import torch
import matplotlib.pyplot as plt

import random

def create_data(w, b, data_num): # 数据生成
x = torch.normal(0, 1, (data_num, len(w)))
y = torch.matmul(x, w) + b # 矩阵相乘再加b

noise = torch.normal(0, 0.01, y.shape)  # 为y添加噪声
y += noise

return x, y

num = 500

true_w = torch.tensor([8.1, 2, 2, 4])
true_b = 1.1

X, Y = create_data(true_w, true_b, num)

plt.scatter(X[:, 3], Y, 1) # 画散点图对X取全部的行的第三列，标签Y，点大小

plt.show()

def data_provider(data, label, batchsize): # 每次取batchsize个数据
length = len(label)
indices = list(range(length))

# 这里需要把数据打乱
random.shuffle(indices)

for each in range(0, length, batchsize):
    get_indices = indices[each: each+batchsize]
    get_data = data[get_indices]
    get_label = label[get_indices]

    yield get_data, get_label  # 有存档点的return

batchsize = 16

for batch_x, batch_y in data_provider(X, Y, batchsize):

print(batch_x, batch_y)

break

定义模型

def fun(x, w, b):
pred_y = torch.matmul(x, w) + b
return pred_y

定义loss

def maeLoss(pre_y, y):
return torch.sum(abs(pre_y-y))/len(y)

sgd(梯度下降)

def sgd(paras, lr):
with torch.nograd(): # 这部分代码不计算梯度
for para in paras:
para -= para.grad lr # 不能写成 para = para - paras.grad lr !!!! 这句相当于要创建一个新的para，会导致报错
para.grad.zero() # 将使用过的梯度归零

lr = 0.01
w_0 = torch.normal(0, 0.01, true_w.shape, requires_grad=True)
b_0 = torch.tensor(0.01, requires_grad=True)
print(w_0, b_0)

epochs = 50
for epoch in range(epochs):
data_loss = 0
for batch_x, batch_y in data_provider(X, Y, batchsize):
pred_y = fun(batch_x, w_0, b_0)
loss = maeLoss(pred_y, batch_y)
loss.backward()
sgd([w_0, b_0], lr)
data_loss += loss

print("epoch %03d: loss: %.6f" % (epoch, data_loss))

print(“真实函数值：”, true_w, true_b)
print(“训练得到的函数值：”, w_0, b_0)

idx = 0
plt.plot(X[:, idx].detach().numpy(), X[:, idx].detach().numpy()*w_0[idx].detach().numpy()+b_0.detach().numpy())
plt.scatter(X[:, idx].detach().numpy(), Y, 1)
plt.show()