Pattern based inductive logic programming
Enrico Pontelli, Chongbing Liu · 2008
A new efficient and effective learning formalism is developed by making explicit and direct use of the concept instances and their patterns. In this formalism, a theory about the target concept is constructed iteratively by first constructing the patterns for all the positive and negative examples using the background knowledge, and then reducing each positive patterns. A system based on this formalism, called PIE, is developed. It is shown to be more efficient and effective than the state of the art ILP system, ALEPH, over a selected set of data sets.