Samuel Leventhal

Will you elaborate on your problem setting? If (a) you have a pre-existing graph which you trained on, and a new node is added to that graph, then no new...

If the node is connected to the graph but labeled 'test' using GraphSage's 'test', train, val' scheme, then the node will be inferred after training. For this look at the...

I'm going to assume you mean one-hop neighbors (correct me if I'm wrong) but this may cause issues as the number of layers determines the aggregation. With graphsage one potential...

I misunderstood your question/problem setting. No, you don't need to remove nodes with only 1 neighbor. If the degree of a node is less than the max degree then graphsage...

I have a similar question as well. To preface I would like to use graphsage for inductive learning. Given a trained model with a learned graph embedding and aggregator functions,...

Hi @shivam1702, your input was very helpful, thank you. After implementing your advice I'm now running into some confusion. Currently, I save the trained graphsage model by checkpoints with tensorflow...

Hi @shivam1702, it's good to hear from you and I appreciate the opportunity to discuss this. Currently, I have created a new inference method that can be called on after...

I have had this same problem and would be very curious to hear a resolution. The issue is the state graph and adjacent matrix are initialized to the original run,...

I believe it's the same ordering, in graphsage/utils you can see how features are treated id none are present as well as in training there is: ``` if not features...

@JingLiofLiJing @LAwwp @ningyuwhut I believe the reason the nodes are labeled and minibatched as such is because edges with adjacent nodes labeled as test or val are grouped in a...