def log_likelihood(self):
# Get Probablities
p = af.logistic(self.features.dot(self.w))
# Get Log Likelihood For Each Row of Dataset
loglikelihood = self.labels * np.log(p + 1e-24) + (1 - self.labels) * np.log(1 - p + 1e-24)
# Return Sum
return -1 * loglikelihood.sum() + self.C * pn.penalty[self.penalty](self.w)
def predict(self, X):
features = (X - self.mean_x) / self.std_x
probs = af.logistic(features.dot(self.w))
for p in probs:
if p[0] >= 0.5: p[0] = 1
else: p[0] = 0
return probs
def predict(self, X):
features = (X - self.mean_x) / self.std_x
probs = af.logistic(features.dot(self.w))
for p in probs:
if p[0] >= 0.5: p[0] = 1
else: p[0] = 0
return probs
def log_likelihood(self):
# Get Probablities
p = af.logistic(self.features.dot(self.w))
# Get Log Likelihood For Each Row of Dataset
loglikelihood = self.labels * np.log(p + 1e-24) + (
1 - self.labels) * np.log(1 - p + 1e-24)
# Return Sum
return -1 * loglikelihood.sum() + self.C * pn.penalty[self.penalty](
self.w)
def log_likelihood_gradient(self):
error = self.labels - af.logistic(self.features.dot(self.w))
product = error * self.features
return product.sum(axis=0).reshape(self.w.shape) - (self.C/self.w.size)* pn.penalty[self.penalty](self.w)
def predict_proba(self, X):
features = (X - self.mean_x) / self.std_x
probs = af.logistic(features.dot(self.w))
return probs
def log_likelihood_gradient(self):
error = self.labels - af.logistic(self.features.dot(self.w))
product = error * self.features
return product.sum(axis=0).reshape(self.w.shape) - (
self.C / self.w.size) * pn.penalty[self.penalty](self.w)
def predict_proba(self, X):
features = (X - self.mean_x) / self.std_x
probs = af.logistic(features.dot(self.w))
return probs