You may want to read the introduction to Bayes net learning from cases.
Netica can learn from a file of cases, or it can learn from cases one-by-one as you enter them. It can also connect directly with a database, or learn from cases in Excel.
The way that Netica learns relies on the concept of experience.
If you are learning from a world that is constantly changing, and you want the net to adapt, fading may be useful.
After learning from some cases (or perhaps manually constructing a net) you may want to test its performance using another set of cases.
Netica versions 5.0 and later allow for basic structure learning.