You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Specifically for the adsorption problem, we could add a feature for whether that atom is part of the adsorbate or catalyst, as well as particular parts of catalyst (e.g. SAC or scaffold, or which layer of MPEA) and preferentially pool those atoms together first (this would obviously also require changes in the ChemistryFeaturization repo)
a more general approach like EigenPooling would also be cool to implement: PyTorch version here (NB: you still need a way to designate subgraphs beforehand with this approach, so perhaps could be combined with the idea above, and/or just use the spectral approach that they adopt in the paper)
there's also this bi-level graph idea (they combined with GAT's though)
The text was updated successfully, but these errors were encountered:
A few ideas here:
The text was updated successfully, but these errors were encountered: