def set_data_set(self, data_set):
self.data_set = data_set
if data_set == bc.DATA_NG:
self.set_data_set_defaults('20ng-2000')
self.loss_function = loss_function.ZeroOneError()
self.cv_loss_function = loss_function.ZeroOneError()
#self.num_labels = [5, 10, 20, 40]
self.target_labels = CR[0]
self.source_labels = np.vstack((CR[1], ST[1]))
self.oracle_labels = CR[1]
elif data_set == bc.DATA_SYNTHETIC_HYP_TRANS_1_1:
self.set_data_set_defaults(
'synthetic_hyp_trans_class500-50-1.0-0.3-1-1')
self.num_labels = [10, 20, 40]
self.target_labels = None
self.source_labels = None
self.oracle_data_set_ids = np.asarray([1])
elif data_set == bc.DATA_SYNTHETIC_HYP_TRANS_2_2:
self.set_data_set_defaults(
'synthetic_hyp_trans_class500-50-1.0-0.3-2-2')
self.num_labels = [10, 20, 40]
self.target_labels = None
self.source_labels = None
self.oracle_data_set_ids = np.asarray([1, 2])
else:
assert False
'''
def set_synthetic_classification(self):
self.loss_function = loss_function.ZeroOneError()
self.data_dir = 'data_sets/synthetic_classification'
self.data_name = 'synthetic_classification'
self.data_set_file_name = 'split_data.pkl'
self.results_dir = 'synthetic_classification'
self.target_labels = np.asarray([1, 2])
#self.target_labels = array_functions.vec_to_2d(self.target_labels).T
self.source_labels = np.asarray([3, 4])
self.source_labels = array_functions.vec_to_2d(self.source_labels).T
self.cv_loss_function = loss_function.LogLoss()
def set_ng_transfer(self):
self.loss_function = loss_function.ZeroOneError()
self.set_ng()
'''
self.target_labels = np.asarray([1,2])
S1 = np.asarray([7,8])
S2 = np.asarray([12,13])
self.source_labels = np.vstack((S1,S2))
'''
self.target_labels = CR[0]
#self.source_labels = CR[1]
self.source_labels = np.vstack((CR[1], ST[1]))
self.oracle_labels = CR[1]
#self.source_labels = ST[1]
#self.oracle_labels = np.empty(0)
#self.cv_loss_function = loss_function.ZeroOneError()
self.cv_loss_function = loss_function.LogLoss()
def set_data_set(self, data_set):
self.data_set = data_set
if data_set == bc.DATA_SYNTHETIC_PIECEWISE:
self.set_data_set_defaults('synthetic_piecewise')
self.num_labels = np.asarray([num_starting_labels])
self.target_labels = np.zeros([1])
self.source_labels = np.ones([1])
elif data_set == bc.DATA_MNIST:
self.set_data_set_defaults('mnist')
self.num_labels = [num_starting_labels / 2]
self.target_labels = np.asarray([1, 3])
self.source_labels = np.asarray([7, 8])
self.loss_function = loss_function.ZeroOneError()
self.cv_loss_function = loss_function.ZeroOneError()
elif data_set == bc.DATA_BOSTON_HOUSING:
self.set_data_set_defaults('boston_housing-13(transfer)')
self.num_labels = [num_starting_labels]
self.target_labels = np.asarray([0])
self.source_labels = np.asarray([1])
elif data_set == bc.DATA_WINE:
self.set_data_set_defaults('wine-small-11')
self.num_labels = [num_starting_labels]
self.target_labels = np.asarray([0])
self.source_labels = np.asarray([1])
elif data_set == bc.DATA_CONCRETE:
self.set_data_set_defaults('concrete-7')
self.num_labels = [num_starting_labels]
self.target_labels = np.asarray([1])
self.source_labels = np.asarray([3])
elif data_set == bc.DATA_CLIMATE_MONTH:
self.set_data_set_defaults('climate-month',
source_labels=[0],
target_labels=[4],
is_regression=True)
self.num_labels = np.asarray([num_starting_labels])
elif data_set == bc.DATA_IRS:
self.set_data_set_defaults('irs-income',
source_labels=[0],
target_labels=[1],
is_regression=True)
self.num_labels = np.asarray([num_starting_labels])
elif data_set == bc.DATA_KC_HOUSING:
self.set_data_set_defaults('kc-housing-spatial-floors',
source_labels=[0],
target_labels=[1],
is_regression=True)
self.num_labels = np.asarray([num_starting_labels])
elif data_set == bc.DATA_ZILLOW:
self.set_data_set_defaults('zillow-traffic',
source_labels=[1],
target_labels=[0],
is_regression=True)
#self.set_data_set_defaults('zillow', source_labels=[1], target_labels=[0], is_regression=True)
self.num_labels = np.asarray([num_starting_labels])
elif data_set == bc.DATA_TAXI:
#self.set_data_set_defaults('taxi2-20', source_labels=[1], target_labels=[0], is_regression=True)
#self.set_data_set_defaults('taxi2-50', source_labels=[1], target_labels=[0], is_regression=True)
#self.set_data_set_defaults('taxi2', source_labels=[0], target_labels=[1], is_regression=True)
#self.set_data_set_defaults('taxi3', source_labels=[1], target_labels=[0], is_regression=True)
self.set_data_set_defaults('taxi',
source_labels=[1],
target_labels=[0],
is_regression=True)
#self.num_labels = np.asarray([5, 10, 20, 40, 100, 200, 400, 800])
self.num_labels = np.asarray([num_starting_labels])
else:
assert False, 'unknown transfer data set'
assert self.source_labels.size > 0
assert self.target_labels.size > 0
self.labels_to_not_sample = self.source_labels.ravel()
a = self.source_labels.ravel()
self.labels_to_keep = np.concatenate((self.target_labels, a))
from configs import base_configs
data_splitter = DataSplitter()
data_splitter.data = data
splits = data_splitter.generate_splits(data.y)
split_data = data_lib.SplitData(data, splits)
use_transfer = True
use_regression = False
m = base_configs.MethodConfigs()
m.use_validation = True
if use_transfer:
assert not use_regression
m.loss_function = loss_function.ZeroOneError()
m.cv_loss_function = loss_function.ZeroOneError()
transfer_learner = transfer_methods.StackingTransfer(deepcopy(m))
transfer_learner.base_learner = method.SKLLogisticRegression(deepcopy(m))
#transfer_learner.source_learner = method.SKLLogisticRegression(deepcopy(m))
transfer_learner.source_learner = method.SKLKNN(deepcopy(m))
transfer_learner.source_learner.configs.use_validation = False
transfer_learner.use_all_source = True
#transfer_learner.target_learner = method.SKLLogisticRegression(deepcopy(m))
transfer_learner.target_learner = method.SKLKNN(deepcopy(m))
#learner = method.SKLKNN(deepcopy(m))
#learner = method.SKLLogisticRegression(deepcopy(m))
#learner = method.SKLRidgeClassification()
if use_regression:
learner = method.SKLKNNRegression(deepcopy(m))