代码拉取完成,页面将自动刷新
import numpy as np
class LinReg(object):
"""
multivariate linear regression using gradient descent
"""
def __init__(self, learning_rate = 0.01, iterations = 50, verbose = True, l2 = 0,
tolerance = 0, intercept = True):
"""
:param learning_rate: learning rate constant
:param iterations: how many epochs
:param tolerance: the error value in which to stop training
:param intercept: whether to fit an intercept
:param verbose: whether to spit out error rates while training
:param l2: L2 regularization term
"""
self.learning_rate = learning_rate
self.iterations = iterations
self.tolerance = tolerance
self.intercept = intercept
self.verbose = verbose
self.l2 = l2
self.theta = None
self.mean = []
self.std = []
def fit(self, X, y):
"""
Gradient descent, loops over theta and updates to
take steps in direction of steepest decrease of J.
:return: value of theta that minimizes J(theta) and J_history
"""
if self.intercept:
intercept = np.ones((np.shape(X)[0],1))
X = np.concatenate((intercept, X), 1)
num_examples, num_features = np.shape(X)
# initialize theta to 1
self.theta = np.ones(num_features)
for i in range(self.iterations):
# make prediction
predicted = np.dot(X, self.theta.T)
# update theta with gradient descent
self.theta = (self.theta * (1 - (self.learning_rate * self.l2))) - self.learning_rate / num_examples * np.dot((predicted - y).T, X)
# sum of squares cost
error = predicted - y
cost = np.sum(error**2) / (2 * num_examples)
if i % 10 == 0 and self.verbose == True:
print 'iteration:', i
print 'theta:', self.theta
print 'cost:', cost
if cost < self.tolerance:
return self.theta
break
return self.theta
def predict(self, X):
"""
Make linear prediction based on cost and gradient descent
:param X: new data to make predictions on
:return: return prediction
"""
if self.intercept:
intercept = np.ones((np.shape(X)[0],1))
X = np.concatenate((intercept, X), 1)
num_examples, num_features = np.shape(X)
prediction = []
for sample in range(num_examples):
yhat = 0
for value in range(num_features):
yhat += X[sample, value] * self.theta[value]
prediction.append(yhat)
return prediction
def demo():
# initialize linear regression parameters
iterations = 2000
learning_rate = 0.1
l2 = 0.0001
linearReg = LinReg(learning_rate = learning_rate, iterations = iterations, verbose = 1, l2 = l2)
data = np.genfromtxt('Data/blood_pressure.csv', delimiter = ',', skip_header = 1)
X = data[:, 1:]
y = data[:, 0]
# scale data
max = np.amax(X)
X /= max
print X
print y
# fit the linear reg
linearReg.fit(X = X, y = y)
# load testing dataset
test = np.genfromtxt('Data/blood_pressure.csv', delimiter = ',', skip_header = 1)
X_test = test[:, 1:]
y_test = test[:, 0]
max = np.amax(X_test)
X_test /= max
print X_test
predictions = np.array(linearReg.predict(X_test))
print 'correct: ', y_test
print 'prediction: ', predictions
if __name__ == '__main__':
demo()
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。