自组织神经网络概念和原理，并重点介绍一下自组织特征映射SOM网络。SOM和现在流行的ANN（MLP）模型在结构上类似，都由非常简单的神经元结构组成，但是SOM是一类“无监督学习”模型，一般的用法是将高维的input数据在低维的空间表示[1]，因此SOM天然是一种降维方法。除了降维，SOM还可以用于数据可视化，以及聚类等应用中。(The concept and principle of self-organizing neural network are introduced, and the self organizing feature mapping SOM network is introduced. SOM and the now popular ANN (MLP) model is similar in structure, the neuron structure very simple, but SOM is a kind of "unsupervised learning" model, the general is the use of input data with high dimension in the low dimensional space for [1], so SOM is a kind of natural dimensionality reduction method. In addition to dimensionality reduction, SOM can also be used in data visualization, clustering and other applications.)