homework3——梯度下降和反向传播

使用python实现梯度下降和反向传播

import torch
import numpy as np

x_data = np.random.randint(1, 6, size=(3))
y_data = (x_data*x_data)*3.0 + x_data*2 + 2.0
print("x_data:", x_data, "y_data:", y_data)

w1 = torch.tensor([1.0], requires_grad=True)  # 设置初值
w2 = torch.tensor([1.0], requires_grad=True)
bias = torch.tensor([1.0], requires_grad=True)


def forward(x):
    return (x*x)*w1 + x*w2 + bias

def loss(x, y):    # MSE
    y_pred = forward(x)
    return ((y_pred - y) ** 2).mean()

for epoch in range(100):
    for x, y in zip(x_data, y_data):
        l = loss(x, y)  # 构建计算图
        l.backward()    # 反向计算梯度并保存
        print("grad: ", x, y, w1.grad.item(), w2.grad.item(), bias.grad.item())  # 打印梯度

        # 更新weights和bias
        w1.data = w1.data - 0.001 * w1.grad.data
        w2.data = w2.data - 0.001 * w2.grad.data
        bias.data = bias.data - 0.001 * bias.grad.data

        # 梯度清零
        w1.grad.data.zero_()
        w2.grad.data.zero_()
        bias.grad.data.zero_()

    print("epoch: {}\tl: {}\tw1: {}\tw2: {}\tbias: {}\t".format(epoch, l.item(), w1.item(), w2.item(), bias.item()))