pytorch：Logistic回归

import matplotlib.pyplot as plt
from torch import nn, optim
import numpy as np
import torch
from torch.autograd import Variable
# 设置字体为中文
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.rcParams['axes.unicode_minus'] = False
# 读取数据
def readText():
    with open('data.txt', 'r') as f:
        data_list = f.readlines()
        data_list = [i.split('\n')[0] for i in data_list]
        data_list = [i.split(',') for i in data_list]
        data = [(float(i[0]), float(i[1]), int(i[2])) for i in data_list]
        x_data = [[float(i[0]), float(i[1])] for i in data_list]
        y_data = [float(i[2]) for i in data_list]
    return data, x_data, y_data
# 原始数据可视化
def visualize(data):
    x0 = list(filter(lambda x: x[-1] == 0.0, data))  # 找出类别为0的数据集
    x1 = list(filter(lambda x: x[-1] == 1.0, data))  # 找出类别为1的数据集
    plot_x0_0 = [i[0] for i in x0]  # 类别0的x
    plot_x0_1 = [i[1] for i in x0]  # 类别0的y
    plot_x1_0 = [i[0] for i in x1]  # 类别1的x
    plot_x1_1 = [i[1] for i in x1]  # 类别1的y
    plt.plot(plot_x0_0, plot_x0_1, 'ro', label='类别 0')
    plt.plot(plot_x1_0, plot_x1_1, 'bo', label='类别 1')
    plt.legend()  # 显示图例
    plt.title(label='原始数据分布情况')
    plt.show()
def visualize_after(data):
    x0 = list(filter(lambda x: x[-1] == 0.0, data))  # 找出类别为0的数据集
    x1 = list(filter(lambda x: x[-1] == 1.0, data))  # 找出类别为1的数据集
    plot_x0_0 = [i[0] for i in x0]  # 类别0的x
    plot_x0_1 = [i[1] for i in x0]  # 类别0的y
    plot_x1_0 = [i[0] for i in x1]  # 类别1的x
    plot_x1_1 = [i[1] for i in x1]  # 类别1的y
    plt.plot(plot_x0_0, plot_x0_1, 'ro', label='类别 0')
    plt.plot(plot_x1_0, plot_x1_1, 'bo', label='类别 1')
    # 绘制分类函数
    w0, w1 = model.lr.weight[0]
    w0 = w0.data.item()
    w1 = w1.data.item()
    b = model.lr.bias.item()
    plot_x = np.arange(30, 100, 0.1)
    plot_y = (-w0 * plot_x - b) / w1
    plt.plot(plot_x, plot_y, 'yo', label='分类线')
    plt.legend()  # 显示图例
    plt.title(label='分类线可视化')
    plt.show()
# Logistic回归模型
class LogisticRegression(nn.Module):
    def __init__(self):
        super(LogisticRegression, self).__init__()
        self.lr = nn.Linear(2, 1)
        self.sm = nn.Sigmoid()  # 激活函数类型为sigmoid,经过激活函数，值控制在0到1之间
    def forward(self, x):
        x = self.lr(x)
        x = self.sm(x)
        return x
# 主函数
if __name__ == '__main__':
    data, x_data, y_data = readText()  # 读取数据

    x_data = torch.from_numpy(np.array(x_data))
    y_data = torch.from_numpy(np.array(y_data))

    visualize(data)  # 可视化（观察原始数据分布情况）
    # 获取到模型
    model = LogisticRegression()
    if torch.cuda.is_available():
        model.cuda()
    # 损失函数以及梯度下降
    criterion = nn.BCELoss()  # 二分类的损失函数
    optimizer = optim.SGD(model.parameters(), lr=1e-3, momentum=0.9)

    ctn = []
    lo = []
    for epoch in range(50000):
        if torch.cuda.is_available():
            x = Variable(x_data).cuda()
            y = Variable(y_data).cuda()
        else:
            x = Variable(x_data)
            y = Variable(y_data)

        x = torch.tensor(x, dtype=torch.float32)
        out = model(x)
        y = torch.tensor(y, dtype=torch.float32)
        loss = criterion(out, y)
        print_loss = loss.data.item()
        mask = out.ge(0.5).float()  # 大于0.5则输出为1
        correct = (mask == y).sum()  # 统计输出为1的个数
        acc = correct.item() / x.size(0)

        optimizer.zero_grad()  # 梯度清零
        loss.backward()  # 反向传播
        optimizer.step()  # 更新梯度

        ctn.append(epoch+1)
        lo.append(print_loss)

        if (epoch+1) % 1000 == 0:
            print('*'*10)
            print('epoch {}'.format(epoch+1))
            print('loss is {:.4f}'.format(print_loss))
            print('acc is {:.4f}'.format(acc))
    visualize_after(data)
    # 绘制训练次数与损失值之间的关系
    plt.plot(ctn,lo)
    plt.title(label='训练次数与损失值关系')
    plt.xlabel('训练次数')
    plt.ylabel('损失值')
    plt.show()

实验数据：

34.62365962451697,78.0246928153624,0
30.2867107622687,43.89499752400101,0
35.84740876993872,72.90219802708364,0
60.18259938620976,86.3855209546826,1
79.0327360507101,75.3443764369103,1
45.08327747668339,56.3163717815305,0
61.10666453684766,96.51142588489624,1
75.02474556738889,46.55401354116538,1
76.09878670226257,87.42056971926803,1
84.43281996120035,43.53339331072109,1
95.86155507093572,38.22527805795094,0
75.01365838958247,30.60326323428011,0
82.30705337399482,76.48196330235604,1
69.36458875970939,97.71869196188608,1
39.53833914367223,76.03681085115882,0
53.9710521485623,89.20735013750265,1
69.07014406283025,52.74046973016765,1
67.9468554771161746,67.857410673128,0

实验结果：

注意事项：

（1）数据太少，没有找到完整数据。

（2）注意类型转换，float64转为float32，对应代码为：

 x = torch.tensor(x, dtype=torch.float32)
 out = model(x)
 y = torch.tensor(y, dtype=torch.float32)