Skip to content

Knowledge Graphs

Knowledge graph embeddings, relational/entity classification, and relational databases.

All categories

  • A GNN made heterogeneous, on a movie graph

    PyG's RelBench example builds a two-layer GraphSAGE and converts it with to_hetero into a model that has separate weights for every node and edge type of a relational database.

    IMDB · HeteroConv · SAGEConv

    Read Tutorial    

  • Entity classification with R-GCN (AIFB)

    Classify the entities of a knowledge graph.

    AIFB (Entities) · FastRGCNConv

    Read Tutorial    

  • Entity classification with relational graph attention (AIFB)

    Classify the entities of a knowledge graph with relational graph attention (Busbridge et al., 2019): attention weights over neighbors that also depend on the relation type of each edge.

    AIFB (Entities) · RGATConv

    Read Tutorial    

  • Knowledge graph completion with R-GCN and DistMult

    Predict missing facts of a knowledge graph.

    WordNet18RR · RGCNConv

    Read Tutorial    

  • Knowledge graph embeddings: TransE, DistMult, ComplEx and RotatE

    Learn an embedding for every entity and relation of a knowledge graph so that true facts score higher than false ones.

    WordNet18RR · ComplEx · DistMult

    Read Tutorial    

  • Relational deep learning with a heterogeneous GraphSAGE

    Relational deep learning (Fey et al., 2024) treats a relational database as a heterogeneous graph: every table row is a node and every foreign-key link an edge.

    DBLP · HeteroConv · SAGEConv

    Read Tutorial