一种用于知识自动获取的多策略学习方法  

A Multi-Strategy Learning Method in the Knowledge Automatic Acquisition

作　　者：孙杰[1] 吴慧中[1] 

机构地区：[1]南京理工大学信息自动化与制造工程学院

出　　处：《南京理工大学学报》1995年第2期101-104,共4页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金

摘　　要：该文分析了机器学习中几种常用学习策略的特性，提出了一种适用于知识自动获取过程的推广一特化知识表示方法，并给出了一个融合演绎、归纳与类比的多策略学习算法。该算法将学习问题的不确定性和多种学习策略溶进命题集合的描述和变换中，强调学习算法与领域专家在知识获取中的协作关系。因此，它是一种面向领域专家的交互式学习方法。通过在机械设计专家系统ＭＤＥＳⅡ知识获取中的实际应用，表明了该算法对设计类知识获取的有效性。This paper analyzes the characteristics of some common learning strategiesin the machine learning,proposes a generalization specialization knowledge representationmethod suitable for the processes of knowledge automatic acquisition ,and presents a multi- strategy learning algorithm in which the deduction , induction and analogy are integrated. The algorithm incorporates the uncertainty of learning problem and several learning strate gies in the description and transformation of the proposition set,and emphasizes the cooperative relationship between the learning algorithm and domain expert.Therefore,it is adomain expert oriented interactive learning approach.The practical application in theknowledge acquisition subsystem of the mechanical design expert system-MDES Ⅱ showsthat the algorithm is effective for the problem of design knowledge acquisition.

关 键 词：学习机 学习系统 多策略学习 知识自动获取 

分 类 号：TP273.22[自动化与计算机技术—检测技术与自动化装置]