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

Algorithm 1

The learning process: CDCN.

Input:network graph G, node attributes matrix T, hyper-parameters α and β, number of communities K and maximum number of iterations maxIter.

Output:the probability distribution matrix U.

Begin

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

WhileimaxIterdo

;

);

;

End While.

End