适应噪声环境的解释学习算法

EXPLANATION-BASED LEARNING ALGORITHM FOR NOISY DATA ENVIRONMENT

在现实世界里，ＡＩ系统难免受到噪声的影响．系统有效工作与否取决于它对噪声的敏感性如何．解释学习ＥＢＬ（ｅｘｐｌａｎａｔｉｏｎ－ｂａｓｅｄｌｅａｒｎｉｎｇ）也不例外．本文探讨了在例子受到噪声影响的情况下，解释学习的处理问题，提出了一个算法ＮＲ－ＥＢＬ（ｎｏｉｓｅ－ｒｅｓｉｓｔａｎｔＥＢＬ）．与现有的解释学习方法不同，ＮＲ－ＥＢＬ在训练例子含有噪声时仍然可以学习，以掌握实际的问题分布；和类似的工作不同，ＮＲ－ＥＢＬ指出了正确识别概念对于噪声规律的依赖性，试图从训练例子集合发现和掌握噪声的规律．可以相信，在识别概念时，借助于对噪声规律的认识，ＮＲ－ＥＢＬ可比ＥＢＬ和类似工作有更高的识别率．ＮＲ－ＥＢＬ是解释学习和统计模式识别思想的结合．它把现有的解释学习模型推广到例子含有噪声的情形，原来的ＥＢＬ算法只是它的特例．

in the real world, Al systems arc constantly and adversely influenced bynoisy data. This is also true of EBL (explanation-based learning). This paper discusseshow to cope with noisy data in explanation - based learning and proposes a NR -EBL(noise resistant explanation - based learning) algorithm. Unlike existing algorithms,NR-EBI. can learn macro rules and find the problem distribution when there is noise intraining examples. Also unlike similar work, NR-EBI. reveals the dependency of classifying examples correctly upon the regularities of noise and attempts to detect noise regularities from a set of training examples. With the help of knowledge of noise regularities,NR -EBI. can have a higher rate of correct recognition than traditional algorithms and previous work. NR EBI. is the combination of explanation - based learning and statisticalpattern recognition. Traditional algorithms are only special cases of NR-EBL when thereis no noise in training examples.