1 import torch
2 import torch.nn.functional as F
3
4 # 1prepare dataset
5 x_data = torch.Tensor([[1.0], [2.0], [3.0]])
6 y_data = torch.Tensor([[0], [0], [1]])
7
8
9 # 2design model using class
10 class LogisticRegressionModel(torch.nn.Module):
11 def __init__(self):
12 super(LogisticRegressionModel, self).__init__()
13 self.linear = torch.nn.Linear(1, 1)
14
15 def forward(self, x):
16 #y_pred = F.sigmoid(self.linear(x))
17 y_pred = torch.sigmoid(self.linear(x))
18 return y_pred
19 model = LogisticRegressionModel()
20
21 # 3construct loss and optimizer
22 # 默认情况下，loss会基于element平均，如果size_average=False的话，loss会被累加。
23 # pytorch版本更新，损失函数更改size_average=False为reduction='sum'
24 # BCELoss是CrossEntropyLoss的一个特例，只用于二分类问题，而CrossEntropyLoss可以用于二分类，也可以用于多分类。
25 criterion = torch.nn.BCELoss(reduction='sum')
26 optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
27
28 # 4training cycle forward, backward, update
29 for epoch in range(1000):
30 y_pred = model(x_data)
31 loss = criterion(y_pred, y_data)
32 print(epoch, loss.item())
33
34 optimizer.zero_grad()
35 loss.backward()
36 optimizer.step()
37
38 print('w = ', model.linear.weight.item())
39 print('b = ', model.linear.bias.item())
40
41 x_test = torch.Tensor([[4.0]])
42 y_test = model(x_test)
43 print('y_pred = ', y_test.data)
