def fit(self, X, y, gradient=False, reg_parameter=0):
"""
Currently, only L2 regularization is implemented.
Args:
X (np.ndarray): Training data of shape[n_samples, n_features]
y (np.ndarray): Target values of shape[n_samples, 1]
gradient (bool): Optional use of gradient descent to calculate weights
if False, uses closed form solution to calculate weights.
reg_parameter (float): float to determine strength of regulatrization penalty
if 0, then no linear regression without regularization is performed
Returns: an instance of self
"""
y = np.asarray(y)
X = np.asarray(X)
X = np.column_stack((np.ones(np.shape(X)[0]), X))
if gradient:
self.weights = gradientdescent(X, y, self.grad, reg_param=reg_parameter)
else:
#Calculate weights (closed form solution)
XtX_lambaI = np.dot(np.transpose(X), X) + reg_parameter * np.identity(len(np.dot(np.transpose(X), X)))
self.weights = np.dot(np.linalg.pinv(XtX_lambaI), np.dot(np.transpose(X), y))
self.learned = True
return self
def fit(self, X, y, reg_parameter=0):
"""
Currently, only L2 regularization is implemented.
Args:
X (np.ndarray): Training data of shape[n_samples, n_features]
y (np.ndarray): Target values of shape[n_samples, 1]
reg_parameter (float): float to determine strength of
regulatrization penalty.
if 0, then no linear regression without regularization
is performed
Returns: an instance of self
Raises:
ValueError if y contains values other than 0 and 1
"""
y = np.asarray(y)
if False in np.in1d(y, [0, 1]):
raise NameError('y required to contain only 0 and 1')
X = np.asarray(X)
X = np.column_stack((np.ones(np.shape(X)[0]), X))
self.weights = gradientdescent(X, y, self.grad,
reg_param=reg_parameter)
self.learned = True
return self
def fit(self, X, y, gradient=False, reg_parameter=0):
"""
Currently, only L2 regularization is implemented.
Args:
X (np.ndarray): Training data of shape[n_samples, n_features]
y (np.ndarray): Target values of shape[n_samples, 1]
gradient (bool): Optional use of gradient descent to
calculate weights.
if False, uses closed form solution to calculate weights.
reg_parameter (float): float to determine strength of
regulatrization penalty if 0, then no linear regression
without regularization is performed.
Returns: an instance of self
"""
y = np.asarray(y)
X = np.asarray(X)
X = np.column_stack((np.ones(np.shape(X)[0]), X))
if gradient:
self.weights = gradientdescent(X, y, self.grad,
reg_param=reg_parameter)
else:
# Calculate weights (closed form solution)
XtX_lambaI = np.dot(np.transpose(X), X) + reg_parameter * \
np.identity(len(np.dot(np.transpose(X), X)))
self.weights = np.dot(np.linalg.pinv(XtX_lambaI),
np.dot(np.transpose(X), y))
self.learned = True
return self
def fit(self, X, y, reg_parameter=0):
"""
Currently, only L2 regularization is implemented.
Args:
X (np.ndarray): Training data of shape[n_samples, n_features]
y (np.ndarray): Target values of shape[n_samples, 1]
reg_parameter (float): float to determine strength of
regulatrization penalty.
if 0, then no linear regression without regularization
is performed
Returns: an instance of self
Raises:
ValueError if y contains values other than 0 and 1
"""
y = np.asarray(y)
if False in np.in1d(y, [0, 1]):
raise NameError('y required to contain only 0 and 1')
X = np.asarray(X)
X = np.column_stack((np.ones(np.shape(X)[0]), X))
self.weights = gradientdescent(X, y, self.grad,
reg_param=reg_parameter)
self.learned = True
return self
