基于神经网络的社团结构提取算法研究

Research on community structure extraction algorithmbased on neural network

针对复杂网络社团结构提取问题提出了离散Hopfield神经网络(DHNN)算法,并对这种算法的稳定性做了理论和实践上的分析。证明了从任意的初值出发,经过若干次迭代后最终收敛到1个吸引子或1个长度为2的极限环;给出了DHNN的能量函数与模块度函数之间的关系,证明了网络的稳定点对应于一个极大的模块度函数Q值。

Aiming at the problem of complex network community structure extraction,a discrete Hopfield neural network(DHNN)algorithm is proposed,and the stability of this algorithm is analyzed theoretically and practically.The results obtained are:It is proved that starting from an arbitrary initial value,after several iterations,it finally converged to an attractor or a limit cycle of length 2;the relationship between the energy function of DHNN and the modularity function is given.It is proved that the stable point of the network corresponds to a maximum Q value of the modularity function.