Neural networks are not generally regarded as fuzzy, but they have
fuzz -- as do all continuous statistical methods. A typical neural
network contains sigmoid functions which can be viewed as expressing
the fuzzy predicate "this big or bigger". The inputs to these units
are weighted sums, and the units are piped into each other, so a
neural network is essentially a fuzzy controller that evaluates
a fuzzy rule that is in general a bit difficult to understand as
such.
Jive