Concept representation and learning is a basic topic of artificial intelligence. The aim of this paper is to explore the representation issue and the learning issue of abstract concepts. In this paper, we first introduce higher-order fuzzy semantic cell models to represent abstract concepts, based on which we develop a hierarchical representation of concepts called abstract concept graphs. Then, we put forward an unsupervised algorithm to learn a second-order abstract concept graph from a given data set. This method combines the feed-upward mechanism and the self-organizing strategy. In addition, we provide an evaluation metric for this learning algorithm. A series of experiments is provided to demonstrate the feasibility and validity of the proposed method. We also conduct a preliminary exploration of the potential application of this method to image segmentation.

概念表示和学习是人工智能的基本主题。本文的目的是探讨抽象概念的表示问题和学习问题。在本文中，我们首先介绍了代表抽象概念的高阶模糊语义单元模型，然后在此基础上开发了概念的分层表示形式，即抽象概念图。然后，我们提出了一种无监督算法来从给定的数据集中学习二阶抽象概念图。该方法结合了前馈机制和自组织策略。另外，我们为该学习算法提供了一个评估指标。提供了一系列实验来证明该方法的可行性和有效性。