Gnn 1.10

Today:Heterogeneous Graphs

  • Goal : so far we only handle graphs with one edge type.

  • Heterogeneous Graphs
    • Relational GCNs
    • Knowledge Graphs
    • Embeddings for KG Completion
  • Block Diagonal Matrices
  • Basis Learning

  • RGCN for Link Prediction

    Knowledge Graph Completion: TransE , TransR ,DistMul,ComplEx

  • Bilinear Modeling DistMult