def fit(self,
theta=None,
x=None,
y=None,
r_xz=None,
t_xz=None,
theta1=None,
batch_size=64,
trainer='adam',
initial_learning_rate=0.001,
final_learning_rate=0.0001,
n_epochs=50,
validation_split=0.2,
early_stopping=True,
alpha=0.01,
learning_curve_folder=None,
learning_curve_filename=None,
**params):
""" Trains ROLR """
logging.info('Training ROLR (ratio regression) with settings:')
logging.info(' theta given: %s', theta is not None)
logging.info(' theta1 given: %s', theta1 is not None)
logging.info(' x given: %s', x is not None)
logging.info(' y given: %s', y is not None)
logging.info(' r_xz given: %s', r_xz is not None)
logging.info(' t_xz given: %s', t_xz is not None)
logging.info(' theta1 given: %s', theta1 is not None)
logging.info(' Samples: %s', x.shape[0])
logging.info(' Parameters: %s', theta.shape[1])
logging.info(' Obserables: %s', x.shape[1])
logging.info(' Batch size: %s', batch_size)
logging.info(' Optimizer: %s', trainer)
logging.info(' Learning rate: %s initially, decaying to %s',
initial_learning_rate, final_learning_rate)
logging.info(' Valid. split: %s', validation_split)
logging.info(' Early stopping: %s', early_stopping)
logging.info(' Epochs: %s', n_epochs)
assert theta is not None
assert x is not None
assert y is not None
assert r_xz is not None
train_model(mode='ratio',
model=self.regressor,
loss_functions=[ratio_mse],
thetas=theta,
xs=x,
ys=y,
r_xzs=r_xz,
batch_size=batch_size,
trainer=trainer,
initial_learning_rate=initial_learning_rate,
final_learning_rate=final_learning_rate,
n_epochs=n_epochs,
validation_split=validation_split,
early_stopping=early_stopping,
learning_curve_folder=learning_curve_folder,
learning_curve_filename=learning_curve_filename)
def fit(self,
theta=None,
x=None,
y=None,
r_xz=None,
t_xz=None,
theta1=None,
batch_size=64,
trainer='adam',
initial_learning_rate=0.001,
final_learning_rate=0.0001,
n_epochs=50,
validation_split=0.2,
early_stopping=True,
alpha=0.01,
learning_curve_folder=None,
learning_curve_filename=None,
**params):
""" Trains SCANDAL """
logging.info('Training SCANDAL (MAF + score) with settings:')
logging.info(' alpha: %s', alpha)
logging.info(' theta given: %s', theta is not None)
logging.info(' x given: %s', x is not None)
logging.info(' y given: %s', y is not None)
logging.info(' r_xz given: %s', r_xz is not None)
logging.info(' t_xz given: %s', t_xz is not None)
logging.info(' theta1 given: %s', theta1 is not None)
logging.info(' Samples: %s', x.shape[0])
logging.info(' Parameters: %s', theta.shape[1])
logging.info(' Obserables: %s', x.shape[1])
logging.info(' Batch size: %s', batch_size)
logging.info(' Optimizer: %s', trainer)
logging.info(' Learning rate: %s initially, decaying to %s',
initial_learning_rate, final_learning_rate)
logging.info(' Valid. split: %s', validation_split)
logging.info(' Early stopping: %s', early_stopping)
logging.info(' Epochs: %s', n_epochs)
assert theta is not None
assert x is not None
assert t_xz is not None
train_model(model=self.maf,
loss_functions=[negative_log_likelihood, score_mse],
loss_weights=[1., alpha],
loss_labels=['nll', 'score'],
thetas=theta,
xs=x,
t_xzs=t_xz,
batch_size=batch_size,
trainer=trainer,
initial_learning_rate=initial_learning_rate,
final_learning_rate=final_learning_rate,
n_epochs=n_epochs,
validation_split=validation_split,
early_stopping=early_stopping,
learning_curve_folder=learning_curve_folder,
learning_curve_filename=learning_curve_filename)
def fit(self,
theta=None,
x=None,
y=None,
r_xz=None,
t_xz=None,
theta1=None,
batch_size=64,
trainer='adam',
initial_learning_rate=0.001,
final_learning_rate=0.0001,
n_epochs=50,
validation_split=0.2,
early_stopping=True,
alpha=0.01,
learning_curve_folder=None,
learning_curve_filename=None,
**params):
""" Trains SCANDAL """
logging.info(
'Training SCANDAL (MAF + score) with ratio control variate with settings:'
)
logging.info(' alpha: %s', alpha)
logging.info(' theta given: %s', theta is not None)
logging.info(' x given: %s', x is not None)
logging.info(' y given: %s', y is not None)
logging.info(' r_xz given: %s', r_xz is not None)
logging.info(' t_xz given: %s', t_xz is not None)
logging.info(' theta1 given: %s', theta1 is not None)
logging.info(' Samples: %s', x.shape[0])
logging.info(' Parameters: %s', theta.shape[1])
logging.info(' Obserables: %s', x.shape[1])
logging.info(' Batch size: %s', batch_size)
logging.info(' Optimizer: %s', trainer)
logging.info(' Learning rate: %s initially, decaying to %s',
initial_learning_rate, final_learning_rate)
logging.info(' Valid. split: %s', validation_split)
logging.info(' Early stopping: %s', early_stopping)
logging.info(' Epochs: %s', n_epochs)
assert theta is not None
assert x is not None
assert t_xz is not None
assert r_xz is not None
assert theta1 is not None
# First, calculate the coefficient for the control variate
cov_r_t = []
var_r = None
for i in range(t_xz.shape[1]):
covariance_matrix = np.cov(1. / r_xz, t_xz[:, i])
var_r = covariance_matrix[0, 0]
cov_r_t.append(covariance_matrix[0, 1])
cov_r_t = np.array(cov_r_t)
cv_coefficients = -cov_r_t / var_r
logging.info('Calculating control variate coefficients')
logging.info(' Variance in 1/r(x,z): %s', var_r)
for i, cov_r_t_i in enumerate(cov_r_t):
logging.info(
' Covariance between t(x,z)_' + str(i) + ', 1/r(x,z): %s',
cov_r_t_i)
for i, cv_ci in enumerate(cv_coefficients):
logging.info(
' CV coefficient for t(x,z)_' + str(i) + ': %s',
cv_ci)
train_model(model=self.maf,
loss_functions=[negative_log_likelihood, score_mse],
loss_weights=[1., alpha],
loss_labels=['nll', 'score'],
thetas=theta,
theta1=theta1,
xs=x,
ys=y,
t_xzs=t_xz,
r_xzs=r_xz,
pre_loss_transformer=t_xz_cv_corrected,
pre_loss_transform_coefficients=cv_coefficients,
batch_size=batch_size,
trainer=trainer,
initial_learning_rate=initial_learning_rate,
final_learning_rate=final_learning_rate,
n_epochs=n_epochs,
validation_split=validation_split,
early_stopping=early_stopping,
learning_curve_folder=learning_curve_folder,
learning_curve_filename=learning_curve_filename)