I am build an auto-encoder for graph embedding.
This is my autoencoder file: https://github.com/hxtruong/dynamic-graph-embedding/blob/master/src/static_ge.py
And this is class which is using Autoencoder to build a model and train: https://github.com/hxtruong/dynamic-graph-embedding/blob/master/src/static_ge.py
I stuck at loss function that doesn’t decrease. I have tried to different learning rate [0.003, 0.001, 0.0005] which seems not improve model.
Note: I am not sure about layer in auto-encoder and loss function as my code are correct.