def __init__(self,
unlabeled_datasets=None,
models=None,
undersample_before_eval=False,
weights=[1, 100],
kernel_type=LINEAR):
BaseLearner.__init__(self,
unlabeled_datasets=unlabeled_datasets,
models=models,
undersample_before_eval=undersample_before_eval)
self.unlabeled_datasets = unlabeled_datasets or []
# params correspond to each of the respective models (one if we're in a single feature space)
# these specify things like what kind of kernel to use. here we just use the default, but
# *you'll probably want to overwrite this* in your subclass. see the libsvm doc for more information
# (in particular, svm_test.py is helpful).
print "%s: using kernel type: %s" % (self.name, kernel_type)
self.weights = weights
self.kernel_type = kernel_type
self.params = [
svm_parameter(weight=self.weights, kernel_type=self.kernel_type)
for d in self.unlabeled_datasets
]
self.div_hash = {}
def __init__(self,
dataset_,
model=None,
rebuild_model_at_each_iter=True,
name=""):
BaseLearner.__init__(self, dataset_, model, rebuild_model_at_each_iter,
"Random" + name)
self.description = "RandomSampling selects new samples at random at each step"
def __init__(self, unlabeled_datasets = [], models = None):
BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets)
# params correspond to each of the respective models (one if we're in a single feature space)
# these specify things like what kind of kernel to use. here we just use the default, but
# *you'll probably want to overwrite this* in your subclass. see the libsvm doc for more information (in particular,
# svm_test.py is helpful).
self.params = [svm_parameter() for d in unlabeled_datasets]
self.div_hash = {}
def __init__(self,
dataset_,
model=None,
rebuild_model_at_each_iter=True,
name=""):
BaseLearner.__init__(self, dataset_, model, rebuild_model_at_each_iter,
"MaxEntropy" + name)
self.description = "MaxEntropySampling chooses those samples from the pool, " \
"where the model entropy of p(y|x) is the highest"
def __init__(self, unlabeled_datasets = [], models = None, undersample_before_eval=False):
BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets,
undersample_before_eval=undersample_before_eval)
# params correspond to each of the respective models (one if we're in a single feature space)
# these specify things like what kind of kernel to use. here we just use the default, but
# *you'll probably want to overwrite this* in your subclass. see the libsvm doc for more information (in particular,
# svm_test.py is helpful).
self.params = [svm_parameter() for d in unlabeled_datasets]
self.div_hash = {}
def __init__(self,
dataset_,
model=None,
rebuild_model_at_each_iter=True,
name=""):
BaseLearner.__init__(self, dataset_, model, rebuild_model_at_each_iter,
"LeastConfident" + name)
if self.dataset.type == "classification" and len(
self.dataset.classes) > 2:
raise Exception(
'Least confident sampling is only defined for 2-class problems, this one has {} classes'
.format(len(self.dataset.classes)))
self.description = "LeastConfidentSampling chooses those samples from the pool, " \
"where the model is least confident"
def __init__(self, unlabeled_datasets = None, models = None, undersample_before_eval = False,
weights=[1,100], kernel_type=LINEAR):
BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval)
self.unlabeled_datasets = unlabeled_datasets or []
# params correspond to each of the respective models (one if we're in a single feature space)
# these specify things like what kind of kernel to use. here we just use the default, but
# *you'll probably want to overwrite this* in your subclass. see the libsvm doc for more information
# (in particular, svm_test.py is helpful).
print "%s: using kernel type: %s" % (self.name, kernel_type)
self.weights = weights
self.kernel_type = kernel_type
self.params = [svm_parameter(weight=self.weights, kernel_type=self.kernel_type) for d in self.unlabeled_datasets]
self.div_hash = {}
def __init__(self,
dataset_,
model=None,
rebuild_model_at_each_iter=True,
update_parameters_sample=True,
min_prob=1e-15,
name=""):
if model is None:
if dataset_.type == "classification":
model = bayesian_models.BinBetaPrior(
same_posterior_params=True)
if dataset_.type == "regression":
model = bayesian_models.GaussianPrior()
BaseLearner.__init__(self, dataset_, model, rebuild_model_at_each_iter,
"LindleyInformation" + name)
self.description = "LindleyInformation chooses those samples from the pool, " \
"which provide maximum expected information gain"
self.model.posterior_parameters_sample = None
self.update_parameters_sample = update_parameters_sample
self.min_prob = min_prob
def __init__(self, unlabeled_datasets, models=None, undersample_before_eval=False): BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models, undersample_before_eval=undersample_before_eval)
def __init__(self, unlabeled_datasets, models=None):
BaseLearner.__init__(self,
unlabeled_datasets=unlabeled_datasets,
models=models)
def __init__(self, unlabeled_datasets): BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets)
def __init__(self, unlabeled_datasets, models=None, undersample_before_eval=False): BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models, undersample_before_eval=undersample_before_eval)
def __init__(self, unlabeled_datasets, models=None): BaseLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models)
