Adversarial Directed Graph Embedding

Node representation learning for directed graphs is critically important to facilitate many graph mining tasks. To capture the edges between nodes, existing methods mostly learn two embedding vectors each node, source vector and target vector. However, these separately. For node with very low indegree or outdegree, corresponding cannot be effectively learned. In this paper, we propose a novel Directed Graph framework based on Generative Adversarial Network, called DGGAN. The main idea use adversarial mechanisms deploy discriminator generators that jointly node’s vectors. given are trained generate its fake neighbor nodes from same underlying distribution, aims distinguish whether real fake. formulated into unified could mutually reinforce other more robust Extensive experiments show DGGAN consistently significantly outperforms state-of-the-art across multiple tasks graphs.