An ILP Algorithm Without Restriction of Constant Ordering
Zhang Run-q · 1999
In this paper, the shortcomings in theory and limitation in applications of FOIL (first-order induc-tive learner) are analyzed. To overcome these difficulties, instance graph H(R,E) and instance order are intro-duced to clarify the relationship between the set R of recursive rules and the instance space E. Based on these concepts, a new ILP (inductive logic programming) algorithm, FOILPlus, is put forward, which prevents the generation of harmful recursive rules by utilizing hung example and hung arc to hold Instance Graph. The algo-rithm can complete learning tasks without the restriction of constant ordering, and does not substantially raise the computational complexity compared with FOIL. FOILPlus has been implemented, and experiments show that it does complete two learning tasks which FOIL fails.