Neuron Models
Aleksander and Morton have greatly simplified the math used to describe the
neuron model
. This may be acceptable, depending on your
purposes, but it makes it harder to judge the limits of where such models may
or may not be applicable. Specifically, they completely ignore the issues
involved in linear approximations. This is the primary problem which
restricts the use of traditional artificial neural net models in many
applications. These issues are covered in some detail, for example, in
Haykin
's thick volume.