机器学习01-线性回归-python

1 算法推导

2. 梯度下降实现

def batchGradientDescent(X, y , theta , alpha , repeat):
    m = len(X)
    cost = []
    for i in range(0 , repeat):
        hyp = np.dot(X , theta)
        loss = hyp - y
        grad = np.dot(X.T ,loss) / m
        theta = theta - alpha * grad
        cost1 = 0.5 * m * np.sum(np.square(np.dot(X , theta.T) - y))
        cost.append(cost1)
    return theta , cost

3. 矩阵实现

def linear_regression_mat(data,target):
    X = np.mat(data)
    X1 = X.T.dot(X)
    if np.linalg.det(X1) == 0:
        theta = np.linalg.solve(X.T , X).dot( X.T).dot( target)
        print 'dddd'
        return theta
    theta = np.linalg.inv(X1).dot(X.T).dot( target)
    #theta = np.linalg.inv(X1) * (X.T) *  target
    return theta

4. sklearn实现

lr = sklearn.linear_model.LinearRegression()
lr.fit(X_train , y_train)
y_pred = lr.predict(X_test)
print lr.coef_
print lr.intercept_

https://github.com/lmm915815/my_ML_python/tree/master/linear_regression