JinheonBaek / RGCN

Pytorch implementation of a RGCN Link Prediction Model
229 stars 41 forks source link

Why building test graph requires train_triplets? #8

Closed zhaohan-xi closed 3 years ago

zhaohan-xi commented 3 years ago

Hi Dear Author, in this line, building the test_graph has train triplets as input, isn't it test triplets actually?

JinheonBaek commented 3 years ago

Thank you for your question.

This is a nice question, and we use the test_graph (all train triplets) to get embeddings of entities on the given graph, which are further used to test the model on the test triplets. In other words, the test triplet's entity embedding is obtained from the RGCN outputs of the test_graph (train_triplets).