def get_predictions(self, target_data):
assert target_data.is_regression
o = self.source_learner.predict(target_data)
is_labeled = target_data.is_labeled
y_s = array_functions.vec_to_2d(o.fu[is_labeled])
y_true = array_functions.vec_to_2d(o.true_y[is_labeled])
return (y_s, y_true)
def get_predictions(self, target_data):
'''
o = self.target_learner.predict_loo(target_data)
o_source = self.source_learner.predict(target_data)
is_labeled = target_data.is_labeled
target_labels = self.configs.target_labels
if self.use_estimated_f:
o = self.target_learner.predict_loo(target_data.get_subset(is_labeled))
if target_data.is_regression:
y_t = array_functions.vec_to_2d(o.fu)
y_s = array_functions.vec_to_2d(o_source.fu[is_labeled])
y_true = array_functions.vec_to_2d(o.true_y)
else:
y_t = o.fu[:,target_labels]
y_s = o_source.fu[:,target_labels]
y_s = y_s[is_labeled,:]
y_true = array_functions.make_label_matrix(o.true_y)[:,target_labels]
y_true = array_functions.try_toarray(y_true)
return (y_t, y_s, y_true)
'''
assert target_data.is_regression
o = self.source_learner.predict(target_data)
is_labeled = target_data.is_labeled
y_s = array_functions.vec_to_2d(o.fu[is_labeled])
y_true = array_functions.vec_to_2d(o.true_y[is_labeled])
return (y_s, y_true)
def combine_predictions(self,x,y_source,y_target):
data = data_lib.Data()
data.x = x
data.is_regression = True
g = self.g_nw.predict(data).fu
a_t = 1 / (1 + g)
b_s = g / (1 + g)
if y_source.ndim > 1:
a_t = array_functions.vec_to_2d(a_t)
b_s = array_functions.vec_to_2d(b_s)
fu = a_t*y_target + b_s * (y_source + self.bias)
else:
fu = np.multiply(a_t, y_target) + np.multiply(b_s, y_source + self.bias)
return fu
def combine_predictions(self, x, y_source, y_target):
data = data_lib.Data()
data.x = x
data.is_regression = True
g = self.g_nw.predict(data).fu
a_t = 1 / (1 + g)
b_s = g / (1 + g)
if y_source.ndim > 1:
a_t = array_functions.vec_to_2d(a_t)
b_s = array_functions.vec_to_2d(b_s)
fu = a_t * y_target + b_s * (y_source + self.bias)
else:
fu = np.multiply(a_t, y_target) + np.multiply(
b_s, y_source + self.bias)
return fu
def viz_features(x, y, domain_ids, feature_names=None, alpha=.1, learner=None):
#y = array_functions.normalize(y)
x = array_functions.vec_to_2d(x)
for i in range(x.shape[1]):
xi = x[:, i]
xi_train = xi
yi = y
ids_i = domain_ids
title = str(i)
density = None
if feature_names is not None:
title = str(i) + ': ' + feature_names[i]
if learner is not None:
xi, yi, ids_i, density = train_on_data(xi, yi, domain_ids, learner)
density = density * 100 + 1
I = array_functions.is_invalid(density)
density[I] = 200
alpha = 1
array_functions.plot_2d_sub(xi,
yi,
alpha=alpha,
title=title,
data_set_ids=ids_i,
sizes=density)
k = 1
array_functions.plot_histogram(xi_train, 100)
k = 1
def train(self, data):
x = data.x
x = array_functions.vec_to_2d(x)
self.model = sm.nonparametric.KDEMultivariate(x,
var_type='c' *
x.shape[1],
bw='cv_ls')
def train(self, data):
I = data.is_train & data.is_labeled
x = data.x[I,:]
n = x.shape[0]
p = x.shape[1]
y = data.y[I]
x_labeled_transform = self.transform.fit_transform(x)
x_all_transform = self.transform.transform(data.x)
x_bias = np.hstack((x_labeled_transform,np.ones((n,1))))
x_all_bias = np.hstack((x_all_transform,np.ones((x_all_transform.shape[0],1))))
O = np.eye(p+1)
O[p,p] = 0
x_L = x_all_bias
if x_L.shape[0] > self.max_n_L:
I_L = np.random.choice(x_L.shape[0], self.max_n_L, replace = False)
x_L = x_L[I_L,:]
L = self.create_laplacian(x_L)
XX = x_bias.T.dot(x_bias)
XLX = x_L.T.dot(L).dot(x_L)
A = XX + self.C*O + self.C2*XLX
v = np.linalg.lstsq(A,x_bias.T.dot(y))
w_anal = array_functions.vec_to_2d(v[0][0:p])
b_anal = v[0][p]
self.w = w_anal
self.b = b_anal
def viz(pc, fig=None, show_histogram=False, show=True):
import create_data_set
from methods import method
source_learner = method.NadarayaWatsonMethod()
target_learner = method.NadarayaWatsonMethod()
#pc = configs_lib.ProjectConfigs()
data = helper_functions.load_object('../' + pc.data_file).data
data.set_train()
source_data = data.get_transfer_subset(pc.source_labels)
source_data.set_target()
target_data= data.get_transfer_subset(pc.target_labels)
target_data.set_target()
source_learner.train_and_test(source_data)
target_learner.train_and_test(target_data)
source_learner.sigma = 10
target_learner.sigma = 10
x = array_functions.vec_to_2d(np.linspace(data.x.min(), data.x.max(), 100))
test_data = data_lib.Data()
test_data.x = x
test_data.is_regression = True
y_s = source_learner.predict(test_data).fu
y_t = target_learner.predict(test_data).fu
#array_functions.plot_line(x,y_t-y_s,pc.data_set,y_axes=np.asarray([-5,5]))
y = y_t-y_s
#y = y - y.mean()
array_functions.plot_line(x,y, title=None ,fig=fig,show=show)
if show_histogram:
array_functions.plot_histogram(data.x,20)
x=1
def make_uniform_data():
X = np.linspace(0, 1, 100)
Y = np.zeros(X.size)
Y[X < .5] = 0
Y[X >= .5] = 1
X = array_functions.vec_to_2d(X)
return X, Y
def train(self, data):
I = data.is_train & data.is_labeled
x = data.x[I, :]
n = x.shape[0]
p = x.shape[1]
y = data.y[I]
x_labeled_transform = self.transform.fit_transform(x)
x_all_transform = self.transform.transform(data.x)
x_bias = np.hstack((x_labeled_transform, np.ones((n, 1))))
x_all_bias = np.hstack(
(x_all_transform, np.ones((x_all_transform.shape[0], 1))))
O = np.eye(p + 1)
O[p, p] = 0
x_L = x_all_bias
if x_L.shape[0] > self.max_n_L:
I_L = np.random.choice(x_L.shape[0], self.max_n_L, replace=False)
x_L = x_L[I_L, :]
L = self.create_laplacian(x_L)
XX = x_bias.T.dot(x_bias)
XLX = x_L.T.dot(L).dot(x_L)
A = XX + self.C * O + self.C2 * XLX
v = np.linalg.lstsq(A, x_bias.T.dot(y))
w_anal = array_functions.vec_to_2d(v[0][0:p])
b_anal = v[0][p]
self.w = w_anal
self.b = b_anal
def plot_g(self):
x = np.linspace(0,1)
x = array_functions.vec_to_2d(x)
g_orig = self.g_learner.predict_g(x)
g = 1 / (1+g_orig)
array_functions.plot_2d(x,g)
pass
def plot_target(self):
x = np.linspace(0,1)
x = array_functions.vec_to_2d(x)
d = data_lib.Data()
d.x = x
d.y = np.nan*np.ones(x.shape[0])
d.is_regression = True
o = self.target_learner.predict(d)
array_functions.plot_2d(x, o.y)
def train_on_data(x,y,domain_ids,learner):
domain_ids = np.squeeze(domain_ids)
data = data_class.Data()
data.is_regression = True
data.x = array_functions.vec_to_2d(x)
data.y = y
data.set_train()
data.set_true_y()
data.set_target()
x_plot = np.zeros((0,1))
y_plot = np.zeros(0)
ids_plot = np.zeros(0)
density_plot = np.zeros(0)
x_test = scipy.linspace(x.min(),x.max(),100)
x_test = array_functions.vec_to_2d(x_test)
data_test = data_class.Data()
data_test.is_regression = True
data_test.x = x_test
data_test.y = np.zeros(x_test.shape[0])
data_test.y[:] = np.nan
from methods import density
kde = density.KDE()
max_n = 200.0
for i in np.unique(domain_ids):
I = domain_ids == i
data_i = data.get_subset(I)
if data_i.n > max_n:
data_i = data_i.rand_sample(max_n/data_i.n)
learner.train_and_test(data_i)
o = learner.predict(data_test)
x_plot = np.vstack((x_plot,x_test))
y_plot = np.hstack((y_plot,o.y))
ids_plot = np.hstack((ids_plot,np.ones(100)*i))
'''
kde.train_and_test(data_i)
dens = kde.predict(data_test)
dens.y = dens.y / dens.y.max()
den_y = dens.y
'''
dens_y = np.ones(data_test.n)
density_plot = np.hstack((density_plot,dens_y))
return x_plot,y_plot,ids_plot,density_plot
def train_on_data(x, y, domain_ids, learner):
domain_ids = np.squeeze(domain_ids)
data = data_class.Data()
data.is_regression = True
data.x = array_functions.vec_to_2d(x)
data.y = y
data.set_train()
data.set_true_y()
data.set_target()
x_plot = np.zeros((0, 1))
y_plot = np.zeros(0)
ids_plot = np.zeros(0)
density_plot = np.zeros(0)
x_test = scipy.linspace(x.min(), x.max(), 100)
x_test = array_functions.vec_to_2d(x_test)
data_test = data_class.Data()
data_test.is_regression = True
data_test.x = x_test
data_test.y = np.zeros(x_test.shape[0])
data_test.y[:] = np.nan
from methods import density
kde = density.KDE()
max_n = 200.0
for i in np.unique(domain_ids):
I = domain_ids == i
data_i = data.get_subset(I)
if data_i.n > max_n:
data_i = data_i.rand_sample(max_n / data_i.n)
learner.train_and_test(data_i)
o = learner.predict(data_test)
x_plot = np.vstack((x_plot, x_test))
y_plot = np.hstack((y_plot, o.y))
ids_plot = np.hstack((ids_plot, np.ones(100) * i))
'''
kde.train_and_test(data_i)
dens = kde.predict(data_test)
dens.y = dens.y / dens.y.max()
den_y = dens.y
'''
dens_y = np.ones(data_test.n)
density_plot = np.hstack((density_plot, dens_y))
return x_plot, y_plot, ids_plot, density_plot
def create_and_save_data(x, y, domain_ids, file):
data = data_class.Data()
data.x = array_functions.vec_to_2d(x)
data.y = y
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
data.data_set_ids = domain_ids
helper_functions.save_object(file, data)
def create_and_save_data(x, y, domain_ids, file):
data = data_class.Data()
data.x = array_functions.vec_to_2d(x)
data.y = y
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
data.data_set_ids = domain_ids
helper_functions.save_object(file, data)
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_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 predict(self, data):
o = self.target_learner.predict(data)
is_target = data.is_target
o_source = self.source_learner.predict(data.get_subset(is_target))
if not data.is_regression:
assert o.fu.ndim == 2
else:
assert np.squeeze(o.fu).ndim == 1
assert np.squeeze(o_source.fu).ndim == 1
o.fu = o.fu.reshape((o.fu.size,1))
o_source.fu = o_source.fu.reshape((o_source.fu.size,1))
for i in range(o.fu.shape[1]):
fu_t = o.fu[is_target,i]
fu_s = o_source.fu[:,i]
if self.g_learner is not None:
pred = self.g_learner.combine_predictions(data.x[is_target,:],fu_s,fu_t)
if data.x.shape[1] == 1:
x = scipy.linspace(data.x.min(),data.x.max(),100)
x = array_functions.vec_to_2d(x)
g = self.g_learner.predict_g(x)
o.x = x
o.g = g
else:
pred = np.multiply(fu_t,1-self.g) + np.multiply(fu_s,self.g)
o.fu[is_target,i] = pred
#o.fu[is_target] = np.multiply(o.fu[is_target],(1-self.g)) + np.multiply(self.g,o_source.fu)
if data.is_regression:
o.y = o.fu
else:
fu = array_functions.replace_invalid(o.fu,0,1)
fu = array_functions.normalize_rows(fu)
o.fu = fu
o.y = fu.argmax(1)
if data.x.shape[1] == 1:
x = array_functions.vec_to_2d(scipy.linspace(data.x.min(),data.x.max(),100))
o.linspace_x = x
o.linspace_g = self.g_learner.predict_g(x)
assert not (np.isnan(o.y)).any()
assert not (np.isnan(o.fu)).any()
return o
def create_concrete(transfer=False):
file = 'concrete/Concrete_Data.csv'
used_field_names, concrete_data = load_csv(file)
data = data_class.Data()
t = ''
if transfer:
feat_ind = 0
domain_ind = (used_field_names == 'age').nonzero()[0][0]
ages = concrete_data[:, domain_ind]
domain_ids = np.zeros(ages.shape)
domain_ids[ages < 10] = 1
domain_ids[(ages >= 10) & (ages <= 28)] = 2
domain_ids[ages > 75] = 3
data.x = concrete_data[:, 0:(concrete_data.shape[1] - 2)]
#0,3,5
#data.x = preprocessing.scale(data.x)
if concrete_num_feats == 1:
data.x = array_functions.vec_to_2d(data.x[:, feat_ind])
t = '-feat=' + str(feat_ind)
elif concrete_num_feats >= data.x.shape[1]:
t = '-' + str(min(data.x.shape[1], concrete_num_feats))
else:
assert False
data.data_set_ids = domain_ids
else:
data.x = concrete_data[:, 0:-1]
data.y = concrete_data[:, -1]
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
viz = False
if viz:
to_use = domain_ids > 0
domain_ids = domain_ids[to_use]
concrete_data = concrete_data[to_use, :]
np.delete(concrete_data, domain_ind, 1)
viz_features(concrete_data, concrete_data[:, -1], domain_ids,
used_field_names)
return
data.x = array_functions.standardize(data.x)
#viz_features(data.x,data.y,data.data_set_ids)
s = concrete_file % t
helper_functions.save_object(s, data)
def create_concrete(transfer=False):
file = "concrete/Concrete_Data.csv"
used_field_names, concrete_data = load_csv(file)
data = data_class.Data()
t = ""
if transfer:
feat_ind = 0
domain_ind = (used_field_names == "age").nonzero()[0][0]
ages = concrete_data[:, domain_ind]
domain_ids = np.zeros(ages.shape)
domain_ids[ages < 10] = 1
domain_ids[(ages >= 10) & (ages <= 28)] = 2
domain_ids[ages > 75] = 3
data.x = concrete_data[:, 0 : (concrete_data.shape[1] - 2)]
# 0,3,5
# data.x = preprocessing.scale(data.x)
if concrete_num_feats == 1:
data.x = array_functions.vec_to_2d(data.x[:, feat_ind])
t = "-feat=" + str(feat_ind)
elif concrete_num_feats >= data.x.shape[1]:
t = "-" + str(min(data.x.shape[1], concrete_num_feats))
else:
assert False
data.data_set_ids = domain_ids
else:
data.x = concrete_data[:, 0:-1]
data.y = concrete_data[:, -1]
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
viz = False
if viz:
to_use = domain_ids > 0
domain_ids = domain_ids[to_use]
concrete_data = concrete_data[to_use, :]
np.delete(concrete_data, domain_ind, 1)
viz_features(concrete_data, concrete_data[:, -1], domain_ids, used_field_names)
return
data.x = array_functions.standardize(data.x)
# viz_features(data.x,data.y,data.data_set_ids)
s = concrete_file % t
helper_functions.save_object(s, data)
def viz_features(x,y,domain_ids,feature_names=None,alpha=.1,learner=None):
#y = array_functions.normalize(y)
x = array_functions.vec_to_2d(x)
for i in range(x.shape[1]):
xi = x[:,i]
xi_train = xi
yi = y
ids_i = domain_ids
title = str(i)
density = None
if feature_names is not None:
title = str(i) + ': ' + feature_names[i]
if learner is not None:
xi,yi,ids_i,density = train_on_data(xi,yi,domain_ids,learner)
density = density*100 + 1
I = array_functions.is_invalid(density)
density[I] = 200
alpha = 1
array_functions.plot_2d_sub(xi,yi,alpha=alpha,title=title,data_set_ids=ids_i,sizes=density)
k = 1
array_functions.plot_histogram(xi_train,100)
k=1
def create_bike_sharing():
file = "bike_sharing/day.csv"
columns = [0] + range(2, 16)
all_field_names = pd.read_csv(file, nrows=1, dtype="string")
all_field_names = np.asarray(all_field_names.keys())
used_field_names = all_field_names[columns]
bike_data = np.loadtxt(file, skiprows=1, delimiter=",", usecols=columns)
domain_ind = used_field_names == "yr"
domain_ids = np.squeeze(bike_data[:, domain_ind])
# inds_to_keep = (used_field_names == 'temp') | (used_field_names == 'atemp')
# bike_data = bike_data[:,inds_to_keep]
# used_field_names = used_field_names[inds_to_keep]
viz = True
to_use = np.asarray([8, 9, 10, 11])
x = bike_data[:, to_use]
used_field_names = used_field_names[to_use]
y = bike_data[:, -1]
if viz:
# learner = make_learner()
learner = None
viz_features(x, y, domain_ids, used_field_names, learner=learner)
field_to_use = 1
x = x[:, field_to_use]
data = data_class.Data()
data.is_regression = True
data.x = array_functions.vec_to_2d(x)
data.x = array_functions.standardize(data.x)
data.y = y
data.y = array_functions.normalize(data.y)
data.set_defaults()
data.data_set_ids = domain_ids
s = bike_file % ("-feat=" + str(field_to_use))
helper_functions.save_object(s, data)
pass
def create_boston_housing(file_dir=""):
boston_data = datasets.load_boston()
data = data_class.Data()
data.x = boston_data.data
data.y = boston_data.target
data.feature_names = list(boston_data.feature_names)
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
s = boston_housing_raw_data_file
x = data.x
y = data.y
if create_transfer_data:
x_ind = 5
domain_ind = 12
domain_ids = np.ones(x.shape[0])
domain_ids = array_functions.bin_data(x[:, domain_ind], num_bins=4)
x = np.delete(x, domain_ind, 1)
# viz_features(x,y,domain_ids,boston_data.feature_names)
data.data_set_ids = domain_ids
if boston_num_feats == 1:
data.x = data.x[:, x_ind]
data.x = array_functions.vec_to_2d(data.x)
s = s % ""
elif boston_num_feats >= data.x.shape[1]:
data.x = array_functions.standardize(data.x)
p = min(boston_num_feats, data.x.shape[1])
s = s % ("-" + str(p))
else:
assert False
else:
s %= ""
if file_dir != "":
s = file_dir + "/" + s
helper_functions.save_object(s, data)
def create_bike_sharing():
file = 'bike_sharing/day.csv'
columns = [0] + range(2, 16)
all_field_names = pd.read_csv(file, nrows=1, dtype='string')
all_field_names = np.asarray(all_field_names.keys())
used_field_names = all_field_names[columns]
bike_data = np.loadtxt(file, skiprows=1, delimiter=',', usecols=columns)
domain_ind = used_field_names == 'yr'
domain_ids = np.squeeze(bike_data[:, domain_ind])
#inds_to_keep = (used_field_names == 'temp') | (used_field_names == 'atemp')
#bike_data = bike_data[:,inds_to_keep]
#used_field_names = used_field_names[inds_to_keep]
viz = True
to_use = np.asarray([8, 9, 10, 11])
x = bike_data[:, to_use]
used_field_names = used_field_names[to_use]
y = bike_data[:, -1]
if viz:
#learner = make_learner()
learner = None
viz_features(x, y, domain_ids, used_field_names, learner=learner)
field_to_use = 1
x = x[:, field_to_use]
data = data_class.Data()
data.is_regression = True
data.x = array_functions.vec_to_2d(x)
data.x = array_functions.standardize(data.x)
data.y = y
data.y = array_functions.normalize(data.y)
data.set_defaults()
data.data_set_ids = domain_ids
s = bike_file % ('-feat=' + str(field_to_use))
helper_functions.save_object(s, data)
pass
def create_boston_housing(file_dir=''):
boston_data = datasets.load_boston()
data = data_class.Data()
data.x = boston_data.data
data.y = boston_data.target
data.feature_names = list(boston_data.feature_names)
data.set_train()
data.set_target()
data.set_true_y()
data.is_regression = True
s = boston_housing_raw_data_file
x = data.x
y = data.y
create_transfer_data = False
create_y_split = True
if create_y_split:
from base import transfer_project_configs as configs_lib
pc = configs_lib.ProjectConfigs()
main_configs = configs_lib.MainConfigs(pc)
learner = main_configs.learner
learner.quiet = True
learner.target_learner[0].quiet = True
learner.source_learner.quiet = True
learner.g_learner.quiet = False
domain_ids = array_functions.bin_data(data.y, num_bins=2)
data.data_set_ids = domain_ids
data.is_train[:] = True
corrs = []
for i in range(x.shape[1]):
corrs.append(scipy.stats.pearsonr(x[:, i], y)[0])
learner.train_and_test(data)
print 'Just playing with data - not meant to save it'
for i, name in enumerate(data.feature_names):
v = learner.g_learner.g[i]
if abs(v) < 1e-6:
v = 0
print name + ': ' + str(v)
exit()
elif create_transfer_data:
x_ind = 5
domain_ind = 12
domain_ids = np.ones(x.shape[0])
domain_ids = array_functions.bin_data(x[:, domain_ind], num_bins=4)
x = np.delete(x, domain_ind, 1)
#viz_features(x,y,domain_ids,boston_data.feature_names)
data.data_set_ids = domain_ids
if boston_num_feats == 1:
data.x = data.x[:, x_ind]
data.x = array_functions.vec_to_2d(data.x)
s = s % ''
elif boston_num_feats >= data.x.shape[1]:
data.x = array_functions.standardize(data.x)
p = min(boston_num_feats, data.x.shape[1])
s = s % ('-' + str(p))
else:
assert False
else:
s %= ''
if file_dir != '':
s = file_dir + '/' + s
helper_functions.save_object(s, data)
def predict(self, data):
x = array_functions.vec_to_2d(data.x)
x = array_functions.vec_to_2d(x)
y = self.model.pdf(x)
o = results.Output(data, y)
return o
