计算机工程与应用 ›› 2006, Vol. 42 ›› Issue (15): 23-.
• 博士论坛 • 上一篇 下一篇
一种自顶向下和自底向上相结合的ILP算法
许中卫,李炜,宋杰,吴建国
收稿日期:
修回日期:
出版日期:
发布日期:
通讯作者:
ZhongWei Xu,,,
Received:
Revised:
Online:
Published:
Contact:
摘要： 归纳逻辑程序设计（ILP）是机器学习的一个重要分支，给定一个样例集和相关背景知识，ILP研究如何构建与其相一致的逻辑程序，这些逻辑程序由有限一阶子句组成。本文描述一种综合当前一些ILP方法多方面优势的算法ICCR, ICCR溶合了以FOIL为代表的自顶向下搜索策略和以GOLEM为代表的自底向上搜索策略，并能根据需要发明新谓词、学习递归逻辑程序，对比实验表明，对相同的样例及背景知识，ICCR比FOIL和GOLEM能学到精度更高的目标逻辑程序。
关键词: 归纳逻辑程序设计, 自顶向下, 自底向上, 一阶子句
Abstract: Inductive Logic Programming (ILP) is an important branch of machine learning. Given a set of examples and background knowledge, ILP studies how to construct a logic program consistent with these data. The logic program consists of some first-order clauses. This paper describes a new algorithm witch integrated some advantage aspects of current ILP methods. In particular, it combines a bottom-up method similar to FOIL with a top-down method similar to GOLEM. It also includes predicate invention and recursive program learning. Contrasting experiments demonstrate the advantages of this method.
Key words: inductive logic programming, top-down, bottom-up, first-order clause
许中卫,李炜,宋杰,吴建国.
ZhongWei Xu,,,. An ILP algorithm that combines top-down and bottom-up methods[J]. Computer Engineering and Applications, 2006, 42(15): 23-.
0 / 推荐
导出引用管理器 EndNote|Ris|BibTeX
链接本文: http://cea.ceaj.org/CN/
http://cea.ceaj.org/CN/Y2006/V42/I15/23
