def get_naive_bayes_conf(self):
name = '-'.join([
'AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num, 'all', 'NaiveBayes'
])
classifier_conf = self.exp.exp_conf.core_conf.classifier_conf
optim_conf = classifier_conf.hyperparam_conf.optim_conf
multiclass = True
hyperparam_conf = HyperparamConf.get_default(
optim_conf.num_folds, optim_conf.n_jobs, multiclass,
GaussianNaiveBayesConf._get_hyper_desc(), self.exp.logger)
naive_bayes_conf = GaussianNaiveBayesConf(multiclass, hyperparam_conf,
self.exp.logger)
test_conf = UnlabeledLabeledConf(self.exp.logger, None)
classification_conf = ClassificationConf(naive_bayes_conf, test_conf,
self.exp.logger)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
classification_conf,
name=name,
parent=self.exp.exp_id)
naive_bayes_exp = DiademExp(exp_conf, session=self.exp.session)
naive_bayes_exp.create_exp()
return naive_bayes_conf
class UpdateModel(CoreUpdateModel):
def __init__(self, iteration):
CoreUpdateModel.__init__(self, iteration)
self.exp = self.iteration.exp
self.model_exp = None
def execute(self):
name = 'AL%d-Iter%d-main' % (self.exp.exp_id, self.iteration.iter_num)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
self.model_conf,
name=name,
parent=self.exp.exp_id)
self.model_exp = DiademExp(exp_conf, session=self.exp.session)
self.model_exp.run(instances=self.iteration.datasets.instances,
cv_monitoring=True)
self._set_exec_time()
self.classifier = self.model_exp.get_train_test_exp('train').classifier
def _set_exec_time(self):
train_monitor = self.model_exp.get_train_test_exp('train').monitoring
test_monitor = self.model_exp.get_train_test_exp('test').monitoring
self.exec_time = sum(
[m.exec_time for m in [train_monitor, test_monitor]])
def monitoring(self, al_dir, iteration_dir):
with_validation = self.iteration.conf.validation_conf is not None
self.monitoring = ModelPerfEvolution(self.iteration.iter_num,
self.model_exp, with_validation)
self.monitoring.generate()
self.monitoring.export(iteration_dir, al_dir)
def execute(self):
name = 'AL%d-Iter%d-main' % (self.exp.exp_id, self.iteration.iter_num)
features_conf = FeaturesConf(
self.exp.exp_conf.features_conf.input_features,
self.exp.exp_conf.features_conf.sparse,
self.exp.exp_conf.features_conf.logger,
filter_in_f=self.exp.exp_conf.features_conf.filter_in_f,
filter_out_f=self.exp.exp_conf.features_conf.filter_out_f)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
features_conf,
self.exp.exp_conf.annotations_conf,
self.model_conf,
None,
name=name,
parent=self.exp.exp_id)
self.model_exp = DiademExp(exp_conf, session=self.exp.session)
classifier_type = get_classifier_type(
self.model_conf.classifier_conf.__class__)
cv_monitoring = classifier_type == ClassifierType.supervised
prev_classifier = None
prev_iter = self.iteration.prev_iter
if prev_iter is not None:
prev_classifier = prev_iter.update_model.classifier
self.model_exp.run(instances=self.iteration.datasets.instances,
cv_monitoring=cv_monitoring,
init_classifier=prev_classifier)
self._set_exec_time()
self.classifier = self.model_exp.get_train_exp().classifier
def execute(self):
name = 'AL%d-Iter%d-main' % (self.exp.exp_id, self.iteration.iter_num)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
self.model_conf,
name=name,
parent=self.exp.exp_id)
self.model_exp = DiademExp(exp_conf, session=self.exp.session)
self.model_exp.run(instances=self.iteration.datasets.instances,
cv_monitoring=True)
self._set_exec_time()
self.classifier = self.model_exp.get_train_exp().classifier
def get_naive_bayes_conf(self):
name = '-'.join([
'AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num, 'all', 'NaiveBayes'
])
classifier_conf = self.exp.exp_conf.core_conf.classifier_conf
optim_conf = classifier_conf.hyperparam_conf.optim_conf
multiclass = True
factory = classifiers.get_factory()
naive_bayes_conf = factory.get_default('GaussianNaiveBayes',
optim_conf.num_folds,
optim_conf.n_jobs, multiclass,
self.exp.logger)
test_conf = UnlabeledLabeledConf(self.exp.logger)
classification_conf = ClassificationConf(naive_bayes_conf, test_conf,
self.exp.logger)
features_conf = FeaturesConf(
self.exp.exp_conf.features_conf.input_features,
self.exp.exp_conf.features_conf.sparse,
self.exp.exp_conf.features_conf.logger,
filter_in_f=self.exp.exp_conf.features_conf.filter_in_f,
filter_out_f=self.exp.exp_conf.features_conf.filter_out_f)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
features_conf,
self.exp.exp_conf.annotations_conf,
classification_conf,
None,
name=name,
parent=self.exp.exp_id)
DiademExp(exp_conf, session=self.exp.session)
return naive_bayes_conf
def _get_multiclass_conf(self):
conf = self.rcd_conf.classification_conf
name = '-'.join([
'AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num, self.label, 'analysis'
])
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
conf,
name=name,
parent=self.exp.exp_id)
self.multiclass_exp = DiademExp(exp_conf, session=self.exp.session)
self.multiclass_exp.create_exp()
return conf
def _run_logistic_regression(self):
name = '-'.join([
'AL%d' % (self.exp.exp_id),
'Iter%d' % (self.iteration.iter_num), 'all', 'LogisticRegression'
])
features_conf = FeaturesConf(
self.exp.exp_conf.features_conf.input_features,
self.exp.exp_conf.features_conf.sparse,
self.exp.exp_conf.features_conf.logger,
filter_in_f=self.exp.exp_conf.features_conf.filter_in_f,
filter_out_f=self.exp.exp_conf.features_conf.filter_out_f)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
features_conf,
self.exp.exp_conf.annotations_conf,
self.exp.exp_conf.core_conf.multiclass_model,
None,
name=name,
parent=self.exp.exp_id)
model_exp = DiademExp(exp_conf, session=self.exp.session)
model_exp.run(instances=self.iteration.datasets.instances,
cv_monitoring=False)
train_exp = model_exp.get_train_exp()
test_exp = model_exp.get_detection_exp('test')
self.lr_predicted_proba = test_exp.predictions.all_probas
self.lr_predicted_labels = test_exp.predictions.values
self.lr_class_labels = train_exp.classifier.class_labels
self.lr_time = train_exp.monitoring.exec_times.total()
self.lr_time += test_exp.monitoring.exec_time.predictions
def _get_multiclass_conf(self):
conf = self.rcd_conf.classification_conf
name = '-'.join(['AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num,
self.label,
'analysis'])
features_conf = FeaturesConf(
self.exp.exp_conf.features_conf.input_features,
self.exp.exp_conf.features_conf.sparse,
self.exp.exp_conf.features_conf.logger,
filter_in_f=self.exp.exp_conf.features_conf.filter_in_f,
filter_out_f=self.exp.exp_conf.features_conf.filter_out_f)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
features_conf,
self.exp.exp_conf.annotations_conf,
conf, None, name=name, parent=self.exp.exp_id)
self.multiclass_exp = DiademExp(exp_conf, session=self.exp.session)
return conf
class UpdateModel(CoreUpdateModel):
def __init__(self, iteration):
CoreUpdateModel.__init__(self, iteration)
self.exp = self.iteration.exp
self.model_exp = None
def execute(self):
name = 'AL%d-Iter%d-main' % (self.exp.exp_id, self.iteration.iter_num)
features_conf = FeaturesConf(
self.exp.exp_conf.features_conf.input_features,
self.exp.exp_conf.features_conf.sparse,
self.exp.exp_conf.features_conf.logger,
filter_in_f=self.exp.exp_conf.features_conf.filter_in_f,
filter_out_f=self.exp.exp_conf.features_conf.filter_out_f)
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
features_conf,
self.exp.exp_conf.annotations_conf,
self.model_conf,
None,
name=name,
parent=self.exp.exp_id)
self.model_exp = DiademExp(exp_conf, session=self.exp.session)
classifier_type = get_classifier_type(
self.model_conf.classifier_conf.__class__)
cv_monitoring = classifier_type == ClassifierType.supervised
prev_classifier = None
prev_iter = self.iteration.prev_iter
if prev_iter is not None:
prev_classifier = prev_iter.update_model.classifier
self.model_exp.run(instances=self.iteration.datasets.instances,
cv_monitoring=cv_monitoring,
init_classifier=prev_classifier)
self._set_exec_time()
self.classifier = self.model_exp.get_train_exp().classifier
def _set_exec_time(self):
training = self.model_exp.get_train_exp().monitoring
training_detect = self.model_exp.get_detection_exp('train').monitoring
detection = self.model_exp.get_detection_exp('test').monitoring
self.exec_time = training.exec_times.total()
self.exec_time += sum(
[m.exec_time.predictions for m in [training_detect, detection]])
def monitoring(self, al_dir, iteration_dir):
with_validation = self.iteration.conf.validation_conf is not None
self.monitoring = ModelPerfEvolution(self.iteration.iter_num,
self.model_exp, with_validation)
self.monitoring.generate()
self.monitoring.export(iteration_dir, al_dir)
class RcdQueries(CoreRcdQueries):
def __init__(self,
iteration,
label,
proba_min=None,
proba_max=None,
input_checking=True):
CoreRcdQueries.__init__(self,
iteration,
label,
proba_min,
proba_max,
input_checking=input_checking)
self.multiclass_exp = None
self.exp = iteration.exp
def generate_query(self,
instance_id,
predicted_proba,
suggested_label,
suggested_family,
confidence=None):
return Query(instance_id,
predicted_proba,
suggested_label,
suggested_family,
confidence=confidence)
def _get_multiclass_conf(self):
conf = self.rcd_conf.classification_conf
name = '-'.join([
'AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num, self.label, 'analysis'
])
exp_conf = DiademConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
conf,
None,
name=name,
parent=self.exp.exp_id)
self.multiclass_exp = DiademExp(exp_conf, session=self.exp.session)
self.multiclass_exp.create_exp()
return conf
def _create_clustering_exp(self):
core_conf = CoreClusteringConf(self.exp.exp_conf.logger,
self.categories.num_categories)
name = '-'.join([
'AL%d' % self.exp.exp_id,
'Iter%d' % self.iteration.iter_num, self.label, 'clustering'
])
exp_conf = ClusteringConf(self.exp.exp_conf.secuml_conf,
self.exp.exp_conf.dataset_conf,
self.exp.exp_conf.features_conf,
self.exp.exp_conf.annotations_conf,
core_conf,
name=name,
parent=self.exp.exp_id)
clustering_exp = ClusteringExperiment(exp_conf,
session=self.exp.session)
clustering_exp.create_exp()
return clustering_exp
def _gen_clustering_visu(self):
if self.families_analysis:
self.clustering_exp = self._create_clustering_exp()
clustering = Clusters(self.categories.instances,
self.categories.assigned_categories)
clustering.generate(None, None)
clustering.export(self.clustering_exp.output_dir())
else:
self.clustering_exp = None
def _set_categories(self, all_instances, assigned_categories,
predicted_proba):
self.categories = Categories(self.multiclass_exp, self.iteration,
all_instances, assigned_categories,
predicted_proba, self.label,
self.multiclass_model.class_labels)
