sparse-settings

This paper introduces PGRE, a probabilistic model for dynamic knowledge graphs that captures inter-relational dependencies using Poisson-Gamma and Markov processes, achieving competitive link prediction performance especially in sparse settings.