average soap about dscribe

[lauri-codes]

Hi @gabor1!

Yes, we indeed simply average over the power spectrum of each atom. The method you described does certainly sound better. Just to be super clear, you mean that instead of simply averaging the power spectrum of each atom (the math will take a second to render):

(where i sums over the atoms and n is the total number of atoms in the structure) we would put the summation inside the power spectrum coefficients:

Can you point to any possible references that discuss this kind form of averaging? Once you verify that I understood correctly, I can think about how to get it implemented. Thanks a lot!

[gabor1]

Your formulae are correct. I think the first mention of this is in here https://pubs.acs.org/doi/10.1021/acs.jpclett.8b01067 (We've had it in the code for a long time but didn't use it very much until recently). Of course if users want the average of the soap vectors they can just call the function without this flag and then average the array themselves. The point is that the kind of average I'm talking about isn't accessible unless you provide those option.

…

-- Gábor

On 13 Feb 2020, at 22:10, Lauri Himanen ***@***.***> wrote:  Hi @gabor1! Yes, we indeed simply average over the power spectrum of each atom. The method you described does certainly sound better. Just to be super clear, you mean that instead of simply averaging the power spectrum of each atom (the math will take a second to render): (where i sums over the atoms and n is the total number of atoms in the structure) we would put the summation inside the power spectrum coefficients: Can you point to any possible references that discuss this kind form of averaging? Once you verify that I understood correctly, I can think about how to get it implemented. Thanks a lot! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[lauri-codes]

Now added this average mode in 0.4.0. It can be activated with average="inner".

[gabor1]

thanks!