def get_data(A):
def sanity_check(M, N):
for m, n in zip(M, N):
if any([mm not in n for mm in m]):
return False
else:
return True
dpath = '../../get_db_fp/{}_fp.db'.format(A)
with FingerprintDB(dpath) as fp:
data = fp.get_fingerprints(params=np.arange(1, 108))
metadata = fp.get_metadata(params=np.arange(1, 6))
dpath = '../../get_db_fp_catlearn/{}_fp_catlearn.db'.format(A)
with FingerprintDB(dpath) as fp:
data_cl = fp.get_fingerprints(params=np.arange(1, 1062))
metadata2 = fp.get_metadata(params=np.arange(1, 6))
target = data[:, -1]
data = data[:, :-1]
print(data.shape)
print(data_cl.shape)
features = np.concatenate((data, data_cl), axis=1)
if sanity_check(metadata2, metadata):
return features, target, metadata
else:
return None, None, None
def get_data(A, l_A, C, l_C):
def get_cond(A, B):
if not set(A) == set(B):
return False
if '|' in A:
if not A.split('|')[1] == B.split('|')[1]:
return False
if not Counter(A.split('|')[0].split('_')) == \
Counter(B.split('|')[0].split('_')):
return False
else:
if not Counter(A.split('_')) == Counter(B.split('_')):
return False
return True
with FingerprintDB(A) as fp:
data_A = fp.get_fingerprints(params=np.arange(1, l_A))
data_A = data_A[:, -1]
metadata_A = fp.get_metadata(params=np.arange(1, 6))
with FingerprintDB(C) as fp:
data_C = fp.get_fingerprints(params=np.arange(1, l_C))
data_C = data_C[:, -1]
metadata_C = fp.get_metadata(params=np.arange(1, 6))
x = []
y = []
metadata = []
for a, b in zip(data_A, metadata_A):
for c, d in zip(data_C, metadata_C):
if all([
b[1] == d[1], b[2] == d[2],
get_cond(b[3], d[3]), b[4] == d[4]
]):
x += [a]
y += [c]
metadata += [[
b[0], b[1], b[2], b[3], b[4], d[0], d[1], d[2], d[3], d[4]
]]
return x, y, metadata
def get_data(A, B):
def get_sym(A):
ordered_metals = [
'Y', 'La', 'Sc', 'Zr', 'Hf', 'Ti', 'Ta', 'Nb', 'V', 'Cr', 'Mo',
'W', 'Re', 'Tc', 'Os', 'Ru', 'Ir', 'Rh', 'Ni', 'Co', 'Fe', 'Mn',
'Pt', 'Pd', 'Au', 'Ag', 'Cu', 'Zn', 'Cd', 'Hg', 'Al', 'Ga', 'In',
'Tl', 'Pb', 'Sn', 'Bi'
]
if '3' in A:
return A, 'L12'
elif A in ordered_metals:
return A, 'A1'
else:
sym = []
for i in ordered_metals:
if i in A:
sym += [i]
if not len(sym) == 2:
A = 'ERROR!!'
else:
return '2'.join(sym) + '2', 'L10'
def get_cond(A, B):
if not set(A) == set(B):
return False
if '|' in A:
if not A.split('|')[1] == B.split('|')[1]:
return False
if not Counter(A.split('|')[0].split('_')) == \
Counter(B.split('|')[0].split('_')):
return False
else:
if not Counter(A.split('_')) == Counter(B.split('_')):
return False
return True
def sanity_check(M, N):
for m, n in zip(M, N):
if any([mm not in n for mm in m]):
return False
else:
return True
dpath = '../../../get_db_fp/{}_fp.db'.format(B)
with FingerprintDB(dpath) as fp:
data = fp.get_fingerprints(params=np.arange(1, 108))
md1 = fp.get_metadata(params=np.arange(1, 6))
dpath = '../../../get_db_fp_catlearn/{}_fp_catlearn.db'.format(B)
with FingerprintDB(dpath) as fp:
data_cl = fp.get_fingerprints(params=np.arange(1, 1062))
md2 = fp.get_metadata(params=np.arange(1, 6))
temp_target = data[:, -1]
data = data[:, :-1]
print(data.shape, data_cl.shape)
temp_features = np.concatenate((data, data_cl), axis=1)
del data, data_cl
if not sanity_check(md2, md1):
sys.exit('Something went wrong in sanity check.')
scaling_data_processed = []
scaling_data = np.load(
'../../../get_data_cathub/{}_data.npy'.format(A))[()]
scaling_data = scaling_data['reactions']['edges']
for d in scaling_data:
if A in ['N', 'H']:
if not '0.5' in d['node']['reactants']:
continue
if A == 'S':
if 'Hstar' in d['node']['products']:
continue
adsorbate = re.findall(r'"([A-Z0-4]+)":', d['node']['sites'])[0]
reactionenergy = d['node']['reactionEnergy']
symbol, sb_symbol = get_sym(d['node']['surfaceComposition'])
site, site_type = re.findall(r'(?<=: )"(.+)"',
d['node']['sites'])[0].split('|', 1)
if site not in ['top', 'bridge', 'hollow']:
print('Site {} not a valid site_type.'.format(site))
continue
if sb_symbol == 'A1':
site_type = site_type.replace('A', symbol)
elif sb_symbol == 'L12':
a, b = symbol.split('3')
site_type = site_type.replace('A', 'aaa').replace('B', 'bbb')
site_type = site_type.replace('aaa', a).replace('bbb', b)
else:
a, b, _ = symbol.split('2')
if a > b:
a, b = b, a
site_type = site_type.replace('A', 'aaa').replace('B', 'bbb')
site_type = site_type.replace('aaa', a).replace('bbb', b)
if ':' in site_type:
print('This is a problematic site type.\n\n\n')
continue
scaling_data_processed += [[
adsorbate, symbol, site, site_type, sb_symbol, reactionenergy
]]
scaling_target = []
metadata = []
features = []
target = []
for a, b, c in zip(md1, temp_features, temp_target):
for d in scaling_data_processed:
if all([
a[1] == d[1], a[2] == d[2],
get_cond(a[3], d[3]), a[4] == d[4]
]):
features += [b]
target += [c]
scaling_target += [d[5]]
metadata += [[
a[0], a[1], a[2], a[3], a[4], d[0], d[1], d[2], d[3], d[4]
]]
features = np.array(features)
target = np.array(target)
scaling_target = np.array(scaling_target)
return features, target, scaling_target, metadata
import os
from collections import Counter
def get_refined_data(A, B):
M = A.copy()
N = B.copy()
mask = np.invert(np.isnan(M) | np.isnan(N))
return M[mask], N[mask]
adsorbates = ['CH', 'CH2', 'CH3', 'OH', 'SH', 'NH']
for ads in adsorbates:
print('Working on adsorbates: {}'.format(ads))
with FingerprintDB('../get_data_object/{}_fp.db'.format(ads)) as f:
data = f.get_fingerprints(params=np.arange(107, 110))
EA = data[:, 1]
EAx = data[:, 2]
E = data[:, 0]
del data
x, y = get_refined_data(EA, E)
jac1 = LinearScaling(np.array(x), np.array(y), 'EA', 'E', 'eV')
x, y = get_refined_data(EAx, E)
jac2 = LinearScaling(np.array(x), np.array(y), 'EAx', 'E', 'eV')
jac1.get_coeff()
jac2.get_coeff()
a1 = jac1.plot_scaling()
a2 = jac2.plot_scaling()
if not os.path.exists('Jacobsen_{}'.format(ads)):
operation_list_local += [
'bonding_convolution',
'layered_sum',
'local_ads_metal_fp'
]
for datum in data:
#generate atoms object
#get data obtained from cathub
w_data = np.load('../get_data_cathub/{}_data.npy'.format(datum))[()]
#connect the database to store the objects
db = connect('{}_atoms.db'.format(datum))
fp_file = '{}_fp.db'.format(datum)
with FingerprintDB(fp_file) as fpd:
for i in range(2):
fpd.parameter_entry('AN_PC{}'.format(i),
'Atomic Number Slab: pc {}'.format(i))
fpd.parameter_entry('AR_PC{}'.format(i),
'Atomic Radius Slab: pc {}'.format(i))
fpd.parameter_entry('DBC_PC{}'.format(i),
'Dband center Slab: pc {}'.format(i))
fpd.parameter_entry('DBW_PC{}'.format(i),
'Dband Width Slab: pc {}'.format(i))
fpd.parameter_entry('DBS_PC{}'.format(i),
'Dband Skewness Slab: pc {}'.format(i))
fpd.parameter_entry('DBK_PC{}'.format(i),
'Dband Kurtosis Slab: pc {}'.format(i))
fpd.parameter_entry('DP_PC{}'.format(i),
'Dipole Polarizability Slab: pc {}'.format(i))
Adata[D][C][3])
except TypeError:
q1 = None
try:
q2 = 0.33 * (Adata[D[0]][A][3] +
Adata[D[0]][B][3] +
Adata[D[0]][C][3])
except TypeError:
q2 = None
return q1, q2
data = ['CH', 'CH2', 'CH3', 'OH', 'NH', 'SH']
jacobsen_data = {d: None for d in data}
for d in data:
#f_loc = '/Users/osmanmamun/Delta_Learning_Paper/get_db_fp/{}_fp.db'
with FingerprintDB('{}_fp.db'.format(d)) as f:
metadata = f.get_metadata(params=np.arange(1, 6))
energies = f.get_fingerprints(params=np.array([107]))
E = []
EA = []
EAx = []
md = []
for i, m in enumerate(metadata, 1):
if m[-1] == 'A1':
continue
if not 'FCC' in m[3]:
continue
A, B, C = m[3].split('|')[0].split('_')
E += [energies[i]]
A, B = get_E(A, B, C, d)
EA += [A]