clarification on the loss function

[weilaigit]

Hi all,

I will appreciate it if someone could clarify the definition of the loss function in the deepmd-kit paper (https://arxiv.org/pdf/1712.03641.pdf)

1. I am assuming epsilon is related to the cell energy, i.e. not energy per atom?
2. What is N here, the number of atoms in a cell, and i is the atom index?
3. Is it practical to call delta F_{i} as rmse of force as it only refers to one atom, if i is indeed the atom index? Mathematically it is correct but it is awkward?

[njzjz]

Hi,
We've updated the equations in the new DeePMD-kit paper. Although we didn't change anything on the loss function, the definition of these symbols should be more clear in the new paper.