-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathDScribe_eigen.py
177 lines (132 loc) · 7.92 KB
/
DScribe_eigen.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
# -*- coding: utf-8 -*-
# A Fourier-transformed feature engineering design for predicting ternary perovskite properties by coupling a two-dimensional convolutional neural
# network with a support vector machine (Conv2D-SVM)
# AUTHOR - (1) * Ericsson Chenebuah, (1) Michel Nganbe and (2) Alain Tchagang
# 1: Department of Mechanical Engineering, University of Ottawa, 75 Laurier Ave. East, Ottawa, ON, K1N 6N5 Canada
# 2: Digital Technologies Research Centre, National Research Council of Canada, 1200 Montréal Road, Ottawa, ON, K1A 0R6 Canada
# * email: echen013@uottawa.ca
# (06-Dec-2022)
"""DScribe_eigen.py
Automatically generated by Colaboratory.
"""
## THIS SCRIPT DESCRIBES THE COMPUTATIONAL CODES FOR THE PERIODIC GLOBAL DESCRIPTORS i.e. COULOMB, EWALD-SUM, AND SINE MATRIX IN THEIR EIGENVALUE FORMS USING "DScribe".
## THE GENERALIZED FEATURE IS CONCATENATED WITH THE EIGENVALUE DESCRIPTORS AND USED TO ENHANCE THE PREDICTIVE CAPABILITY OF THE PERIODIC GLOBAL DESCRIPTOR.
# The data file to run this code is separated and compressed as "DScribe Data" on the github repository.
# All calculations related to the atomic coordinates were converted from Angstroms to Atomic Units
#! pip install dscribe
import pandas as pd
import numpy as np
from dscribe.descriptors import CoulombMatrix, EwaldSumMatrix, SineMatrix
from ase import Atoms
# 5 Number of atoms in unit cell
df1 = pd.read_csv('DScribe_5.csv')
data_cm1 = []
data_esm1 = []
data_sine1 = []
for i in range (0, 13195):
perov = Atoms(symbols=[df1.iloc[(i,0)], df1.iloc[(i,1)], df1.iloc[(i,2)], df1.iloc[(i,3)], df1.iloc[(i,4)]],
positions=[(df1.iloc[i, 5:8]), (df1.iloc[i, 8:11]), (df1.iloc[i, 11:14]), (df1.iloc[i, 14:17]), (df1.iloc[i, 17:20])],
cell=[df1.iloc[(i,20)], df1.iloc[(i,21)], df1.iloc[(i,22)], df1.iloc[(i,23)], df1.iloc[(i,24)], df1.iloc[(i,25)]], pbc=True)
cm= CoulombMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
cm_out = cm.create(perov)
data_cm1.append(cm_out)
esm= EwaldSumMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
esm_out = esm.create(perov)
data_esm1.append(esm_out)
sine= SineMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
sine_out = sine.create(perov)
data_sine1.append(sine_out)
data_cm1 = np.array(data_cm1)
data_esm1 = np.array(data_esm1)
data_sine1 = np.array(data_sine1)
# 10 Number of atoms in unit cell
df2 = pd.read_csv('DScribe_10.csv')
data_cm2 = []
data_esm2 = []
data_sine2 = []
for i in range (0, 10277):
perov = Atoms(symbols=[df2.iloc[(i,0)], df2.iloc[(i,1)], df2.iloc[(i,2)], df2.iloc[(i,3)], df2.iloc[(i,4)], df2.iloc[(i,5)], df2.iloc[(i,6)], df2.iloc[(i,7)], df2.iloc[(i,8)],
df2.iloc[(i,9)]],
positions=[(df2.iloc[i, 10:13]), (df2.iloc[i, 13:16]), (df2.iloc[i, 16:19]), (df2.iloc[i, 19:22]), (df2.iloc[i, 22:25]), (df2.iloc[i, 25:28]),
(df2.iloc[i, 28:31]), (df2.iloc[i, 31:34]), (df2.iloc[i, 34:37]), (df2.iloc[i, 37:40])],
cell=[df2.iloc[(i,40)], df2.iloc[(i,41)], df2.iloc[(i,42)], df2.iloc[(i,43)], df2.iloc[(i,44)], df2.iloc[(i,45)]], pbc=True)
cm= CoulombMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
cm_out = cm.create(perov)
data_cm2.append(cm_out)
esm= EwaldSumMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
esm_out = esm.create(perov)
data_esm2.append(esm_out)
sine= SineMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
sine_out = sine.create(perov)
data_sine2.append(sine_out)
data_cm2 = np.array(data_cm2)
data_esm2 = np.array(data_esm2)
data_sine2 = np.array(data_sine2)
# 15 Number of atoms in unit cell
df3 = pd.read_csv('DScribe_15.csv')
data_cm3 = []
data_esm3 = []
data_sine3 = []
for i in range (0, 49):
perov = Atoms(symbols=[df3.iloc[(i,0)], df3.iloc[(i,1)], df3.iloc[(i,2)], df3.iloc[(i,3)], df3.iloc[(i,4)], df3.iloc[(i,5)], df3.iloc[(i,6)], df3.iloc[(i,7)], df3.iloc[(i,8)],
df3.iloc[(i,9)], df3.iloc[(i,10)], df3.iloc[(i,11)], df3.iloc[(i,12)], df3.iloc[(i,13)], df3.iloc[(i,14)]],
positions=[(df3.iloc[i, 15:18]), (df3.iloc[i, 18:21]), (df3.iloc[i, 21:24]), (df3.iloc[i, 24:27]), (df3.iloc[i, 27:30]), (df3.iloc[i, 30:33]), (df3.iloc[i, 33:36]),
(df3.iloc[i, 36:39]), (df3.iloc[i, 39:42]), (df3.iloc[i, 42:45]), (df3.iloc[i, 45:48]), (df3.iloc[i, 48:51]), (df3.iloc[i, 51:54]), (df3.iloc[i, 54:57]),
(df3.iloc[i, 57:60])],
cell=[df3.iloc[(i,60)], df3.iloc[(i,61)], df3.iloc[(i,62)], df3.iloc[(i,63)], df3.iloc[(i,64)], df3.iloc[(i,65)]], pbc=True)
cm= CoulombMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
cm_out = cm.create(perov)
data_cm3.append(cm_out)
esm= EwaldSumMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
esm_out = esm.create(perov)
data_esm3.append(esm_out)
sine= SineMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
sine_out = sine.create(perov)
data_sine3.append(sine_out)
data_cm3 = np.array(data_cm3)
data_esm3 = np.array(data_esm3)
data_sine3 = np.array(data_sine3)
# 20 Number of atoms in unit cell
df4 = pd.read_csv('DScribe_20.csv')
data_cm4 = []
data_esm4 = []
data_sine4 = []
for i in range (0, 4066):
perov = Atoms(symbols=[df4.iloc[(i,0)], df4.iloc[(i,1)], df4.iloc[(i,2)], df4.iloc[(i,3)], df4.iloc[(i,4)], df4.iloc[(i,5)], df4.iloc[(i,6)], df4.iloc[(i,7)], df4.iloc[(i,8)],
df4.iloc[(i,9)], df4.iloc[(i,10)], df4.iloc[(i,11)], df4.iloc[(i,12)], df4.iloc[(i,13)], df4.iloc[(i,14)], df4.iloc[(i,15)], df4.iloc[(i,16)], df4.iloc[(i,17)],
df4.iloc[(i,18)], df4.iloc[(i,19)]],
positions=[(df4.iloc[i, 20:23]), (df4.iloc[i, 23:26]), (df4.iloc[i, 26:29]), (df4.iloc[i, 29:32]), (df4.iloc[i, 32:35]), (df4.iloc[i, 35:38]), (df4.iloc[i, 38:41]),
(df4.iloc[i, 41:44]), (df4.iloc[i, 44:47]), (df4.iloc[i, 47:50]), (df4.iloc[i, 50:53]), (df4.iloc[i, 53:56]), (df4.iloc[i, 56:59]), (df4.iloc[i, 59:62]),
(df4.iloc[i, 62:65]), (df4.iloc[i, 65:68]), (df4.iloc[i, 68:71]), (df4.iloc[i, 71:74]), (df4.iloc[i, 74:77]), (df4.iloc[i, 77:80])],
cell=[df4.iloc[(i,80)], df4.iloc[(i,81)], df4.iloc[(i,82)], df4.iloc[(i,83)], df4.iloc[(i,84)], df4.iloc[(i,85)]], pbc=True)
cm= CoulombMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
cm_out = cm.create(perov)
data_cm4.append(cm_out)
esm= EwaldSumMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
esm_out = esm.create(perov)
data_esm4.append(esm_out)
sine= SineMatrix(n_atoms_max=20, flatten=False, permutation='eigenspectrum')
sine_out = sine.create(perov)
data_sine4.append(sine_out)
data_cm4 = np.array(data_cm4)
data_esm4 = np.array(data_esm4)
data_sine4 = np.array(data_sine4)
# concatenate all matrices
eigen_cm=np.abs(np.concatenate([data_cm1, data_cm2, data_cm3, data_cm4], axis=0))
eigen_esm=np.abs(np.concatenate([data_esm1, data_esm2, data_esm3, data_esm4], axis=0))
eigen_sine=np.abs(np.concatenate([data_sine1, data_sine2, data_sine3, data_sine4], axis=0))
from sklearn.preprocessing import MinMaxScaler
scaler_cm = MinMaxScaler()
eigen_cm = np.asarray(scaler_cm.fit_transform(eigen_cm))
scaler_esm = MinMaxScaler()
eigen_esm = np.asarray(scaler_esm.fit_transform(eigen_esm))
scaler_sine = MinMaxScaler()
eigen_sine = np.asarray(scaler_sine.fit_transform(eigen_sine))
## GENERALIZED FEATURES
GF = np.asarray(pd.read_csv('gen_feat.csv').astype('float32'))
scaler_gen = MinMaxScaler()
GF = np.asarray(scaler_gen.fit_transform(GF))
# Concatenated periodic input descriptor
x_cm = np.concatenate((GF,eigen_cm), axis=-1) #Coulomb matrix eigen representation with generalized features
x_esm = np.concatenate((GF,eigen_esm), axis=-1) #Ewald-sum matrix eigen representation with generalized features
x_sine = np.concatenate((GF,eigen_sine), axis=-1) #Sine matrix eigen representation with generalized features