11/24/03
12
Goal of Perceptron Learning
¥
Suppose we have training patterns
x
1
,
x
2
,
É,
x
P
with corresponding desired outputs
y
1
,
y
2
, É,
y
P
¥
where
x
p
ë
{0, 1}
n
,
y
p
ë
{0, 1}
¥
We want to find
w
,
q
such that
y
p
=
Q
(
w
×
x
p
Ð
q
) for
p
= 1,
É
,
P