python实现梯度下降求解二元二次方程的最小值

 

之前已经有发过求解一元一次的最小值。先理解好那个，才好理解这个。梯度下降不难的。

方程 x**2 + y**2 + 4

import numpy as np
import matplotlib.pyplot as plt


# 定义二元一次方程
def function(x, y):
    return x**2 + y**2 + 4

# 梯度函数
def gradient(x, y):
    return 2 * x, 2 * y

# 梯度下降法
def gradient_descent(learning_rate, max_iterations):
    point = [8, 8]  # 初始点
    trajectory = [point]

    for _ in range(max_iterations):
        grad = gradient(*point)
      
        point = [point[i] - learning_rate * grad[i] for i in range(len(point))]
        trajectory.append(point)

    min_point = trajectory[-1]
    min_value = function(*min_point)
    print(f"最小点：({min_point[0]}, {min_point[1]})")
    print(f"最小值：{min_value}")


    return np.array(trajectory)

# 设置参数
learning_rate = 0.1
max_iterations = 100

# 运行梯度下降法
trajectory = gradient_descent(learning_rate, max_iterations)

# 可视化结果
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')

x = np.linspace(-10, 10, 100)
y = np.linspace(-10, 10, 100)
X, Y = np.meshgrid(x, y)
Z = function(X, Y)

ax.plot_surface(X, Y, Z, cmap='viridis', alpha=0.8)
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('Function Value')

# 绘制梯度下降轨迹
trajectory = np.array(trajectory)
ax.scatter(trajectory[:, 0], trajectory[:, 1], function(trajectory[:, 0], trajectory[:, 1]), color='r')

plt.show()

输出数值结果：

输出可视化结果：