Rex Ying

That would work as well!

1. Networkx has a very easy-to-use interface for constructing graphs from data. You could certainly do it relatively fast. Also note that since networkx (naive python implementation in general) is...

Sorry for the late response. Suppose that (v1, v2, ... ) are your minibatch of nodes. You need to find all nodes in the neighborhood of each of (v1, v2,...

Hi, i think for dict order doesn't matter. as long as you map the feature/index that corresponds to id. The updated networkx should be able to handle non-int keys.

You can also find the datasets at http://snap.stanford.edu/graphsage/ In the dataset section, we listed the Reddit and PPI datasets.

I think the PPI dataset does not have time information. These interactions are biochemical effects of proteins on other proteins common to many animals of the same species.

Does that mean you obtain higher F1 compared to what we reported in the paper? We only got ~0.6. I'm not too sure what's going on. Potentially due to different...

Hi I can't really tell if your snippet makes sense, since I don't know the context, but tf.get_variable seems to be the more common way of retrieving the variables imo....

when you compute messages for each edge from a neighbor in the neighborhood to the center node, you can concat the edge weights / attributes with the node features.

Identity features can be used. The experiment is done in Kipf & Welling's paper. They show performance comparison over baselines such as Planetoid and DeepWalk. We did not want to...