this is the code for NoiGAN we do some modification on KnowledgeGraphEmbedding the data
-
test how noise affects performance
- different models(TransE, RotatE, DisMult, ComplEx)
- different datasets(FB15k-237 and WN18RR)
- different embedding dims(50, 100, 250, 500, 750, 1000)
-
main goal make generator work
- get a better paper submission
- performance drop is not good enough. especially in wn18rr.
- study of generator:
- what is good noise?