Would angluar shifts from 0 to +pi be sufficient?

isayev / ASE_ANI

ANI-1 neural net potential with python interface (ASE)

MIT License

220 stars 56 forks source link

Would angluar shifts from 0 to +pi be sufficient? #20

Closed bgobbi closed 6 years ago

bgobbi commented 6 years ago

Thanks fro making this available this is very exciting research!

Ever since reading the papers I have been asking myself if it is really necessary to sample the angular shift from -PI to +PI? It seems to me that you can not distinguish positive from negative angles and therefore should be able to just sample from 0 to PI. Since you are looking at triples of atoms I see no way to define right hand versus left hand angles.

Could you comment on this? Thanks, Alberto

Jussmith01 commented 6 years ago

Hi Alberto,

The representation in earlier code worked in 0 to Pi but externally -Pi to Pi was passed in. This has since been changed in the newest code (in the 'test_branch' branch) which is currently undergoing tests and will soon be merged into master.

You can see this in the following parameters file: https://github.com/isayev/ASE_ANI/blob/test_branch/ani_models/ani-1ccx_8x/rHCNO-5.2R_16-3.5A_a4-8.params'

Best, Justin

bgobbi commented 6 years ago

Thanks @Jussmith01!