map the output of model back to Euclidean space

HazyResearch / hgcn

Hyperbolic Graph Convolutional Networks in PyTorch.

599 stars 107 forks source link

map the output of model back to Euclidean space #32

Open mumu-mu opened 3 years ago

mumu-mu commented 3 years ago

Hi!

I had seen your 'HGCN' on github. I'd like to reimplement it, but I'm confused that why you map the output of model back to Euclidean space when computing loss. Could you please explain it?

Thanks!