def fit(self, train, target, alpha=0.0001, max_loop=130, converge=0.001):
m, n = train.shape
self.weights = np.zeros(n)
for k in range(max_loop):
prev_error = mse(self.predict(train), target)
self.print_progress(k, prev_error)
for t in xrange(m):
data_point = train[t]
predict = self.predict(data_point)
error = predict - target[t]
self.weights -= alpha * error * predict * (1.0 - predict) * data_point
if abs(prev_error - mse(self.predict(train), target)) <= converge:
break
return self
def fit(self, train, target, alpha=0.0001, max_loop=130, converge=0.001):
m, n = train.shape
self.weights = np.zeros(n)
for k in range(max_loop):
prev_error = mse(self.predict(train), target)
self.print_progress(k, prev_error)
for t in range(m):
data_point = train[t]
error = self.predict(data_point) - target[t]
self.weights -= alpha * error * data_point
print self.weights
if abs(prev_error - mse(self.predict(train), target)) <= converge:
break
return self
def fit(self, train, target, alpha=0.0001, max_loop=1300, converge=0.0001, beta=10):
m, n = train.shape
self.weights = np.ones(n)
for k in range(max_loop):
prev_error = mse(self.predict(train), target)
self.print_progress(k, prev_error)
predict = self.predict(train)
error = predict - target
self.weights -= alpha * (train.T.dot(error) + beta * self.weights)
return self
def fit(self,
train,
target,
alpha=0.0001,
max_loop=1300,
converge=0.0001,
beta=10):
m, n = train.shape
self.weights = np.ones(n)
for k in range(max_loop):
prev_error = mse(self.predict(train), target)
self.print_progress(k, prev_error)
predict = self.predict(train)
error = predict - target
self.weights -= alpha * (train.T.dot(error) + beta * self.weights)
return self