Skip to content

Latest commit

 

History

History
14 lines (11 loc) · 820 Bytes

ML-for-the-solution-of-the-Schrödin.md

File metadata and controls

14 lines (11 loc) · 820 Bytes

Machine learning for the solution of the Schrödinger equation

Code: N/A Is-Survey: Yes URL: https://iopscience.iop.org/article/10.1088/2632-2153/ab7d30/pdf Year: 2021

Zhou et al [100] [[Toward-the-Exact-Exchange–Correlation-Potential]] machine-learned the exchange-correlation potential for real molecules. They showed that using a convolutional NN that learns the exchange correlation potential from the electron density based on small molecule CCSD data, one can achieve portability to larger molecules. The convolutional NN effectively generated a density-derived description inside itself. More accurate bond lengths and vibrational frequencies were obtained with the NN functional than with B3LYP for small molecules. The NN was also able to learn the weak vdW (van der Waals) interactions from CCSD data. L