本文主要描述了一个增量式混合型多概念获取系统ＨＭＣＡＳ．它提出了一个基于概率论的符号学习与神经网络学习相结合的学习算法，能从隶属于某个概念集的实例集中归纳出满足用户精度要求的、以混合型判定树表示的概念描述．在ＨＭＣＡＳ中，符号学习与神经网络学习具有结合紧密和转换灵活等特点，具有较高的学习效率和较强的归纳能力以及增量学习能力．ＨＭＣＡＳ的神经网络学习可选择ＢＰ网络或ＦＴＡＲＴ网络，其推理机制提供了混合型判定树推理 This article describes an incremental hybrid multi-concept acquisition system HMCAS. It proposes a learning algorithm combining symbolic learning and neural network learning based on probability theory, which can summarize the concept description of hybrid decision tree satisfying the users’ precision requirements from the instances belonging to a certain concept set. In HMCAS, symbolic learning and neural network learning are characterized by a combination of tightness and flexible transformation, with high learning efficiency and strong induction and incremental learning ability. HMCAS neural network learning can choose BP network or FTART network, its reasoning mechanism provides a mixed decision tree reasoning