Modifications for LMBTR for Element Encoding about dscribe HOT 4 CLOSED

Hi!

Yes, it might be reasonable to come up with some form of a k=1 term for LMBTR as well. I guess we could just add the possibility of using a radial weighting from the local center that would be applied to the Gaussians. Is that something you would be interested in? The resulting spectrum would crudely tell which atomic species are most prominent near a point.

from dscribe.

Yes, that would definitely be very useful. Do you have any suggestions on how this functionality could be added on the user end?

from dscribe.

You mean how the end user would configure the weighting? I would implement it like this:

lmbtr = LMBTR(
    species=["H", "O"],
    k1={
        "geometry": {"function": "atomic_number"},
        "grid": {"min": 0, "max": 100, "n": 100, "sigma": 0.1},
        "weighting": {"function": "exponential", "scale": 0.5, "cutoff": 1e-3},
    }, ...
)

We can also quite easily add new weighting functions. Any suggestions for them are welcome.

from dscribe.

Closing for now, reopen again if still relevant.

from dscribe.