It is proved that if the cost function is not zero after the algorithm converges, then all hidden units will move to the weighted cluster centers of sample inputs.

• 主要结论包括：如果算法收敛后损失函数不为零，则各隐节点将位于样本输入的加权聚类中心；