PAC学习模型研究  

Research of PAC Learning Model

作　　者：何志国[1] 

机构地区：[1]攀枝花学院电信系

出　　处：《微机发展》2004年第8期52-54,70,共4页Microcomputer Development

摘　　要：计算学习理论为比较两算法的性能提供了形式化的框架,并能确定某概念类的计算复杂度和样本复杂度。而PAC学习模型是计算学习理论的基础,它为研究学习及泛化问题提供了一种基本的概率框架。先介绍了基本的PAC学习模型并对其进行了深入的分析,给出了判断一概念类是否是PAC学习的方法;然后针对基本PAC学习模型的不足进行了相应的扩展;最后介绍了计算学习理论中的一些其它模型。Computational learning theory provides formal framework for comparing the performance of two algorithms,and can determine the sample complexity and the computational complexity of some concept class.PAC learning model is the fundamental of computational learning theory,it provides a probabilistic framework for the study of learning and generalization.First,the basic PAC learning model is introduced and analyzed deeply,and a method is given for deciding whether the concept class is PAC learning or not; then extends it for its fault;lastly introduces some other learning models in computational learning theory.

关 键 词：PAC学习 VC维 样本复杂度 计算复杂度 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]