def get_data():
"""Simple function to pull some training and test data."""
# Attach the database.
dd = DescriptorDatabase(db_name='{}/vec_store.sqlite'.format(wkdir),
table='FingerVector')
# Pull the features and targets from the database.
names = dd.get_column_names()
features, targets = names[1:-1], names[-1:]
feature_data = dd.query_db(names=features)
target_data = np.reshape(dd.query_db(names=targets),
(np.shape(feature_data)[0], ))
# Split the data into so test and training sets.
train_features = feature_data[:train_size, :n_features]
train_targets = target_data[:train_size]
test_features = feature_data[test_size:, :n_features]
test_targets = target_data[test_size:]
return train_features, train_targets, test_features, test_targets
def todb(self, features, targets):
"""Function to convert numpy arrays to basic db."""
data = np.concatenate((features, np.reshape(targets,
(len(targets), 1))),
axis=1)
uid = [str(uuid.uuid4()) for _ in range(len(targets))]
data = np.concatenate((np.reshape(uid, (len(uid), 1)), data), axis=1)
descriptors = ['f' + str(i) for i in range(np.shape(features)[1])]
targets = ['target']
names = descriptors + targets
# Set up the database to save system descriptors.
dd = DescriptorDatabase(db_name=self.db_name, table=self.table)
dd.create_db(names=names)
# Fill the database with the data.
dd.fill_db(descriptor_names=names, data=data)
def test_expand(self):
"""Generate an extended feature space."""
# Attach the database.
dd = DescriptorDatabase(db_name='{}/vec_store.sqlite'.format(wkdir),
table='FingerVector')
# Pull the features and targets from the database.
names = dd.get_column_names()
features, targets = names[1:-1], names[-1:]
feature_data = dd.query_db(names=features)
target_data = np.reshape(dd.query_db(names=targets),
(np.shape(feature_data)[0], ))
# Split the data into so test and training sets.
train_features = feature_data[:train_size, :]
train_targets = target_data[:train_size]
test_features = feature_data[test_size:, :]
d, f = np.shape(train_features)
td, tf = np.shape(test_features)
# Make some toy names.
names = ['f{}'.format(i) for i in range(f)]
# Perform feature engineering.
extend = fe.single_transform(train_features)
self.assertTrue(np.shape(extend) == (d, f * 3))
extend = fe.get_order_2(train_features)
ext_n = fe.get_labels_order_2(names, div=False)
self.assertTrue(np.shape(extend) == (d, f * (f + 1) / 2))
self.assertTrue(len(ext_n) == np.shape(extend)[1])
extend = fe.get_div_order_2(train_features)
ext_n = fe.get_labels_order_2(names, div=True)
self.assertTrue(np.shape(extend) == (d, f**2))
self.assertTrue(len(ext_n) == np.shape(extend)[1])
extend = fe.get_order_2ab(train_features, a=2, b=4)
ext_n = fe.get_labels_order_2ab(names, a=2, b=4)
self.assertTrue(np.shape(extend) == (d, f * (f + 1) / 2))
self.assertTrue(len(ext_n) == np.shape(extend)[1])
extend = fe.get_ablog(train_features, a=2, b=4)
ext_n = fe.get_labels_ablog(names, a=2, b=4)
self.assertTrue(np.shape(extend) == (d, f * (f + 1) / 2))
self.assertTrue(len(ext_n) == np.shape(extend)[1])
p = train_features[:3, :10]
fe.generate_features(p,
max_num=2,
max_den=0,
log=False,
sqrt=False,
exclude=False,
s=True)
fe.generate_features(p,
max_num=2,
max_den=1,
log=True,
sqrt=True,
exclude=True,
s=True)
self.__class__.train_features = train_features
self.__class__.train_targets = train_targets
self.__class__.test_features = test_features
def test_storage(self):
"""Test database functions."""
# Define variables for database to store system descriptors.
db_name = '/vec_store.sqlite'
descriptors = ['f' + str(i) for i in range(np.shape(self.data)[1])]
targets = ['Energy']
names = descriptors + targets
# Set up the database to save system descriptors.
dd = DescriptorDatabase(db_name=wkdir + db_name, table='FingerVector')
dd.create_db(names=names)
# Put data in correct format to be inserted into database.
print('Generate the database')
new_data = []
for i, a in zip(self.data, self.all_cand):
d = []
d.append(a.info['unique_id'])
for j in i:
d.append(j)
d.append(a.info['key_value_pairs']['raw_score'])
new_data.append(d)
# Fill the database with the data.
dd.fill_db(descriptor_names=names, data=new_data)
# Test out the database functions.
train_fingerprint = dd.query_db(names=descriptors)
train_target = dd.query_db(names=targets)
print('\nfeature data for candidates:\n', train_fingerprint,
'\ntarget data for candidates:\n', train_target)
cand_data = dd.query_db(unique_id='7a216711c2eae02decc04da588c9e592')
print('\ndata for random candidate:\n', cand_data)
all_id = dd.query_db(names=['uuid'])
dd.create_column(new_column=['random'])
for i in all_id:
dd.update_descriptor(descriptor='random',
new_data=random(),
unique_id=i[0])
print('\nretrieve random vars:\n', dd.query_db(names=['random']))
print('\nretrieved column names:\n', dd.get_column_names())