reorder the update code

TrustAGI-Lab / ARGA

This is a TensorFlow implementation of the Adversarially Regularized Graph Autoencoder(ARGA) model as described in our paper: Pan, S., Hu, R., Long, G., Jiang, J., Yao, L., & Zhang, C. (2018). Adversarially Regularized Graph Autoencoder for Graph Embedding, [https://www.ijcai.org/proceedings/2018/0362.pdf].

MIT License

185 stars 58 forks source link

reorder the update code #3

Open dawnranger opened 5 years ago

dawnranger commented 5 years ago

I think the code is somehow disordered for two reasons:

set dropout=0 right after dropout=FLAGS.dropout is meaningless
the updated emb should be returned rather than the untrained one