#導(dǎo)入相關(guān)包
import torch as t
import matplotlib.pyplot as plt
 
#構(gòu)造數(shù)據(jù)
def get_fake_data(batch_size = 8):
 #設(shè)置隨機(jī)種子數(shù)，這樣每次生成的隨機(jī)數(shù)都是一樣的
 t.manual_seed(10)
 #產(chǎn)生隨機(jī)數(shù)據(jù):y = 2*x+3,加上了一些噪聲
 x = t.rand(batch_size,1) * 20
 #randn生成期望為0方差為1的正態(tài)分布隨機(jī)數(shù)
 y = x * 2 + (1 + t.randn(batch_size,1)) * 3 
 return x,y
 
#查看生成數(shù)據(jù)的分布
x,y = get_fake_data()
plt.scatter(x.squeeze().numpy(),y.squeeze().numpy())
 
#線性回歸
 
#隨機(jī)初始化參數(shù)
w = t.rand(1,1)
b = t.zeros(1,1)
#學(xué)習(xí)率
lr = 0.001 
 
for i in range(10000):
 x,y = get_fake_data()
 
 #forward:計(jì)算loss
 y_pred = x.mm(w) + b.expand_as(y)
 
 #均方誤差作為損失函數(shù)
 loss = 0.5 * (y_pred - y)**2 
 loss = loss.sum()
 
 #backward:手動(dòng)計(jì)算梯度
 dloss = 1
 dy_pred = dloss * (y_pred - y)
 dw = x.t().mm(dy_pred)
 db = dy_pred.sum()
 
 #更新參數(shù)
 w.sub_(lr * dw)
 b.sub_(lr * db)
 
 if i%1000 == 0:
 #畫圖
 plt.scatter(x.squeeze().numpy(),y.squeeze().numpy())
 
 x1 = t.arange(0,20).float().view(-1,1)
 y1 = x1.mm(w) + b.expand_as(x1)
 plt.plot(x1.numpy(),y1.numpy()) #predicted
 plt.show()
 #plt.pause(0.5)
 print(w.squeeze(),b.squeeze())

#導(dǎo)入相關(guān)包
import torch as t
from torch.autograd import Variable as V
import matplotlib.pyplot as plt
 
#構(gòu)造數(shù)據(jù)
def get_fake_data(batch_size=8):
 t.manual_seed(10) #設(shè)置隨機(jī)數(shù)種子
 x = t.rand(batch_size,1) * 20
 y = 2 * x +(1 + t.randn(batch_size,1)) * 3
 return x,y
 
#查看產(chǎn)生的x,y的分布是什么樣的
x,y = get_fake_data()
plt.scatter(x.squeeze().numpy(),y.squeeze().numpy())
 
#線性回歸
 
#初始化隨機(jī)參數(shù)
w = V(t.rand(1,1),requires_grad=True)
b = V(t.rand(1,1),requires_grad=True)
lr = 0.001
for i in range(8000):
 x,y = get_fake_data()
 x,y = V(x),V(y)
 y_pred = x * w + b
 loss = 0.5 * (y_pred-y)**2
 loss = loss.sum()
 
 #自動(dòng)計(jì)算梯度
 loss.backward()
 #更新參數(shù)
 w.data.sub_(lr * w.grad.data)
 b.data.sub_(lr * b.grad.data)
 
 #梯度清零,不清零梯度會(huì)累加的
 w.grad.data.zero_()
 b.grad.data.zero_()
 
 if i%1000==0:
 #predicted
 x = t.arange(0,20).float().view(-1,1)
 y = x.mm(w.data) + b.data.expand_as(x)
 plt.plot(x.numpy(),y.numpy())
 
 #true data
 x2,y2 = get_fake_data()
 plt.scatter(x2.numpy(),y2.numpy())
 plt.show()
print(w.data[0],b.data[0])