A CLAUSE LEARNING ALGORITHM COMBINING IMMUNE MECHANISM TO INVENT PREDICATE

摘要：

Aiming at the larger search space of learning clause in Inductive Logic Programming, we put forward the definitions of the predicate template and the clause template to reduce search space.We suggest IMPI algorithm, which is a genetic algorithm combining immune mechanism and uses the clause template as genetic code to learn the clause template of required clause.IMPI uses immune mechanism to invent new predicates, which can extend the hypothesis language, find better results.Correspondingly we design when to invent predicates.After obtaining the clause template, we use a general method based on generalization and information gain sampling to convert clause template to clause.We design the corresponding fitness function and genetic operator.It indicates that this algorithm can reduce the search space, improve the efficiency of search algorithm and can learn recursion clause by theoretical analysis and experiment comparison.It is an effective clause learning algorithm.

关键词： Inductive logic programming Clause template Immune mechanism Recursion clause Information gain Genetic algorithm

作者: PENG YU DA-YOU LIU

作者单位: College of Computer Science and Technology, Jilin University, Changchun 130012, China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, Jilin Uni

会议类型: 国际会议

会议名称: 2007 International Conference on Machine Learning and Cybernetics(IEEE第六届机器学习与控制论国际会议)

会议地点: 香港

会议语种:英文

页码: 3490-3495

在线出版日期: 2007-08-19（万方平台首次上网日期，不代表论文的发表时间）