def activeLearning(data_d, data_model, labelPairFunction, num_questions) :
training_data = []
duplicates = []
nonduplicates = []
num_iterations = 100
pairs = blocking.allCandidates(data_d)
record_distances = core.recordDistances(pairs, data_d, data_model)
for _ in range(num_questions) :
print "finding the next uncertain pair ..."
uncertain_indices = findUncertainPairs(record_distances, data_model)
record_distances = record_distances[: , uncertain_indices]
uncertain_pairs = record_distances['pairs'][0:1]
record_distances = record_distances[1:]
labeled_pairs = labelPairFunction(uncertain_pairs, data_d, data_model)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs, training_data, data_model)
data_model = core.trainModel(training_data, num_iterations, data_model)
training_pairs = {0 : nonduplicates, 1 : duplicates}
return(training_data, training_pairs, data_model)
def activeLearning(candidates,
data_model,
labelPairFunction,
training_data,
training_pairs=None):
"""
Ask the user to label the record pair we are most uncertain of. Train the
data model, and update our uncertainty. Repeat until user tells us she is
finished.
"""
fields = [field for field in data_model['fields']
if data_model['fields'][field]['type'] != 'Missing Data']
duplicates = []
nonduplicates = []
if training_pairs:
nonduplicates.extend(training_pairs[0])
duplicates.extend(training_pairs[1])
finished = False
import time
t_train = time.time()
field_distances = core.fieldDistances(candidates, data_model)
logging.info('calculated fieldDistances in %s seconds',
str(time.time() - t_train))
seen_indices = set()
while finished == False:
logging.info('finding the next uncertain pair ...')
uncertain_indices = findUncertainPairs(field_distances, data_model)
for uncertain_index in uncertain_indices:
if uncertain_index not in seen_indices:
seen_indices.add(uncertain_index)
break
uncertain_pairs = [(candidates[uncertain_index][0][1],
candidates[uncertain_index][1][1])]
(labeled_pairs, finished) = labelPairFunction(uncertain_pairs, fields)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs, data_model, training_data)
if len(training_data) > 0:
data_model = core.trainModel(training_data, data_model, 1)
else:
raise ValueError('No training pairs given')
training_pairs = {0: nonduplicates, 1: duplicates}
return (training_data, training_pairs, data_model)
def activeLearning(candidates,
data_model,
labelPairFunction,
training_data,
training_pairs=None):
"""
Ask the user to label the record pair we are most uncertain of. Train the
data model, and update our uncertainty. Repeat until user tells us she is
finished.
"""
fields = [field for field in data_model['fields']
if data_model['fields'][field]['type'] != 'Missing Data']
duplicates = []
nonduplicates = []
if training_pairs:
nonduplicates.extend(training_pairs[0])
duplicates.extend(training_pairs[1])
finished = False
import time
t_train = time.time()
field_distances = core.fieldDistances(candidates, data_model)
logging.info('calculated fieldDistances in %s seconds',
str(time.time() - t_train))
seen_indices = set()
while finished == False:
logging.info('finding the next uncertain pair ...')
uncertain_indices = findUncertainPairs(field_distances, data_model)
for uncertain_index in uncertain_indices:
if uncertain_index not in seen_indices:
seen_indices.add(uncertain_index)
break
uncertain_pairs = [(candidates[uncertain_index][0][1],
candidates[uncertain_index][1][1])]
(labeled_pairs, finished) = labelPairFunction(uncertain_pairs, fields)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs, data_model, training_data)
if len(training_data) > 0:
data_model = core.trainModel(training_data, data_model, 1)
else:
raise ValueError('No training pairs given')
training_pairs = {0: nonduplicates, 1: duplicates}
return (training_data, training_pairs, data_model)
def activeLearning(data_d,
data_model,
labelPairFunction,
training_data,
training_pairs = None,
key_groups = []
):
duplicates = []
nonduplicates = []
if training_pairs :
nonduplicates.extend(training_pairs[0])
duplicates.extend(training_pairs[1])
finished = False
candidates = blocking.allCandidates(data_d, key_groups)
import time
t_train = time.time()
record_distances = core.recordDistances(candidates, data_model)
print 'calculated recordDistances in ', time.time() - t_train, 'seconds'
while finished == False :
print 'finding the next uncertain pair ...'
uncertain_indices = findUncertainPairs(record_distances,
data_model)
# pop the next most uncertain pair off of record distances
record_distances = record_distances[:, uncertain_indices]
uncertain_pair_ids = (record_distances['pairs'])[0:1]
record_distances = record_distances[1:]
uncertain_pairs = []
for pair in uncertain_pair_ids :
record_pair = [data_d[instance] for instance in pair]
record_pair = tuple(record_pair)
uncertain_pairs.append(record_pair)
labeled_pairs, finished = labelPairFunction(uncertain_pairs,
data_model)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs,
data_model,
training_data)
if len(training_data) > 0 :
data_model = core.trainModel(training_data, data_model, 1)
else :
raise ValueError("No training pairs given")
training_pairs = {0: nonduplicates, 1: duplicates}
return (training_data, training_pairs, data_model)
def train(self, data_d, training_source=None, key_groups=[]) :
"""
Learn field weights and blocking predicate from file of
labeled examples or round of interactive labeling
Keyword arguments:
data_d -- a dictionary of records
training_source -- either a path to a file of labeled examples or
a labeling function
In the dictionary of records, the keys are unique identifiers
for each record, the values are a dictionary where the keys
are the names of the record field and values are the record
values.
For Example,
{
854: {'city': 'san francisco',
'address': '300 de haro st.',
'name': "sally's cafe & bakery",
'cuisine': 'american'},
855: {'city': 'san francisco',
'address': '1328 18th st.',
'name': 'san francisco bbq',
'cuisine': 'thai'}
}
The labeling function will be used to do active learning. The
function will be supplied a list of examples that the learner
is the most 'curious' about, that is examples where we are most
uncertain about how they should be labeled. The labeling function
will label these, and based upon what we learn from these
examples, the labeling function will be supplied with new
examples that the learner is now most curious about. This will
continue until the labeling function sends a message that we
it is done labeling.
The labeling function must be a function that takes two
arguments. The first argument is a sequence of pairs of
records. The second argument is the data model.
The labeling function must return two outputs. The function
must return a dictionary of labeled pairs and a finished flag.
The dictionary of labeled pairs must have two keys, 1 and 0,
corresponding to record pairs that are duplicates or
nonduplicates respectively. The values of the dictionary must
be a sequence of records pairs, like the sequence that was
passed in.
The 'finished' flag should take the value False for active
learning to continue, and the value True to stop active learning.
i.e.
labelFunction(record_pairs, data_model) :
...
return (labeled_pairs, finished)
For a working example, see consoleLabel in training_sample
Labeled example files are typically generated by saving the
examples labeled in a previous session. If you need details
for this file see the method writeTraining.
"""
if (training_source.__class__ is not str
and not isinstance(training_source, types.FunctionType)):
raise ValueError
# data_d = core.sampleDict(data_d, 700) #we should consider changing this
print "data_d length: ", len(data_d)
self.data_d = dict([(key, core.frozendict(value)) for key, value in data_d.iteritems()])
if training_source.__class__ is str:
print 'reading training from file'
if not hasattr(self, 'training_data'):
self.initializeTraining(training_source)
self.training_pairs, self.training_data = self._readTraining(training_source,
self.training_data)
elif isinstance(training_source, types.FunctionType) :
if not hasattr(self, 'training_data'):
self.initializeTraining()
(self.training_data,
self.training_pairs,
self.data_model) = training_sample.activeLearning(self.data_d,
self.data_model,
training_source,
self.training_data,
self.training_pairs,
key_groups)
self.alpha = crossvalidation.gridSearch(self.training_data,
core.trainModel,
self.data_model,
k=20)
self.data_model = core.trainModel(self.training_data,
self.data_model,
self.alpha)
self._printLearnedWeights()
def activeLearning(candidates,
data_model,
labelPairFunction,
training_data,
training_pairs=None):
"""
Ask the user to label the record pair we are most uncertain of. Train the
data model, and update our uncertainty. Repeat until user tells us she is
finished.
"""
fields = [
field for field in data_model['fields']
if data_model['fields'][field]['type'] not in ('Missing Data',
'Interaction',
'Higher Categories')
]
duplicates = []
nonduplicates = []
if training_pairs:
nonduplicates.extend(training_pairs[0])
duplicates.extend(training_pairs[1])
if training_data.shape[0] == 0:
rand_int = random.randint(0, len(candidates))
exact_match = candidates[rand_int]
training_data = addTrainingData({
1: [exact_match] * 2,
0: []
}, data_model, training_data)
data_model = core.trainModel(training_data, data_model, .1)
finished = False
import time
t_train = time.time()
field_distances = core.fieldDistances(candidates, data_model)
logging.info('calculated fieldDistances in %s seconds',
str(time.time() - t_train))
seen_indices = set()
while finished == False:
logging.info('finding the next uncertain pair ...')
uncertain_indices = findUncertainPairs(field_distances, data_model,
(len(duplicates) /
(len(nonduplicates) + 1.0)))
for uncertain_index in uncertain_indices:
if uncertain_index not in seen_indices:
seen_indices.add(uncertain_index)
break
uncertain_pairs = [candidates[uncertain_index]]
(labeled_pairs, finished) = labelPairFunction(uncertain_pairs, fields)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs, data_model,
training_data)
if len(training_data) > 0:
data_model = core.trainModel(training_data, data_model, .1)
else:
raise ValueError('No training pairs given')
training_pairs = {0: nonduplicates, 1: duplicates}
return (training_data, training_pairs, data_model)
def train(self, data_sample, training_source=None, key_groups=[]):
"""
Learn field weights and blocking predicate from file of
labeled examples or round of interactive labeling
Keyword arguments:
data_sample -- a sample of record pairs
training_source -- either a path to a file of labeled examples or
a labeling function
In the sample of record_pairs, each element is a tuple of two
records. Each record is, in turn, a tuple of the record's key and
a record dictionary.
In in the record dictionary the keys are the names of the
record field and values are the record values.
For example, a data_sample with only one pair of records,
[
(
(854, {'city': 'san francisco',
'address': '300 de haro st.',
'name': "sally's cafe & bakery",
'cuisine': 'american'}),
(855, {'city': 'san francisco',
'address': '1328 18th st.',
'name': 'san francisco bbq',
'cuisine': 'thai'})
)
]
The labeling function will be used to do active learning. The
function will be supplied a list of examples that the learner
is the most 'curious' about, that is examples where we are most
uncertain about how they should be labeled. The labeling function
will label these, and based upon what we learn from these
examples, the labeling function will be supplied with new
examples that the learner is now most curious about. This will
continue until the labeling function sends a message that we
it is done labeling.
The labeling function must be a function that takes two
arguments. The first argument is a sequence of pairs of
records. The second argument is the data model.
The labeling function must return two outputs. The function
must return a dictionary of labeled pairs and a finished flag.
The dictionary of labeled pairs must have two keys, 1 and 0,
corresponding to record pairs that are duplicates or
nonduplicates respectively. The values of the dictionary must
be a sequence of records pairs, like the sequence that was
passed in.
The 'finished' flag should take the value False for active
learning to continue, and the value True to stop active learning.
i.e.
labelFunction(record_pairs, data_model) :
...
return (labeled_pairs, finished)
For a working example, see consoleLabel in training
Labeled example files are typically generated by saving the
examples labeled in a previous session. If you need details
for this file see the method writeTraining.
"""
self.data_sample = data_sample
if training_source.__class__ is not str and not isinstance(training_source, types.FunctionType):
raise ValueError
if training_source.__class__ is str:
logging.info("reading training from file")
if not hasattr(self, "training_data"):
self.initializeTraining(training_source)
self.training_pairs, self.training_data = self._readTraining(training_source, self.training_data)
elif isinstance(training_source, types.FunctionType):
if not hasattr(self, "training_data"):
self.initializeTraining()
(self.training_data, self.training_pairs, self.data_model) = training.activeLearning(
self.data_sample, self.data_model, training_source, self.training_data, self.training_pairs, key_groups
)
self.alpha = crossvalidation.gridSearch(self.training_data, core.trainModel, self.data_model, k=20)
self.data_model = core.trainModel(self.training_data, self.data_model, self.alpha)
self._logLearnedWeights()
def activeLearning(candidates,
data_model,
labelPairFunction,
training_data,
training_pairs=None):
"""
Ask the user to label the record pair we are most uncertain of. Train the
data model, and update our uncertainty. Repeat until user tells us she is
finished.
"""
fields = [field for field in data_model['fields']
if data_model['fields'][field]['type'] not in ('Missing Data',
'Interaction',
'Higher Categories')]
duplicates = []
nonduplicates = []
if training_pairs:
nonduplicates.extend(training_pairs[0])
duplicates.extend(training_pairs[1])
if training_data.shape[0] == 0 :
rand_int = random.randint(0, len(candidates))
exact_match = candidates[rand_int]
training_data = addTrainingData({1:[exact_match]*2,
0:[]},
data_model,
training_data)
data_model = core.trainModel(training_data, data_model, .1)
finished = False
import time
t_train = time.time()
field_distances = core.fieldDistances(candidates, data_model)
logging.info('calculated fieldDistances in %s seconds',
str(time.time() - t_train))
seen_indices = set()
while finished == False:
logging.info('finding the next uncertain pair ...')
uncertain_indices = findUncertainPairs(field_distances,
data_model,
(len(duplicates)/
(len(nonduplicates)+1.0)))
for uncertain_index in uncertain_indices:
if uncertain_index not in seen_indices:
seen_indices.add(uncertain_index)
break
uncertain_pairs = [candidates[uncertain_index]]
(labeled_pairs, finished) = labelPairFunction(uncertain_pairs, fields)
nonduplicates.extend(labeled_pairs[0])
duplicates.extend(labeled_pairs[1])
training_data = addTrainingData(labeled_pairs, data_model, training_data)
if len(training_data) > 0:
data_model = core.trainModel(training_data, data_model, .1)
else:
raise ValueError('No training pairs given')
training_pairs = {0: nonduplicates, 1: duplicates}
return (training_data, training_pairs, data_model)