mihailbogojeski / ml-dft Goto Github PK

Dear Mihail,

I am Chenru Duan, a graduate student at Prof. Heather Kulik group at MIT. I read your NatComm2020 paper about predicting the DFT/CC energies with electron densities and found it very interesting! Impressive work! It is also very kind of you that you open-sourced the code and dataset to let other people play with it. I am able to reproduce the results using your code and plug it into a workflow that I am building up. But I have a question related:

I am a bit puzzled by the electron density file (dft_densities.npy) that you shared, especially by the fact that the density contains negative values in the 3D tensor and that one cannot see the shape of a water molecule in it. Is there any post-processing step needed from the raw valence electron density to derive the density you uploaded?

Thanks in advance,
Chenru

Hi，we are following your work on Nat. Commun.. The papar is very inspiring as well as this repo framework.

However, I have a problem on the meaning of the v in krr_param_grid. It seems the water example provided in the repo doesn't utilize this part. So could you explain the meaning of this parameter or provide some literatures?

Thank you.