论文部分内容阅读
ART-2是一种基于自适应谐振理论的自组织神经网络,广泛应用于模式聚类与识别等方面.本文介绍原始的 ART-2的结构和运算过程,分析它的训练算法,探讨其固有局限性.归纳总结各主要改进 ART-2的背景、目标和实现,评述它们的特征及适应场合.最后指出进一步改进 ART-2的一些思路,在解决具体问题运用各方法的一些参考原则和 ART-2的理论应用价值.
ART-2 is a self-organizing neural network based on adaptive resonance theory, which is widely used in pattern clustering and recognition, etc. This paper introduces the structure and operation of the original ART-2, analyzes its training algorithm, and discusses its inherent The limitations, the background, goal and realization of each major improvement of ART-2 are summarized and their characteristics and adaptation are summarized.Finally some ideas of further improvement ART-2, some reference principles of using each method in solving specific problems and ART -2 theoretical application value.