机器学习之多项式回归

多项式回归的实现

// 多项式回归
import numpy as np
import matplotlib.pyplot as plt

np.random.seed(0)
x0 = np.random.random(50)
y = 2 * x0 ** 3 + 1.2 * x0 ** 2 + 1.5


def degreed(x0):
    degreed = 3
    onex = np.c_[np.ones(len(x0))]
    for i in range(1, degreed+1):
        onex = np.c_[onex, x0**i]
    return onex


onex = degreed(x0)
y = np.c_[y]
m, n = onex.shape

#前向传播
def model(theta):
    h = onex.dot(theta)
    return h

#损失函数
def computeCost(h):
    cost = 0.5*np.mean((h-y)**2)
    return cost

#梯度下降
def gradtheta(h):
    d_theta = (1/m)*np.dot(onex.T, h-y)
    return d_theta

#训练模型
def trainModel():
    alpha = 1
    iters = 10000
    Jcost = []
    theta = np.zeros((n, 1))
    for i in range(iters):
        h = model(theta)
        jcost = computeCost(h)
        d_theta = gradtheta(h)
        theta -= alpha*d_theta
        if i % 100 == 0:
            Jcost.append(jcost)
    return theta, Jcost


theta, Jcost = trainModel()
print(theta)

plt.plot(x0, y, 'rx')
x0 = np.sort(x0)
# y0 = theta[0] + theta[1] * x0 + theta[2] * x0 ** 2 + theta[3] * x0 ** 3
onex = degreed(x0)
y0 = model(theta)
plt.plot(x0, y0)
plt.show()

学习讨论群请加QQ群:521364338,扫码进群领取人工智能学习资料,转载标明出处,侵权必究!