Torch learning notes (6) -- logistic regression model (self training)

import torch
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt
#  Logical regression （Logistic Regression）
# LR It is a linear binary classification model 
# y = f(wx+b) f(x) = 1/(1+e**-x) f(x) Also known as Sigmoid Function or Logistic function 
# LR effect ： Map input data to [0,1]
#  Process quantity classification adopts “ rounding , Realize overall two classification 
#  Linear regression is the analysis of independent variables x And scalar y The method of the relationship between 
# LR Is the analysis independent variable x And probability y The method of the relationship between 
# LR It can also be generated by variant “ Logarithmic probability regression model ” ln(y/1-y) = wx+b
#  Machine learning steps ：data( collection 、 cleaning 、 Divide 、 Preprocessing )、model、Loss、optim

# LR Case study 
# data
sample_num = torch.tensor([100.], requires_grad=True)
mean_value = 1.7
bias = 1
n_data = torch.normal(sample_num, 2)
x0 = torch.normal(mean_value * n_data, 1) + bias  #  Normal distribution generates 
y0 = torch.zeros(100)
x1 = torch.normal(-mean_value * n_data, 1) + bias
y1 = torch.ones(100)
train_x = torch.cat((x0, x1), 0)  #  Splicing variables 
train_y = torch.cat((y0, y1), 0)


# model
class LR(nn.Module):
    def __init__(self):
        super(LR, self).__init__()
        self.features = nn.Linear(2, 1)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        x = self.features(x)
        x = self.sigmoid(x)


#  Instantiation LR
LR = LR()

# Loss
loss_fn = nn.BCELoss()
# optim
lr = 0.01  #  Learning rate 
optimizer = torch.optim.SGD(LR.parameters(), lr=lr, momentum=0.9)

# train
for iteration in range(1000):
    y_pred = LR(train_x)
    loss = loss_fn(y_pred.squeeze(), train_y)
    loss.backward()
    optimizer.step()
    #  mapping 
    if iteration % 20 == 0:
        mask = y_pred.ge(0.5).float().squeeze()  # 0.5 Classify thresholds 
        correct = (mask == train_y).sum()  #  Number of correct classifications 
        acc = correct.item() / train_y.size(0)  #  precision 

        plt.scatter(x0.data.numpy()[:, 0], x0.data.numpy()[:, 1], c='r', label='class 0')
        plt.scatter(x1.data.numpy()[:, 0], x1.data.numpy()[:, 1], c='b', label='class 1')

        w0, w1 = LR.features.weight[0]
        w0, w1 = float(w0.item()), float(w1.item())
        plot_b = float(LR.features.bias[0].item())
        plot_x = np.arange(-6, 6, 0.1)
        plot_y = (-w0 * plot_x - plot_b) / w1

        plt.xlim(-5, 7)
        plt.xlim(-5, 7)
        plt.plot(plot_x, plot_y)
        plt.legend()
        plt.show()
        plt.pause(0.5)

        if acc >= 0.99: break