Gaussian descriptor¶

Custom parameters¶

The Gaussian descriptor creates feature vectors based on the Behler scheme, and defaults to a small set of reasonable values. The values employed are always written to the log file and within saved instances of Amp calculators. You can specify custom parameters for the elements of the feature vectors as listed in the documentation of the Gaussian class.

There is also a helper function make_symmetry_functions() within the amp.descriptor.gaussian module to assist with this. An example of making a custom fingerprint is given below for a two-element system.

import numpy as np
from amp import Amp
from amp.descriptor.gaussian import Gaussian, make_symmetry_functions
from amp.model.neuralnetwork import NeuralNetwork

elements = ['Cu', 'Pt']
G = make_symmetry_functions(elements=elements, type='G2',
etas=np.logspace(np.log10(0.05), np.log10(5.),
num=4))
G += make_symmetry_functions(elements=elements, type='G4',
etas=[0.005],
zetas=[1., 4.],
gammas=[+1., -1.])

G = {'Cu': G,
'Pt': G}
calc = Amp(descriptor=Gaussian(Gs=G),
model=NeuralNetwork())

To include angular symmetry functions of triplets inside the cutoff sphere but with distances larger than the cutoff radius you need to slightly modify the snippet above:

import numpy as np
from amp import Amp
from amp.descriptor.gaussian import Gaussian, make_symmetry_functions
from amp.model.neuralnetwork import NeuralNetwork

elements = ['Cu', 'Pt']
G = make_symmetry_functions(elements=elements, type='G2',
etas=np.logspace(np.log10(0.05), np.log10(5.),
num=4))
G += make_symmetry_functions(elements=elements, type='G5',
etas=[0.005],
zetas=[1., 4.],
gammas=[+1., -1.])

G = {'Cu': G,
'Pt': G}
calc = Amp(descriptor=Gaussian(Gs=G),
model=NeuralNetwork())