机器学习中的概念聚类

Conceptual Clustering in Machine Learning

一、引言机器学习是人工智能发展中一个十分活跃的领域,机器学习的研究就是希望计算机能够像人类那样具有从现实世界获取知识的能力。学习应是一切智能系统的重要特征之一,没有学习能力的系统都不堪称为智能系统。因此不少学者认为,如果说八十年代是专家系统的年代,那么九十年代将是机器学习的年代。归纳学习是机器学习研究中最为困难,然而却最为诱人的一个方面。从科学哲学的观点来看,归纳能够帮助人们学习。

Conceptual Clustering is an important method of machine learning,especially learning by observation.In this paper,we overtiew various views of conceptual clustering,mainly including extended numerical taxonomy,concept learning and concept discovery based on conceptual clustering,discussing a number of issues in conceptual clustering with several famous machine learning systems such as DIS- CON,CLUSTER,MITT,INDUCE,AQ 15,UNIMEM,GLAUBER,IPP,and so on, analysising the differences between conceptual clustering and learning from exam- ples,addressing two processes of conceptual clustering:concept sorting and concept description.At last,we briefly consider the levels of conceptual clust- ering algorithms,and summarize various search technologies used in conceptual clustering systems.