An Efficient Multiple Predicate Learner  

作　　者：张晓龙 MasayukiNumao 

机构地区：[1]DepartmentofComputerScience,TokyoInstituteofTechnology,Japan [2]DepartmentofComputerScience,TokyoInstitu

出　　处：《Journal of Computer Science & Technology》1998年第3期268-278,共11页计算机科学技术学报（英文版）

摘　　要：In this papers we examine the issue of learning multiple predicates from given training examples. A proposed MPL-CORE algorithm efficiently induces Horn clauses from examples and background knowledge by employing a single predicate learning module CORE. A fast failure mechanism is also proposed which contributes learning efficiency and learnability to the algorithm. MPL-CORE employs background knowledge that can be represented in intensional (Horn clauses) or extensional (ground atoms) forms during its learning process. With the fast failure mechanism, MPL-CORE outperforms previous multiple predicate learning systems in both the computational complexity and learnability.In this papers we examine the issue of learning multiple predicates from given training examples. A proposed MPL-CORE algorithm efficiently induces Horn clauses from examples and background knowledge by employing a single predicate learning module CORE. A fast failure mechanism is also proposed which contributes learning efficiency and learnability to the algorithm. MPL-CORE employs background knowledge that can be represented in intensional (Horn clauses) or extensional (ground atoms) forms during its learning process. With the fast failure mechanism, MPL-CORE outperforms previous multiple predicate learning systems in both the computational complexity and learnability.

关 键 词：Machine learning inductive logic programming multiple predicate learning shift of bias 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]