The Symmetry of the Structure of DHNN with Even Classification
Dong Sheng Ji · Journal of Xiamen University · 2000
The topological structure of the evenly classified DHNN whose synaptic matrix is clipped by a non linear function is studied and geometrized in Euclidean space. In this way, we can see that the visualized structure of the evenly classified network is rotational symmetric. And the rotational symmetry of the geometrized structure graph is coincident with the permutation symmetry of the synaptic matrix of the network, so we can calculate symmetry of the network with the geometrized structure graph.