Practical Neural Network Recipies in C++

Probabilistic

Neural Networks

221

When to Use a Probabilistic Neural Network

This model is fairly specialized, not having the wide applicability of

the multiple-layer feedforward network. However, in some situations,

it is ideal.

First, this model is intrinsically a classifier. Modifications exist

that allow it to interpolate between decisions, making it a more

universal function approximator, but these tend to be less than

elegant. Also, as described above, we can force it into an auto-

associative mode. However, this is often inelegant. Thus, in general,

we should primarily consider this model for classification problems.

The principal advantag ...