def svr_space_2D():
from sklearn.svm import SVR
return hp.Obj(SVR)(
C = hp.Float( 0.01, 1000, hp.log_scale ),
gamma = hp.Float( 10**-5, 1000, hp.log_scale ),
)
def gbc_space_3D():
return hp.Obj(GBC)(
max_depth = hp.Int(1, 15 ),
learning_rate = hp.Float( 0.01, 1, hp.log_scale ),
max_features = hp.Float( 0.001, 1 ),
n_estimators = 100,
)
def forest_nll_space_4D(n_estimators = 100):
return hp.Obj(Forest)(
n_estimators = n_estimators,
base_prob = hp.Float(1e-8,1e-1, hp.log_scale),
max_features = hp.Float(0.0001,1),
bootstrap = hp.Enum(True, False),
algo = hp.Enum('RFC','eTreeC'),
return_proba = True,
)
def gbr_space_3D():
from sklearn.ensemble import GradientBoostingRegressor
return hp.Obj(GradientBoostingRegressor)(
max_depth = hp.Int(1, 15 ),
learning_rate = hp.Float( 0.01, 1, hp.log_scale ),
alpha = hp.Float( 0.0001, 1 ),
loss = 'huber',
n_estimators = 100,
)
def forest_regressor_space_3D():
return hp.Obj(Forest)(
n_estimators = 100,
max_features = hp.Float(0.0001,1),
bootstrap = hp.Enum(True, False),
algo = hp.Enum('RFR','eTreeR')
)
def forest_classifier_space_3D(n_estimators = 100):
return hp.Obj(Forest)(
n_estimators = n_estimators,
max_features = hp.Float(0.0001,1),
bootstrap = hp.Enum(True, False),
algo = hp.Enum('RFC','eTreeC')
)
def svc_space_1d():
from sklearn.svm import SVC
return hp.Obj(SVC)(
# C = hp.Float( 0.01, 1000, hp.log_scale ),
C = 1000,
gamma = hp.Float( 10**-5, 1000, hp.log_scale ),
)