There are a ton of interesting graph neural network architectures that have been developed over the last 5-6 years, they are getting pretty good at modeling the

There are a ton of interesting graph neural network architectures that have been developed over the last 5-6 years, they are getting pretty good at modeling the properties of molecules and crystals Edit: saw your other comment on GNNs. I don’t think it’s like a solved problem or anything, but I also don’t think it’s intractable. e.g. https://arxiv.org/abs/2012.00094 https://arxiv.org/abs/2012.00094 seems relevant