Thomas Kipf

Maybe your label processing scheme introduced some issues? You can use multi-label targets by replacing the loss function accordingly. On Fri 29. Mar 2019 at 11:45 lanlin wrote: > Hi...

I’ve trained GCN models with features derived from graph structure (e.g. node degree, or DeepWalk embeddings) before and this generally worked quite well. If your GCN model is performing worse...

The approach we introduce in the Neural Relational Inference paper ( https://arxiv.org/abs/1802.04687) can be used for edge classification. Code is linked in the paper - I’m working on a more...

If `adj` is symmetric (which we assume) then both are identical. On Thu, Jan 2, 2020 at 10:30 AM kentfly wrote: > (A*B)T = BT * AT > so.. "adj.dot(d_mat_inv_sqrt).transpose().dot(d_mat_inv_sqrt)"...

I think this should be easiest to implement in the new DGL library: https://www.dgl.ai On Sun, Jan 20, 2019 at 8:39 AM RookieYin wrote: > So, did you solve these...

You can use either model - R-GCN probably fits that problem quite well out of the box, but GCN also works on directed graphs if you normalize the adjacency matrix...

Can you be a bit more specific? I am not sure if I understand the question. On Thu, May 9, 2019 at 6:35 AM sukannyapurkayastha wrote: > Hi @tkipf ,...

I do not remember seeing such a suggestion. Not sure exactly if I understand what you mean, but if you want such an approach to work you would probably have...

Maybe this paper is helpful? https://arxiv.org/abs/1806.07243 On Wed, Apr 3, 2019 at 4:51 PM hahahahah712 wrote: > hello，thanks for your amazing work！ > > I am doubt that if i...

Yes, this is possible- feel free to implement it! On Mon 27. May 2019 at 03:42 mrano wrote: > @tkipf Is it possible to implement masked_auroc > in metrics.py? Cause...