Linear regression training energy

[lammps-vasp]

DP developer, I want to know if demd can be used to train energy alone, which is similar to machine learning. Now I use energy alone for training, and the effect is too poor. As the developer said, [] is equivalent to linear regression. Is there a problem in parameter adjustment, or can't I train with energy alone? Thank you for your advice

Looking forward to your reply!

[AnguseZhang]

What's the meaning of training energy alone?

[lammps-vasp]

Observe whether the energy can be trained well. The energy data is from MOLPRO

[AnguseZhang]

Do you have force data? Training with energy alone may cause overfitting problem.

[lammps-vasp]

There is no force data, there is only energy, but there is no over fitting, but the result is very poor, not on the line y = x

[AnguseZhang]

Then you should check the quality of training data, including first-principles convergence, data conversion, e.t.c. and traning parameters.