def _ll_grad(w, X, X2, y, l2):
"""
Logistic Regression loglikelihood gradient given, the weights w, the data
X, and the labels y.
"""
w1 = w[:X.shape[1]]
w2 = w[X.shape[1]:]
z1 = np.dot(w1, np.transpose(X))
z2 = np.dot(w2, np.transpose(X2))
p1 = invlogit_vect(z1)
p2 = invlogit_vect(z2)
p = np.multiply(p1, p2)
g1 = np.dot(np.transpose(X),
np.multiply(invlogit_vect(-z1),
np.divide(np.subtract(y, p),
np.subtract(1, p))))
g2 = np.dot(np.transpose(X2),
np.multiply(invlogit_vect(-z2),
np.divide(np.subtract(y, p),
np.subtract(1, p))))
g = np.concatenate((g1, g2))
g -= np.multiply(l2, w)
return -1 * g
def predict_proba(self, X, X2):
"""
Returns the probability of class 1 for each x in X.
"""
try:
getattr(self, "intercept1_")
getattr(self, "intercept2_")
getattr(self, "coef1_")
getattr(self, "coef2_")
except AttributeError:
raise RuntimeError(
"You must train classifer before predicting data!")
X = check_array(X)
X2 = check_array(X2)
if self.fit_first_intercept:
X = np.insert(X, 0, 1, axis=1)
if self.fit_second_intercept:
X2 = np.insert(X2, 0, 1, axis=1)
w = np.insert(self.coef1_, 0, self.intercept1_)
w2 = np.insert(self.coef2_, 0, self.intercept2_)
return (invlogit_vect(np.dot(w, np.transpose(X))) *
invlogit_vect(np.dot(w2, np.transpose(X2))))
def _ll_grad(w, X, X2, y, l2):
"""
Logistic Regression loglikelihood gradient given, the weights w, the data
X, and the labels y.
"""
w1 = w[:X.shape[1]]
w2 = w[X.shape[1]:]
z1 = np.dot(w1, np.transpose(X))
z2 = np.dot(w2, np.transpose(X2))
p1 = invlogit_vect(z1)
p2 = invlogit_vect(z2)
p = np.multiply(p1, p2)
g1 = np.dot(np.transpose(X),
np.multiply(invlogit_vect(-z1),
np.divide(np.subtract(y, p),
np.subtract(1, p))))
g2 = np.dot(np.transpose(X2),
np.multiply(invlogit_vect(-z2),
np.divide(np.subtract(y, p),
np.subtract(1, p))))
g = np.concatenate((g1, g2))
g -= np.multiply(l2, w)
return -1 * g
def _ll(w, X, X2, y, l2):
"""
Logistic Regression loglikelihood given, the weights w, the data X, and the
labels y.
"""
w1 = w[:X.shape[1]]
w2 = w[X.shape[1]:]
p1 = invlogit_vect(np.dot(w1, np.transpose(X)))
p2 = invlogit_vect(np.dot(w2, np.transpose(X2)))
p = np.multiply(p1, p2)
ll = -sum(np.multiply(y, np.log(p)))
ll -= sum(np.multiply(1-y, np.log(1-p)))
ll += np.dot(np.divide(l2, 2), np.multiply(w, w))
return ll
def _ll(w, X, X2, y, l2):
"""
Logistic Regression loglikelihood given, the weights w, the data X, and the
labels y.
"""
w1 = w[:X.shape[1]]
w2 = w[X.shape[1]:]
p1 = invlogit_vect(np.dot(w1, np.transpose(X)))
p2 = invlogit_vect(np.dot(w2, np.transpose(X2)))
p = np.multiply(p1, p2)
ll = -sum(np.multiply(y, np.log(p)))
ll -= sum(np.multiply(1-y, np.log(1-p)))
ll += np.dot(np.divide(l2, 2), np.multiply(w, w))
return ll
def _ll_grad(w, X, y, l2):
"""
Logistic Regression loglikelihood gradient given, the weights w, the data
X, and the labels y.
"""
p = invlogit_vect(np.dot(w, np.transpose(X)))
g = np.dot(np.transpose(X), np.subtract(y, p))
g -= np.multiply(l2, w)
return -1 * g
def _ll_grad(w, X, y, l2):
"""
Logistic Regression loglikelihood gradient given, the weights w, the data
X, and the labels y.
"""
p = invlogit_vect(np.dot(w, np.transpose(X)))
g = np.dot(np.transpose(X), np.subtract(y, p))
g -= np.multiply(l2, w)
return -1 * g
def predict_proba(self, X, X2):
"""
Returns the probability of class 1 for each x in X.
"""
try:
getattr(self, "intercept1_")
getattr(self, "intercept2_")
getattr(self, "coef1_")
getattr(self, "coef2_")
except AttributeError:
raise RuntimeError("You must train classifer before predicting data!")
X = check_array(X)
X2 = check_array(X2)
if self.fit_first_intercept:
X = np.insert(X, 0, 1, axis=1)
if self.fit_second_intercept:
X2 = np.insert(X2, 0, 1, axis=1)
w = np.insert(self.coef1_, 0, self.intercept1_)
w2 = np.insert(self.coef2_, 0, self.intercept2_)
return (invlogit_vect(np.dot(w, np.transpose(X))) *
invlogit_vect(np.dot(w2, np.transpose(X2))))
def predict_proba(self, X):
"""
Returns the probability of class 1 for each x in X.
"""
try:
getattr(self, "intercept_")
getattr(self, "coef_")
except AttributeError:
raise RuntimeError("You must train classifer before predicting data!")
X = check_array(X)
if self.fit_intercept:
X = np.insert(X, 0, 1, axis=1)
w = np.insert(self.coef_, 0, self.intercept_)
return invlogit_vect(np.dot(w, np.transpose(X)))
def predict_proba(self, X):
"""
Returns the probability of class 1 for each x in X.
"""
try:
getattr(self, "intercept_")
getattr(self, "coef_")
except AttributeError:
raise RuntimeError(
"You must train classifer before predicting data!")
X = check_array(X)
if self.fit_intercept:
X = np.insert(X, 0, 1, axis=1)
w = np.insert(self.coef_, 0, self.intercept_)
return invlogit_vect(np.dot(w, np.transpose(X)))