Problemstatement: Given observations, what
causal relations can be learnt?
The data here presented is drawn from the well-known Alarm Network (see previous topic). Run de model
construction algorithm step-by-step. For boolean data, the treshold for
KDE is set to 0.03. This is
however still to large to recognize all independencies, because chances
of a Burglary or earthquake are very small. Lower the treshold to about
0.01 for correct learning.