What's the coolest/clearest application of graph neural networks?
-
-
W odpowiedzi do @roydanroy
Where does one begin? But here is one: A successful case of antibiotic designhttps://www.sciencedirect.com/science/article/abs/pii/S0092867420301021 …
4 odpowiedzi 4 podane dalej 55 polubionych -
W odpowiedzi do @_onionesque @roydanroy
Another one is simulating physics, which is particularly compelling e.g. the middle of https://arxiv.org/abs/1806.01261 has some very nice examples.
1 odpowiedź 1 podany dalej 8 polubionych -
W odpowiedzi do @_onionesque @roydanroy
Sorry for the haphazard replies (in my defense just woke up). I really like the use of GNNs for PDE solvers, or in conjunction with PDE solvers e.g. https://arxiv.org/abs/2006.09535 and http://proceedings.mlr.press/v119/de-avila-belbute-peres20a.html … My personal favourite is learning algebraic multigrid https://arxiv.org/abs/2003.05744
1 odpowiedź 0 podanych dalej 3 polubione -
W odpowiedzi do @_onionesque @roydanroy
The link in the first reply had made quite a splash in case you missed it when it came out e.g. https://news.mit.edu/2020/artificial-intelligence-identifies-new-antibiotic-0220 … at the core of this is a message passing network.
1 odpowiedź 0 podanych dalej 1 polubiony -
W odpowiedzi do @_onionesque @roydanroy
Lastly in healthcare in the learning and use of EHR graphs, construction of knowledge graphs, they are starting to have an impact (maybe
@david_sontag has favourites to share, as my familiarity is only through talks and colleagues).2 odpowiedzi 0 podanych dalej 1 polubiony -
EHR Graphs? Do you mean knowledge graphs by NLP to EHR data?
2 odpowiedzi 0 podanych dalej 0 polubionych -
I mean both. But I also mean that EHR data can have complicated graphical structure too e.g. (image).pic.twitter.com/nzUMow6pHQ
1 odpowiedź 0 podanych dalej 1 polubiony -
Oh that’s cool. From what I recall about EHR data there’s huge issues with data silos, poor integration and implementation. But if the limitations were overcome they could open up incredible possibilities.
1 odpowiedź 0 podanych dalej 1 polubiony -
Shubhendu Trivedi podał/a dalej Narges Razavian
@narges_razavian has given a nice example in the same threadhttps://twitter.com/narges_razavian/status/1339038999168622593?s=20 …Shubhendu Trivedi dodał/a,
2 odpowiedzi 0 podanych dalej 2 polubione
Predicting Alzheimer’s; a graph network built upon EHR data that infers labels of AD node by learning graph structures and representations of other nodes. Solving two intrinsic problems for exploiting EHR data - the sparsity and lack of understandable connection among features.
Wydaje się, że ładowanie zajmuje dużo czasu.
Twitter jest przeciążony lub wystąpił chwilowy problem. Spróbuj ponownie lub sprawdź status Twittera, aby uzyskać więcej informacji.
to 
