Variational Approach for Learning Community Structures

Algorithm 1

Variational Graph Autoencoder for Community Detection (VGAECD).

Input: Features , Adjacency matrix ,

Hyperparameters: learning rate , epochs , size of layer 1 and 2.

Output: Community Assignment Probability and Reconstructed Adjacency matrix

fordo

fordo

Sample

Sample

Obtain reconstructed ⊳  Decoder

Compute loss, ⊳  From (14)

and backpropagate gradients.

end for

end for

Extract community assignment via ⊳  From (18)

Return ,