This is an impressive piece of work. It’s not a complete breakthrough like AlphaFold was for protein folding, but it’s a major step in a field that has been arguably rather stagnant. There were previous attempts at this (machine-learned density functionals), but none really challenged the existing methods until now. I strongly believe that it’s just the beginning of machine learning for electronic structure calculations.
2 comments
[ 4.8 ms ] story [ 14.4 ms ] thread