Select structures

This example shows how to select structures from dataset

NEP 3 calculator with 1 symbols: C
------------------------------
  zbl                 : False
  radial_cutoff       : 4.2
  angular_cutoff      : 3.7
  n_max_radial        : 10
  n_max_angular       : 8
  basis_size_radial   : 10
  basis_size_angular  : 8
  l_max_3body         : 4
  num_node            : 65
  num_para            : 6903
  element_list        : ['C']
------------------------------

from pynep.calculate import NEP
from pynep.select import FarthestPointSample
from ase.io import read, write
import numpy as np
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA


a = read('data.traj', ':')
calc = NEP("C_2022_NEP3.txt")
print(calc)
des = np.array([np.mean(calc.get_property('descriptor', i), axis=0) for i in a])
sampler = FarthestPointSample(min_distance=0.05)
selected_i = sampler.select(des, [])
write('selected.traj', [a[i] for  i in selected_i])

reducer = PCA(n_components=2)
reducer.fit(des)
proj = reducer.transform(des)
plt.scatter(proj[:,0], proj[:,1], label='all data')
selected_proj = reducer.transform(np.array([des[i] for i in selected_i]))
plt.scatter(selected_proj[:,0], selected_proj[:,1], label='selected data')
plt.legend()
plt.axis('off')
plt.savefig('select.png')