二元函数梯度下降寻找最值python示例代码

import numpy as np
import matplotlib.pyplot as plt
import sympy as sp

# 设置 matplotlib 参数以支持中文显示
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False

# 定义函数和梯度计算
x, y = sp.symbols('x y')
f = (x - 1) ** 2 + (y - 2) ** 2
grad_f = sp.derive_by_array(f, [x, y])

def substitute_values(gradient, values):
    """
    对梯度表达式进行代值
    :param gradient: 梯度表达式
    :param values: 变量及其对应值的字典
    :return: 代值后的梯度向量
    """
    return [g.subs(values) for g in gradient]


def plot_contour_and_path(func, path, xlim=(-1, 3), ylim=(0, 4)):
    """
    绘制等高线图和梯度下降路径
    :param func: 目标函数
    :param path: 梯度下降路径
    :param xlim: x 轴范围
    :param ylim: y 轴范围
    """
    x_vals = np.linspace(xlim[0], xlim[1], 400)
    y_vals = np.linspace(ylim[0], ylim[1], 400)
    X, Y = np.meshgrid(x_vals, y_vals)
    Z = np.array([[func.subs({x: X[i, j], y: Y[i, j]}) for j in range(X.shape[1])] for i in range(X.shape[0])])

    plt.contour(X, Y, Z, 20)
    path = np.array(path)
    plt.plot(path[:, 0], path[:, 1], 'ro-', label='Gradient Descent Path')
    plt.scatter([1], [2], color='blue', label='Global Minimum (1, 2)')
    plt.xlim(xlim)
    plt.ylim(ylim)
    plt.xlabel('x轴')
    plt.ylabel('y轴')
    plt.title('二元函数梯度下降')
    plt.legend()
    plt.grid()
    plt.show()


# 梯度下降算法
def gradient_descent(grad, start_point, learning_rate, num_iterations):
    """
    梯度下降算法
    :param grad: 梯度表达式
    :param start_point: 起始点
    :param learning_rate: 学习率
    :param num_iterations: 迭代次数
    :return: 梯度下降路径
    """
    path = [start_point]
    point = start_point
    for _ in range(num_iterations):
        grad_values = substitute_values(grad, {x: point[0], y: point[1]})
        point = [point[0] - learning_rate * grad_values[0],
                 point[1] - learning_rate * grad_values[1]]
        path.append(point)
        print(point)
    return path

# 可视化结果
def plot_gradient_descent_path(path, xlim=(-1, 3), ylim=(0, 4)):
    """
    绘制梯度下降路径
    :param path: 梯度下降路径
    :param xlim: x 轴范围
    :param ylim: y 轴范围
    """
    path = np.array(path)
    plt.plot(path[:, 0], path[:, 1], 'ro-', label='Gradient Descent Path')
    plt.scatter([1], [2], color='blue', label='Global Minimum (1, 2)')
    plt.xlim(xlim)
    plt.ylim(ylim)
    
    # 标注每个点的迭代次数
    for i, (px, py) in enumerate(path):
        plt.text(px, py, f'{i}', fontsize=9, verticalalignment='bottom', horizontalalignment='right')
    
    plt.xlabel('x轴')
    plt.ylabel('y轴')
    plt.title('二元函数梯度下降')
    plt.legend()
    plt.grid()
    plt.show()

# 参数设置
start_point = [2.5, 3.5]
learning_rate = 0.1
num_iterations =100

# 执行梯度下降
path = gradient_descent(grad_f, start_point, learning_rate, num_iterations)

plot_gradient_descent_path(path)

运行结果