Community Detection Based on Self-Organizing Map Clustering of Graph’s Embeddings
This is a community detection model (VGASOM), that employs a variational graph autoencoder neural network with two graph convolutional layers to encode the high-dimensional graph features into a lower-dimensional embedding, which is then clustered using self-organizing maps. Community detection is an unsupervised learning task that has quite a number of challenges to be dealt with. Our experiments show that our approach achieves better performances than the baseline methods and shows great potential for the community detection task.
Our work is based on https://github.com/DaehanKim/vgae_pytorch and https://github.com/JustGlowing/minisom