Both the multiplicative and z-axis normalization types allow a group of
outputs to be adjusted so that the "vector length" is 1. Both go about it
in different ways. Certain types of neural networks require a vector length
The multiplicative normalization is more simple than Z-Axis normalization.
Almost always Z=Axis normalization is a better choice. However,
multiplicative can perform better than Z-Axis when all of the values
are near zero most of the time. This can cause the "synthetic value"
that z-axis uses to dominate and skew the answer.
Multiplicative normalization works by calculating the vector length of
the input fields and dividing each by that value. This also presents
a problem, as the magnitude of the original fields is not used. For
example, multiplicative normalization would not distinguish between
(-2,1,3) and (-10,5,15). Both would result in the same output.