LFIO:基于输入输出的Datalog归纳学习系统

LFIO:An Inductive Learning System Based on Input and Output

归纳学习是一种重要的知识发现手段。本文介绍一种新型的基于Datalog程序的归纳学习系统LFIO,在此归纳学习系统中,对于给定的背景知识B和例子集E,其中每个例子都是一对输入输出的(I,O)原子集对子,其中I表示输入,O表示对应的输出,LFIO归纳学习出一个Datalog程序P,对于E中的每个例子(I,O),O都是B∪P∪I的最小(Herbrand)模型。本文详细描述了LFIO的系统设计及实现,初步的实验结果展示了该归纳学习系统的有效性。

An inductive learning system for Datalog programs LFIO was introduced. In this inductive learning system,a background theory is a set of rules,and an example is a pair,where and are sets of ground atoms,standing for input and output respectively. LFIO computes a Datalog program P for a given background theory B and a set E of examples such that B is the least（ Herbrand） model of for every example. This paper describes the design and the implementation of the system. Experiment results show the effectiveness of this inductive learning system.