How far can we get when predicting properties without crystal structures? [working title]

Rhys Goodall
University of Cambridge
Department of Physics

I will introduce an end-to-end stiochiometry to property model for inorganic crystals based on message passing and compare it against more typical descriptor based approaches. If I have managed it before the talk, I will also compare my model against schnet to try and interpret how much worse we are/where the most significant failings are.


Back to Machine Learning for Physics and the Physics of Learning