Deep MinCut: Learning Node Embeddings by Detecting Communities.

• Our node embeddings are both interpretable and competent for classification tasks. • Our graph representation learning process is scalable. • Our interpretable node embeddings outperform baselines. • Our technique is robust to different experimental settings. • Our embedding reveals graph's community structure, especially a hierarchy one. We present Deep MinCut (DMC), an unsupervised approach to learn node embeddings for graph-structured data. It derives node representations based on their membership in communities. As such, the embeddings directly provide insights into the graph structure, so that a separate clustering step is no longer needed. DMC learns both, node embeddings and communities, simultaneously by minimizing the mincut loss , which captures the number of connections between communities. Striving for high scalability, we also propose a training process for DMC based on minibatches. We provide empirical evidence that the communities learned by DMC are meaningful and that the node embeddings are competitive in different node classification benchmarks. [ABSTRACT FROM AUTHOR]