自定義數據集 使用pytorch框架實現邏輯回歸并保存模型，然后保存模型后再加載模型進行預測
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optimizer
import matplotlib.pyplot as pltclass1_points = np.array([[2.1, 1.8],[1.9, 2.4],[2.2, 1.2],[1.8, 1.5],[1.3, 1.7],[1.6, 2.1],[1.7, 1.4]])class2_points = np.array([[3.5, 3.4],[3.8, 2.7],[3.4, 2.9],[3.1, 3.6],[3.9, 2.4],[4.0, 2.8],[3.3, 2.5]])x_train = np.concatenate((class1_points, class2_points), axis=0)
y_train = np.concatenate((np.zeros(len(class1_points)), np.ones(len(class2_points))))x_train_tensor = torch.tensor(x_train, dtype=torch.float32)
y_train_tensor = torch.tensor(y_train, dtype=torch.float32)seed = 42
torch.manual_seed(seed)class LogisticRegreModel(nn.Module):def __init__(self):super(LogisticRegreModel, self).__init__()self.fc = nn.Linear(2, 1)def forward(self, x):x = self.fc(x)x = torch.sigmoid(x)return xmodel = LogisticRegreModel()cri = nn.BCELoss()
lr = 0.05
optimizer = optimizer.SGD(model.parameters(), lr=lr)fig, (ax1, ax2) = plt.subplots(1, 2)
epoch_list = []
epoch_loss = []epoches = 1000
for epoch in range(1, epoches + 1):y_pre = model(x_train_tensor)loss = cri(y_pre, y_train_tensor.unsqueeze(1))optimizer.zero_grad()loss.backward()optimizer.step()if epoch % 50 == 0 or epoch == 1:print(f"epoch:{epoch},loss:{loss.item()}")w1, w2 = model.fc.weight.data[0]b = model.fc.bias.data[0]slope = -w1 / w2intercept = -b / w2x_min, x_max = 0, 5x = np.array([x_min, x_max])y = slope * x + interceptax1.clear()ax1.plot(x, y, 'r')ax1.scatter(x_train[:len(class1_points), 0], x_train[:len(class1_points), 1])ax1.scatter(x_train[len(class1_points):, 0], x_train[len(class1_points):, 1])ax2.clear()epoch_list.append(epoch)epoch_loss.append(loss.item())ax2.plot(epoch_list, epoch_loss, 'b')plt.pause(1)
運行結果如下
)
