SOM的学习规则有三个主要阶段：

1）寻找与输入模式xk最接近的连接权向量Wj*=（wj*1, wj*2,….., wj*N）

2）将该连接权向量Wj*进一步朝向与输入模式xk接近的方向调整

3）除调整连接权向量Wj*外，还调整邻域内的各个连接权向量，并随着学习次数的增加，逐渐缩小邻域范围-SOM learning rules are three main stages : 1) Find the input mode xk closest to the connection weights vector Wj *= (wj * 1, * 2 wj ... .., wj * N) 2) for the right to connect Vector further towards * Wj with input mode xk close reorientation 3) Subject to adjustment weights Vector * Wj, also adjusted neighborhood in all the right connections Vector, and with the increase in the number of learning, gradually narrowing the scope Neighborhood