A deep latent space model for interpretable representation learning on directed graphs

Hanxuan Yang, Qingchao Kong, Wenji Mao

Abstract: Highlights•The proposed DLSM combines LSMs and GNNs for interpretable representation learning on directed graphs.•Interpretable variables are generated to capture statistical properties of directed graphs.•A lattice VAE architecture is established to support interactions between variables.•Experiments demonstrate the state-of-the-art performances on downstream tasks.