We don't just use #NeuralNetworks in a naive way. We get low-cost features using semi-empirical electronic structure calculations: uses domain knowledge and also uses known symmetries in atomic orbitals. We show zero-shot generalization to much larger molecules than training data
-
-
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Very interesting. Does this speedup include data synthesis and training time?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.