You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With #110 it is now easy to make a Rnl basis, or more generally a scalar basis P_v of the form
P_v = sum_n U_vn J_n
where v is a general multi-index. The parameters U_vn can then be learned as part of the parameter estimation for example.
But we also want Rnla bases where a is the species of the neighbouring atom!! This is a problem because the species is not an index but it is part of the argument.
But is it really a problem? Or is the 1-hot embedding designed to precisely overcome this? So this may or may not really be an issue, but I'm now posting this here to make sure somebody thinks this through and implements an example of such a Rnla basis.
The text was updated successfully, but these errors were encountered:
With #110 it is now easy to make a
Rnl
basis, or more generally a scalar basisP_v
of the formwhere v is a general multi-index. The parameters
U_vn
can then be learned as part of the parameter estimation for example.But we also want
Rnla
bases wherea
is the species of the neighbouring atom!! This is a problem because the species is not an index but it is part of the argument.But is it really a problem? Or is the 1-hot embedding designed to precisely overcome this? So this may or may not really be an issue, but I'm now posting this here to make sure somebody thinks this through and implements an example of such a
Rnla
basis.The text was updated successfully, but these errors were encountered: