Unsupervised graphsage

I want to find similar implementation of original unsupervised graphsage implementation https://github.com/williamleif/GraphSAGE in DGL.
Is there an existing implementation for this?

There is a built-in GraphSAGE module which implements a single layer of GraphSAGE, plus a complete example that uses this module.

Those are supervised examples. But it doesn’t have any unsupervised examples

Hi,

We have a unsupervised example by link prediction you can find at the later section in https://github.com/dglai/DGL_devday_tutorial/blob/master/BasicTasks_pytorch.ipynb

1 Like

@VoVAllen @ghaffarian
Sorry for not being specific, I’m looking for unsupervised training for heterogenous graphs, basically nodes with different initial features