无它,关键思想就是梯度的数学原理
梯度下降算法二维代码
import numpy as np
import matplotlib.pyplot as plt
def f(x):
return (x - 3)**2
def grad_f(x):
return 2 * (x - 3)
def gradient_descent(starting_point, learning_rate, num_iterations):
x = starting_point
x_history = [x]
for _ in range(num_iterations):
gradient = grad_f(x)
x = x - learning_rate * gradient
x_history.append(x)
return x, x_history
starting_point = 0 # 初始点
learning_rate = 0.1 # 学习率
num_iterations = 20 # 迭代次数
optimal_x, x_history = gradient_descent(starting_point, learning_rate, num_iterations)
# 绘制目标函数
x_values = np.linspace(-1, 5, 400)
y_values = f(x_values)
plt.figure(figsize=(10, 6))
plt.plot(x_values, y_values, label='Objective Function $(x-3)^2$', color='blue')
plt.scatter(x_history, [f(x) for x in x_history], color='red', label='Gradient Descent Steps')
plt.plot(x_history, [f(x) for x in x_history], linestyle='--', color='red')
plt.axvline(x=3, color='green', linestyle='--', label='Minimum x=3')
plt.xlabel('x')
plt.ylabel('f(x)')
plt.title('Gradient Descent Visualization')
plt.legend()
plt.grid()
plt.show()
梯度下降算法三维代码
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
# 目标函数
def f(x, y):
return (x - 2)**2 + (y - 3)**2
# 目标函数的梯度
def grad_f(x, y):
df_dx = 2 * (x - 2)
df_dy = 2 * (y - 3)
return df_dx, df_dy
# 梯度下降算法
def gradient_descent(starting_point, learning_rate, num_iterations):
x, y = starting_point
x_history, y_history = [x], [y]
for _ in range(num_iterations):
df_dx, df_dy = grad_f(x, y)
x = x - learning_rate * df_dx
y = y - learning_rate * df_dy
x_history.append(x)
y_history.append(y)
return (x, y), x_history, y_history
# 参数设置
starting_point = (0, 0) # 初始点
learning_rate = 0.1 # 学习率
num_iterations = 30 # 迭代次数
# 执行梯度下降
optimal_point, x_history, y_history = gradient_descent(starting_point, learning_rate, num_iterations)
# 绘制目标函数的3D图
x = np.linspace(0, 4, 100)
y = np.linspace(0, 6, 100)
X, Y = np.meshgrid(x, y)
Z = f(X, Y)
fig = plt.figure(figsize=(12, 8))
ax = fig.add_subplot(111, projection='3d')
# 绘制目标函数的表面图
ax.plot_surface(X, Y, Z, cmap='viridis', alpha=0.7)
# 绘制梯度下降的路径
ax.plot(x_history, y_history, f(np.array(x_history), np.array(y_history)), color='red', marker='o')
# 绘制目标函数的最小值位置
ax.scatter(2, 3, f(2, 3), color='green', s=100, label='Minimum (2, 3)')
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('f(X, Y)')
ax.set_title('3D Gradient Descent Visualization')
ax.legend()
plt.show()