
Artificial Neural Networks ◾ 183
represents the data in a distributed fashion since each cell contains infor-
mation which is aected by all the data samples. Note that the learning
rule (4.10) ensures that the weights between neurons that typically have
the same value will be high.
is learning procedure is called Hebb’s rule, named aer the psy-
chologist Donald Hebb who postulated that in the brain the connection
strength between cells is correlated with the frequency of them being
active together. is idea is oen summarized by the slogan “Neurons that
re together wire together.” Obviously, the simple learning rule in (4.1