def __init__(self,
unlabeled_datasets=None,
models=None,
undersample_before_eval=False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(
self,
unlabeled_datasets=unlabeled_datasets,
models=models,
undersample_before_eval=undersample_before_eval)
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
# there's no reason to rebuild the models during active learning,
# since we're just randomly picking things to label
self.rebuild_models_at_each_iter = False
print "switching query function to RANDOM"
#
# Here we switch the query function to randomly sampling. Note that
# this function actually lives in the base_learner parent class, because
# it may very well be useful for other learners to request ids for random
# unlabeled examples
#
self.query_function = self.get_random_unlabeled_ids
self.name = "Random"
print "%s learner intialized with %s labeled examples" % (
self.name, self.labeled_datasets[0].size())
def __init__(self, unlabeled_datasets = [], experts=None,
models=None, undersample_before_eval = False, request_path=None):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval)
#
# most importantly we change the query function to SIMPLE here
#
self.query_function = self.DT
def __init__(self, unlabeled_datasets = None, models=None, undersample_before_eval = False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval)
#
# most importantly we change the query function to max_contention here
#
print "switching query function to max contention!"
self.query_function = self.max_contention
self.name = "CoTester_max_contention"
def __init__(self, unlabeled_datasets = [], models=None, undersample_before_eval = False, request_path=None):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval)
#
# most importantly we change the query function to SIMPLE here
#
self.query_function = self.faker
self.name ="FAKER"
self.get_these = eval(open(request_path, 'r').readline())
def __init__(self, learners, unlabeled_datasets=None, undersample_before_eval = False, kernel_type=RBF,
weights=[1,100], use_raw=False, name="STACKED"):
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets,
undersample_before_eval = undersample_before_eval)
self.learners = learners
self.labeled_datasets = None# stacked dataset
self.unlabeled_datasets = None
self.kernel_type = kernel_type
self.weights = weights
# If use_raw is true, the meta features will be the raw (signed) distances of
# points from the hyperplanes, rather than the binary 1/-1 predictions.
self.use_raw = use_raw
self.name = name
def __init__(self, unlabeled_datasets = [], models=None, undersample_before_eval = False, kernel_type=LINEAR, svm_params=None):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval, kernel_type=kernel_type,
svm_params=svm_params)
#
# most importantly we change the query function to SIMPLE here
#
print "switching query function to SIMPLE!"
self.query_function = self.SIMPLE
self.name = "SIMPLE"
def __init__(self, unlabeled_datasets = [], models=None):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models)
#super(SimpleLearner, self).__init__()
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
print "switching query function to SIMPLE!"
#
# most importantly we change the query function to SIMPLE here
#
self.query_function = self.SIMPLE
self.name = "SIMPLE"
def __init__(self, unlabeled_datasets = [], models=None):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models)
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
print "switching query function to RANDOM"
#
# Here we switch the query function to randomly sampling. Note that
# this function actually lives in the base_learner parent class, because
# it may very well be useful for other learners to request ids for random
# unlabeled examples
#
self.query_function = self.get_random_unlabeled_ids
self.name = "Random"
def __init__(self,
unlabeled_datasets=None,
models=None,
undersample_before_eval=False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(
self,
unlabeled_datasets=unlabeled_datasets,
models=models,
undersample_before_eval=undersample_before_eval)
#
# most importantly we change the query function to max_contention here
#
print "switching query function to max contention!"
self.query_function = self.max_contention
self.name = "CoTester_max_contention"
def __init__(self, unlabeled_datasets = [], models=None, undersample_before_eval=False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval=undersample_before_eval)
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
print "switching query function to RANDOM"
#
# Here we switch the query function to randomly sampling. Note that
# this function actually lives in the base_learner parent class, because
# it may very well be useful for other learners to request ids for random
# unlabeled examples
#
self.query_function = self.get_random_unlabeled_ids
self.name = "Random"
def __init__(self, unlabeled_datasets = None, models=None, undersample_before_eval = False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(self, unlabeled_datasets=unlabeled_datasets, models=models,
undersample_before_eval = undersample_before_eval)
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
# there's no reason to rebuild the models during active learning,
# since we're just randomly picking things to label
self.rebuild_models_at_each_iter = False
print "switching query function to RANDOM"
#
# Here we switch the query function to randomly sampling. Note that
# this function actually lives in the base_learner parent class, because
# it may very well be useful for other learners to request ids for random
# unlabeled examples
#
self.query_function = self.get_random_unlabeled_ids
self.name = "Random"
print "%s learner intialized with %s labeled examples" % (self.name, self.labeled_datasets[0].size())
def __init__(self,
unlabeled_datasets=[],
models=None,
undersample_before_eval=False):
#
# call the BaseLearner constructor to initialize various globals and process the
# datasets, etc.; of course, these can subsequently be overwritten.
BaseSVMLearner.__init__(
self,
unlabeled_datasets=unlabeled_datasets,
models=models,
undersample_before_eval=undersample_before_eval)
# ovewrite svm parameters here
self.params = [svm_parameter() for d in unlabeled_datasets]
print "switching query function to SIMPLE!"
#
# most importantly we change the query function to SIMPLE here
#
self.query_function = self.SIMPLE
self.name = "SIMPLE"
def __init__(self,
learners,
unlabeled_datasets=None,
undersample_before_eval=False,
kernel_type=RBF,
weights=[1, 100],
use_raw=False,
name="STACKED"):
BaseSVMLearner.__init__(
self,
unlabeled_datasets=unlabeled_datasets,
undersample_before_eval=undersample_before_eval)
self.learners = learners
self.labeled_datasets = None # stacked dataset
self.unlabeled_datasets = None
self.kernel_type = kernel_type
self.weights = weights
# If use_raw is true, the meta features will be the raw (signed) distances of
# points from the hyperplanes, rather than the binary 1/-1 predictions.
self.use_raw = use_raw
self.name = name