Closed mufeili closed 5 years ago
Hi,
Thank you for the kind words, and your interest in GAT!
The initialisation is implicit---every layer we have used from tf.layers
will utilise Glorot (Xavier) initialisation by default, without us having to do anything.
Thanks, Petar
Got it, thanks a lot! @PetarV-
Hi, thank you for the great work and releasing your code. In the paper you mentioned that the model parameters were initialized with Glorot initialization, but I cannot find it in the code. As I'm not very familiar with TensorFlow, maybe I just missed that part. Could you please elaborate on how did you do this? Thank you!