About semi-supervised learning

[junkang Wu]

Hi, here!
Recently, I'm confused about semi-supervised classification in GCN.

• In other words, how the lose function (1)$\lambda \times L_{reg}$ is implemented? I can't find it in the code. May I ask you about its implementation?

• As it's mentioned in the paper, semi-supervised is focused on some datas which are unlabeled.But all data in cora, pubmed and citeseer are labeled. so it is depend on the sample_mask? Or they are masked while training? Could you explain it for me in detailed or give an simple example?
  Thanks a lot ! Looking forward to your help!