Temporal network embedding using graph attention network

Abstract Graph convolutional network (GCN) has made remarkable progress in learning good representations from graph-structured data. The layer-wise propagation rule of conventional GCN is designed such a way that the feature aggregation at each node depends on features one-hop neighbouring nodes. Adding an attention layer over can allow to provide different importance within various neighbours. These methods capture properties static network, but not well suited temporal patterns time-varying networks. In this work, we propose graph (TempGAN), where aim learn continuous-time by preserving proximity between nodes network. First, perform walk generate positive pointwise mutual information matrix (PPMI) which denote correlation Furthermore, design TempGAN architecture uses both adjacency and PPMI embeddings Finally, conduct link prediction experiments designing autoencoder evaluate quality embedding generated, results are compared with other state-of-the-art methods.