def get_original_training(path_to_features, method='Bazin', screen=False):
"""Read original full light curve training sample
Parameters
----------
path_to_features: str
Complete path to file holding full light curve features.
method: str (optional)
Feature extraction method. Only option implemented is "Bazin".
screen: bool (optional)
If true, show on screen comments on the dimensions of some key elements.
Returns
-------
snactclass.DataBase
Information about the original full light curve analys.
"""
data = DataBase()
data.load_features(path_to_features, method=method, screen=screen)
data.build_samples(initial_training='original', screen=screen)
return data
def learn_loop(nloops: int, strategy: str, path_to_features: str,
output_metrics_file: str, output_queried_file: str,
features_method='Bazin', classifier='RandomForest',
training='original', batch=1, screen=True):
"""Perform the active learning loop. All results are saved to file.
Parameters
----------
nloops: int
Number of active learning loops to run.
strategy: str
Query strategy. Options are 'UncSampling' and 'RandomSampling'.
path_to_features: str
Complete path to input features file.
output_metrics_file: str
Full path to output file to store metric values of each loop.
output_queried_file: str
Full path to output file to store the queried sample.
features_method: str (optional)
Feature extraction method. Currently only 'Bazin' is implemented.
classifier: str (optional)
Machine Learning algorithm.
Currently only 'RandomForest' is implemented.
training: str or int (optional)
Choice of initial training sample.
If 'original': begin from the train sample flagged in the file
If int: choose the required number of samples at random,
ensuring that at least half are SN Ia
Default is 'original'.
batch: int (optional)
Size of batch to be queried in each loop. Default is 1.
screen: bool (optional)
If True, print on screen number of light curves processed.
"""
# initiate object
data = DataBase()
# load features
data.load_features(path_to_features, method=features_method,
screen=screen)
# separate training and test samples
data.build_samples(initial_training=training)
for loop in range(nloops):
if screen:
print('Processing... ', loop)
# classify
data.classify(method=classifier)
# calculate metrics
data.evaluate_classification()
# choose object to query
indx = data.make_query(strategy=strategy, batch=batch)
# update training and test samples
data.update_samples(indx, loop=loop)
# save metrics for current state
data.save_metrics(loop=loop, output_metrics_file=output_metrics_file,
batch=batch, epoch=loop)
# save query sample to file
data.save_queried_sample(output_queried_file, loop=loop,
full_sample=False)
def time_domain_loop(days: list, output_metrics_file: str,
output_queried_file: str,
path_to_features_dir: str, strategy: str,
batch=1, canonical = False, classifier='RandomForest',
features_method='Bazin', path_to_canonical="",
path_to_full_lc_features="", queryable=True,
screen=True, training='original'):
"""Perform the active learning loop. All results are saved to file.
Parameters
----------
days: list
List of 2 elements. First and last day of observations since the
beginning of the survey.
output_metrics_file: str
Full path to output file to store metrics for each loop.
output_queried_file: str
Full path to output file to store the queried sample.
path_to_features_dir: str
Complete path to directory holding features files for all days.
strategy: str
Query strategy. Options are 'UncSampling' and 'RandomSampling'.
batch: int (optional)
Size of batch to be queried in each loop. Default is 1.
canonical: bool (optional)
If True, restrict the search to the canonical sample.
classifier: str (optional)
Machine Learning algorithm.
Currently only 'RandomForest' is implemented.
features_method: str (optional)
Feature extraction method. Currently only 'Bazin' is implemented.
path_to_canonical: str (optional)
Path to canonical sample features files.
It is only used if "strategy==canonical".
path_to_full_lc_features: str (optional)
Path to full light curve features file.
Only used if training is a number.
queryable: bool (optional)
If True, allow queries only on objects flagged as queryable.
Default is True.
screen: bool (optional)
If True, print on screen number of light curves processed.
training: str or int (optional)
Choice of initial training sample.
If 'original': begin from the train sample flagged in the file
If int: choose the required number of samples at random,
ensuring that at least half are SN Ia
Default is 'original'.
"""
# initiate object
data = DataBase()
# load features for the first day
path_to_features = path_to_features_dir + 'day_' + str(int(days[0])) + '.dat'
data.load_features(path_to_features, method=features_method,
screen=screen)
# change training
if training == 'original':
data.build_samples(initial_training=0, screen=screen, queryable=queryable)
full_lc_features = get_original_training(path_to_features=path_to_full_lc_features,
screen=screen)
ini_train_ids = full_lc_features.train_metadata['id'].values
data.train_metadata = full_lc_features.train_metadata
data.train_labels = full_lc_features.train_labels
data.train_features = full_lc_features.train_features
# remove repeated ids
test_flag = np.array([item not in ini_train_ids
for item in data.test_metadata['id'].values])
data.test_metadata = data.test_metadata[test_flag]
data.test_labels = data.test_labels[test_flag]
data.test_features = data.test_features[test_flag]
else:
data.build_samples(initial_training=int(training), screen=screen,
queryable=queryable)
ini_train_ids = []
# get list of canonical ids
if canonical:
canonical = DataBase()
canonical.load_features(path_to_file=path_to_canonical)
data.queryable_ids = canonical.queryable_ids
for night in range(int(days[0]), int(days[-1]) - 1):
if screen:
print('Processing night: ', night)
print(' ... train: ', data.train_metadata.shape[0])
print(' ... test: ', data.test_metadata.shape[0])
print(' ... queryable_ids: ', data.queryable_ids.shape[0])
# cont loop
loop = night - int(days[0])
# classify
data.classify(method=classifier, screen=screen)
# calculate metrics
data.evaluate_classification(screen=screen)
# choose object to query
indx = data.make_query(strategy=strategy, batch=batch, screen=screen)
# update training and test samples
data.update_samples(indx, loop=loop, screen=screen)
# save metrics for current state
data.save_metrics(loop=loop, output_metrics_file=output_metrics_file,
batch=batch, epoch=night)
# save query sample to file
data.save_queried_sample(output_queried_file, loop=loop,
full_sample=False)
# load features for next day
path_to_features2 = path_to_features_dir + 'day_' + str(night + 1) + '.dat'
data_tomorrow = DataBase()
data_tomorrow.load_features(path_to_features2, method=features_method,
screen=False)
data_tomorrow.build_samples(initial_training=0, screen=screen)
# remove training samples from new test
for obj in data.train_metadata['id'].values:
if obj in data_tomorrow.test_metadata['id'].values:
indx_tomorrow = list(data_tomorrow.test_metadata['id'].values).index(obj)
if obj not in ini_train_ids:
# remove old features from training
indx_today = list(data.train_metadata['id'].values).index(obj)
data.train_metadata = data.train_metadata.drop(data.train_metadata.index[indx_today])
data.train_labels = np.delete(data.train_labels, indx_today, axis=0)
data.train_features = np.delete(data.train_features, indx_today, axis=0)
# update new features of the training with new obs
flag = np.arange(0, data_tomorrow.test_metadata.shape[0]) == indx_tomorrow
data.train_metadata = pd.concat([data.train_metadata, data_tomorrow.test_metadata[flag]], axis=0,
ignore_index=True)
data.train_features = np.append(data.train_features,
data_tomorrow.test_features[flag], axis=0)
data.train_labels = np.append(data.train_labels,
data_tomorrow.test_labels[flag], axis=0)
# remove from new test sample
data_tomorrow.test_metadata = data_tomorrow.test_metadata.drop(data_tomorrow.test_metadata.index[indx_tomorrow])
data_tomorrow.test_labels = np.delete(data_tomorrow.test_labels, indx_tomorrow, axis=0)
data_tomorrow.test_features = np.delete(data_tomorrow.test_features, indx_tomorrow, axis=0)
# use new test data
data.test_metadata = data_tomorrow.test_metadata
data.test_labels = data_tomorrow.test_labels
data.test_features = data_tomorrow.test_features
if strategy == 'canonical':
data.queryable_ids = canonical.queryable_ids
else:
data.queryable_ids = data_tomorrow.queryable_ids
if queryable:
queryable_flag = data_tomorrow.test_metadata['queryable'].values
data.queryable_ids = data_tomorrow.test_metadata['id'].values[queryable_flag]
else:
data.queryable_ids = data_tomorrow.metadata['id'].values[~train_flag]
# check if there are repeated ids
for name in data.train_metadata['id'].values:
if name in data.test_metadata['id'].values:
raise ValueError('End of time_domain_loop: ' + \
'Object ', name, ' found in test and training sample!')
# check if all queried samples are in the training sample
for i in range(len(data.queried_sample)):
for j in range(len(data.queried_sample[i])):
if data.queried_sample[i][j][1] not in data.train_metadata['id'].values:
raise ValueError('Object ', name, 'was queried but is missing from training!')
if screen:
print('\n End of time domain loop:')
print(' ... train: ', data.train_metadata.shape[0])
print(' ... test: ', data.test_metadata.shape[0])
print(' ... queryable: ', data.queryable_ids.shape[0], '\n')
def time_domain_loop(days: list,
output_diag_file: str,
output_queried_file: str,
path_to_features_dir: str,
strategy: str,
batch=1,
canonical=False,
classifier='RandomForest',
features_method='Bazin',
path_to_canonical="",
path_to_full_lc_features="",
screen=True,
training='original'):
"""Perform the active learning loop. All results are saved to file.
Parameters
----------
days: list
List of 2 elements. First and last day of observations since the
beginning of the survey.
output_diag_file: str
Full path to output file to store diagnostics of each loop.
output_queried_file: str
Full path to output file to store the queried sample.
path_to_features_dir: str
Complete path to directory holding features files for all days.
strategy: str
Query strategy. Options are 'UncSampling' and 'RandomSampling'.
batch: int (optional)
Size of batch to be queried in each loop. Default is 1.
canonical: bool (optional)
If True, restrict the search to the canonical sample.
classifier: str (optional)
Machine Learning algorithm.
Currently only 'RandomForest' is implemented.
features_method: str (optional)
Feature extraction method. Currently only 'Bazin' is implemented.
path_to_canonical: str (optional)
Path to canonical sample features files.
It is only used if "strategy==canonical".
path_to_full_lc_features: str (optional)
Path to full light curve features file.
Only used if training is a number.
screen: bool (optional)
If True, print on screen number of light curves processed.
training: str or int (optional)
Choice of initial training sample.
If 'original': begin from the train sample flagged in the file
If int: choose the required number of samples at random,
ensuring that at least half are SN Ia
Default is 'original'.
"""
# initiate object
data = DataBase()
# load features for the first day
path_to_features = path_to_features_dir + 'day_' + str(int(
days[0])) + '.dat'
data.load_features(path_to_features, method=features_method, screen=screen)
# change training
if training == 'original':
data.build_samples(initial_training='original')
full_lc_features = get_original_training(
path_to_features=path_to_full_lc_features)
data.train_metadata = full_lc_features.train_metadata
data.train_labels = full_lc_features.train_labels
data.train_features = full_lc_features.train_features
else:
data.build_samples(initial_training=int(training))
# get list of canonical ids
if canonical:
canonical = DataBase()
canonical.load_features(path_to_file=path_to_canonical)
data.queryable_ids = canonical.queryable_ids
for night in range(int(days[0]), int(days[-1]) - 1):
if screen:
print('Processing night: ', night)
# cont loop
loop = night - int(days[0])
# classify
data.classify(method=classifier)
# calculate metrics
data.evaluate_classification()
# choose object to query
indx = data.make_query(strategy=strategy, batch=batch)
# update training and test samples
data.update_samples(indx, loop=loop)
# save diagnostics for current state
data.save_metrics(loop=loop,
output_metrics_file=output_diag_file,
batch=batch,
epoch=night)
# save query sample to file
data.save_queried_sample(output_queried_file,
loop=loop,
full_sample=False)
# load features for next day
path_to_features2 = path_to_features_dir + 'day_' + str(night +
1) + '.dat'
data_tomorrow = DataBase()
data_tomorrow.load_features(path_to_features2,
method=features_method,
screen=False)
# notice that original here corresponds to original in the file
# not original full light curve training
data_tomorrow.build_samples('original')
# use new test data
data.test_metadata = data_tomorrow.test_metadata
data.test_labels = data_tomorrow.test_labels
data.test_features = data_tomorrow.test_features
if strategy == 'canonical':
data.queryable_ids = canonical.queryable_ids
else:
data.queryable_ids = data_tomorrow.queryable_ids