The first number (PA) represents predictive
accuracy (in %) on the test set after training. The
second number (TS) is the number of instances used to
train the system. The third number (NS) is the size of
the final network in nodes. The last column reports
predictive accuracy and network size when precepts
are added before training for five of the datasets.
Currently, the precepts are obtained from domain
knowledge provided with the application or generated
from the authors commonsense. They are detailed in
Appendix B. Other, more automatic ways of
generating precepts are the topic of future research