4 ms·
What's the current killer app of GNNs? By that I mean, in which tasks is using a GNN-based approach obviously better? Conceptually, just by framing the graph a
by vslira 5y ago
What's the current killer app of GNNs? By that I mean, in which tasks is using a GNN-based approach obviously better?
Conceptually, just by framing the graph as a mxm matrix I assume the most obvious comparison would be against some sort of dimensionality reduction for node embedding. Should I see graph(node) embedding as an alternative to e.g. PCA, LDA, t-SNE and UMAP and variants?
- entee 5y agoIt’s pretty clearly the current best way to do deep learning on molecules in chemistry. Transformers do really well in predicting reaction outcomes but GCNNs are still better when it comes to “is this a good molecule according to this label” type questions. That said, we have added attention layers to GCNNs so I think the lines get blurry.
- FL33TW00D 5y agoThe Stanford Open Graph Benchmark shows the current best performers on some different tasks: https://ogb.stanford.edu/docs/leader_graphprop/ https://ogb.stanford.edu/docs/leader_graphprop/
- rmac 5y agowhen your input makes more sense as a graph, or your desired output resembles predicting node labels or edges?
- lysecret 5y agoFraud detection as well as Facebook haha
- Voloskaya 5y agoAlphafold. And generally any problem where you care about symmetries in space, like 3D point cloud representation.