CDCN: A New NMF-Based Community Detection Method with Community Structures and Node Attributes

Algorithm 2

The optimization process of CDCN.

Input:network graph G, hyper-parameters β, number of communities K and maximum number of iterations maxIter.

Output:the probability distribution matrix U.

Begin

According to the network graph G and Eq.(1), generate the community structure matrix and randomly initialize the probability distribution matrix

WhileimaxIterdo

Update U⁽ⁱ⁺¹⁾ according to Eq.(21);

Update V⁽ⁱ⁺¹⁾ according to Eq.(22);

End While

End