python 共轭梯度下降

import numpy as np
import matplotlib.pyplot as plt

# initial point x0
x0 = np.zeros((4,1))

# positive symmetric matrix H
H = np.array([[150, 20, 90, 101],
           [20, 36, 46, 61],
           [90, 46, 306, 156],
           [101, 61, 156, 245]])

# vector g
g = np.array([8, -5, 1, 6]).reshape(-1, 1)

def f(x):
    return 1/2*np.dot(np.dot(x.T,H),x)
y_data=np.zeros((7,1))

# x←x0
x = x0
# compute the gradient eta
eta = np.dot(H, x)-g
d = -eta
i = 0

epsilon = 1e-4

while not np.linalg.norm(eta, ord=2) <= epsilon:
    i = i+1
    print("iter=", i)
    # compute alpha
    alpha = -(np.dot(eta.T, d)/np.dot(np.dot(d.T, H), d))
    print("alpha:", alpha)
    x = x+alpha*d
    print("x=", x)
    # renew the gradient-更新梯度
    eta = np.dot(H, x)-g
    print("eta=", eta)
    # renew the descent direction-更新下降方向
    d = -eta+(np.dot(np.dot(eta.T, H), d)/np.dot(np.dot(d.T, H), d))*d
    print("d=", d)

print("x=", x)