笔记: 随机梯度下降参数求解实现

"""
随机梯度下降

描述：每次选取一个样本参与梯度计算
场景：一元线性回归
"""

import numpy as np

np.random.seed(1)
X = 2 * np.random.rand(100, 1)
Y = 4 + 3 * np.random.randn(100, 1)
X = np.c_[np.ones((100, 1)), X]
m = 100  # 样本数
theta = np.random.randn(2, 1)
learning_rate = 0.001
n_epochs = 10000  # 随机梯度下降次数

if __name__ == "__main__":

    for epoch in range(n_epochs):
        for i in range(m):
            random_index = np.random.randint(m)
            xi = X[random_index:random_index+1]
            yi = Y[random_index:random_index+1]
            gradients = xi.T.dot(xi.dot(theta) - yi)
            # 统计更新theta
            theta -= learning_rate * gradients


    print(theta)

[[ 4.65856506]
 [-0.47926501]]
虽然没调参,但这个是真不稳...