Differentiable Meta Multigraph Search with Partial Message Propagation on Heterogeneous Information Networks

Heterogeneous information networks (HINs) are widely employed for describing real-world data with intricate entities and relationships. To automatically utilize their semantic information, graph neural architecture search has recently been developed various tasks of HINs. Existing works, on the other hand, show weaknesses in instability inflexibility. address these issues, we propose a novel method called Partial Message Meta Multigraph (PMMM) to optimize design Specifically, learn how (GNNs) propagate messages along types edges, PMMM adopts an efficient differentiable framework meaningful meta multigraph, which can capture more flexible complex relations than graph. The typically suffers from performance instability, so further stable algorithm partial message ensure that searched multigraph consistently surpasses manually designed meta-structures, i.e., meta-paths. Extensive experiments six benchmark datasets over two representative tasks, including node classification recommendation, demonstrate effectiveness proposed method. Our approach outperforms state-of-the-art heterogeneous GNNs, finds out multigraphs, is significantly stable. code available at https://github.com/JHL-HUST/PMMM.