3 ms·
In my use case I care not one bit about the overall graph structure, but only about the inter-node interactions/relationships. But often times you will get grap
by dauertewigkeit 3y ago
In my use case I care not one bit about the overall graph structure, but only about the inter-node interactions/relationships. But often times you will get graphs where you have 1% type A interactions, 12% type B interactions and 87% type C interactions and that makes training quite challenging.
But I am not sure if you can stay that that is over-fitting the graph structure rather than simply learning the most common interactions in the dataset.
- lmeyerov 3y agothe paper is on my queue, but yeah, these kind of class imbalance & class dependency problems are why things like RGCNs and negative sampling are so popular. It's almost the typical case in cyber, fraud, supply chain, social, etc. As others posted here, ideas like metapaths, transformer equivalence, etc are common building blocks in the GNN world because of these kinds of things. Important areas so def curious.