Closed junkangwu closed 3 years ago
Hi @Wjk666 ,
Observe that M_r is a diagonal matrix in DistMult which makes both the forms similar. Typically, I think the authors observed that having a matrix as a relation embedding can be highly overparameterized, hence the diagonalized version was introduced. This is a good survey to read about the common KG embedding methods which we followed.
According to the model in
distmult
paper, $$score = h_s M_r h_o^T$$. However, in the most papers about knowledge graph completion including the model in CompGCN, they are realized as a form of $$score = h_s r h_o^T$$. May I ask you about the insight behind your realization form?