论文部分内容阅读
本文提出了一个混合型多概念获取算法ＨＭＣＡＰ，它将符号学习ＣＡＰ算法的主要思想和ＢＰ神经网络有机结合，以状态在实例空间出现的概率为启发信息，以提供的混合实例集为分类依据，并具有增量学习能力．ＨＭＣＡＰ所依据的实例集既可具有离散属性又可有连续属性，并且能根据用户的要求得到不同精度的结合ＢＰ网的二叉多分类判定树．本文还给出ＨＭＣＡＰ的算法应用实例，ＨＭＣＡＰ可用于自动知识获取系统．
This paper presents a hybrid multi-concept acquisition algorithm HMCAP, which combines the main idea of symbol learning CAP algorithm and BP neural network organically, takes the probability of the state in the instance space as the inspiration information and the provided mixed instance set as the classification basis, And with incremental learning ability. HMCAP based on the set of examples can have discrete attributes but also have continuous attributes, and can be based on user requirements with different precision BP neural network with binary multi-class decision tree. This article also gives an example of HMCAP algorithm, HMCAP can be used for automatic knowledge acquisition system.