Gaussian descriptor¶
Custom parameters¶
The Gaussian descriptor creates feature vectors based on the Behler scheme, and defaults to values used in Nano Letters 14:2670, 2014. 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(80.),
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())