – 1) ) pj( q )
32
Iteration 1
0 – 1 Assumption:
0 0
p 1 = 0 p 1 = 1
p 2 = –1 p 2 = 1 measured values are
0 1 available only on
even iterations.
=1
0 0 0
a 1 = s atli ns 0 + 0 1 = 0 ( no r esponse )
0 0 0
0 0 0 0
w 1 1 = w1 0 + a 1 – w 1 0 p 1 = 0 + 0 – 0 1 = 0
0 0 0 0
33
Convergence
– 1 0 –1
a 2 = sa tlins – 1 + 0 1 = –1 ( me asurem ent s given)
1 0 1
0 – 1 0 –1
w 1 2 = w1 1 + a 2 – w1 1 p 2 = 0 + – 1 – 0 1 = –1
0 1 0 1
0 – 1 –1
a 3 = s atli ns 0 + – 1 1 = – 1 ( me asurem ent s re calle d)
0 1 1
34
• The network is now able to recall the measurements
of the pineapple when it sees it, even though the
measurements system fails.
• From now on, the weights will no longer change
values unless a pineapple is seen with different
measurements.
– 1 –1 – 1 –1
w 1 3 = w1 2 + a 2 – w1 2 p 2 = – 1 + –1 – – 1 1 = – 1
1 1 1 1
nnd15os
35