研究了经特定非线性函数剪切后能够均匀分类的ＤＨＮＮ的拓扑结构，并在欧氏空间中对它进行几何化．实验表明，几何化后的拓扑结构图具有旋转对称性．这种对称性与网络连接矩阵的置换对称性是一致的，因此，我们可以借助这种几何结构图来研究网络的对称性及其相关的问题． The topology structure of DHNN which can be uniformly classified after being cut by a specific nonlinear function is studied and it is geometrically transformed in Euclidean space. Experiments show that the geometric topological map has rotational symmetry. This symmetry is consistent with the permutation symmetry of the network connection matrix, so we can use this geometry diagram to study the network symmetry and its related problems.