generalizations. In this case, the bias can be thought
of as a general rule or precept resulting from some
instantiation of domain knowledge or commonsense.
It thus "suggests something advisory and not
obligatory communicated typically through teaching"
(Webster's Dictionary). For instance, when dealing
with how weather conditions affect what one wears, a
possible precept is: "if it rains, take an umbrella,"
regardless of the temperature, the day of the week, or
any other unnecessary (don't-care) detail. It is our
contention that the encoding and use of such precepts
improves an inductive learning system's overall
efficiency. In particular, it increases learning speed by
pruning and constraining the search in the input
space, reduces memory requirements, and improves
overall predictive accuracy. The precept-driven
learning algorithm (PDLA) discussed in [7] is a proof
of concept for the above contention.