resnet18模型中加入正则化的代码

可以通过在网络的定义中加入正则化方法来实现。常用的正则化方法包括L1正则化和L2正则化，可以在网络定义中通过调用相关的函数来实现。比如，在PyTorch中，在定义网络时可以在每个卷积层或全连接层后添加一个正则化方法，如下所示：

import torch.nn as nn

class ResNet18(nn.Module):
    def __init__(self, num_classes=1000):
        super(ResNet18, self).__init__()
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.relu1 = nn.ReLU(inplace=True)
        
        # Residual block 1
        self.conv2_1 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2_1 = nn.BatchNorm2d(64)
        self.relu2_1 = nn.ReLU(inplace=True)
        self.conv2_2 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2_2 = nn.BatchNorm2d(64)
        self.relu2_2 = nn.ReLU(inplace=True)        
        self.residual1 = nn.Sequential(
            nn.Conv2d(64, 64, kernel_size=1, stride=1, bias=False),
            nn.BatchNorm2d(64)
        )
        
        # Residual block 2
        self.conv3_1 = nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1, bias=False)
        self.bn3_1 = nn.BatchNorm2d(128)
        self.relu3_1 = nn.ReLU(inplace=True)
        self.conv3_2 = nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn3_2 = nn.BatchNorm2d(128)
        self.relu3_2 = nn.ReLU(inplace=True)
        self.residual2 = nn.Sequential(
            nn.Conv2d(64, 128, kernel_size=1, stride=2, bias=False),
            nn.BatchNorm2d(128)
        )
        
        # Residual block 3
        self.conv4_1 = nn.Conv2d(128, 256, kernel_size=3, stride=2, padding=1, bias=False)
        self.bn4_1 = nn.BatchNorm2d(256)
        self.relu4_1 = nn.ReLU(inplace=True)
        self.conv4_2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn4_2 = nn.BatchNorm2d(256)
        self.relu4_2 = nn.ReLU(inplace=True)
        self.residual3 = nn.Sequential(
            nn.Conv2d(128, 256, kernel_size=1, stride=2, bias=False),
            nn.BatchNorm2d(256)
        )
        
        # Residual block 4
        self.conv5_1 = nn.Conv2d(256, 512, kernel_size=3, stride=2, padding=1, bias=False)
        self.bn5_1 = nn.BatchNorm2d(512)
        self.relu5_1 = nn.ReLU(inplace=True)
        self.conv5_2 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn5_2 = nn.BatchNorm2d(512)
        self.relu5_2 = nn.ReLU(inplace=True)
        self.residual4 = nn.Sequential(
            nn.Conv2d(256, 512, kernel_size=1, stride=2, bias=False),
            nn.BatchNorm2d(512)
        )
        
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(512, num_classes)
        
        self.dropout = nn.Dropout(p=0.5)  # 添加dropout正则化
        self.relu = nn.ReLU(inplace=True)
        
        self.init_weights()
        
    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu1(x)

        # Residual block 1
        residual = self.residual1(x)
        x = self.conv2_1(x)
        x = self.bn2_1(x)
        x = self.relu2_1(x)
        x = self.conv2_2(x)
        x = self.bn2_2(x)
        x += residual
        x = self.relu2_2(x)

        # Residual block 2
        residual = self.residual2(x)
        x = self.conv3_1(x)
        x = self.bn3_1(x)
        x = self.relu3_1(x)
        x = self.conv3_2(x)
        x = self.bn3_2(x)
        x += residual
        x = self.relu3_2(x)

        # Residual block 3
        residual = self.residual3(x)
        x = self.conv4_1(x)
        x = self.bn4_1(x)
        x = self.relu4_1(x)
        x = self.conv4_2(x)
        x = self.bn4_2(x)
        x += residual
        x = self.relu4_2(x)

        # Residual block 4
        residual = self.residual4(x)
        x = self.conv5_1(x)
        x = self.bn5_1(x)
        x = self.relu5_1(x)
        x = self.conv5_2(x)
        x = self.bn5_2(x)
        x += residual
        x = self.relu5_2(x)

        x = self.avgpool(x)
        x = x.view(x.size(0), -1)
        
        x = self.fc(x)
        x = self.dropout(x)  # 加上dropout
        
        return x

    def init_weights(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)
            elif isinstance(m, nn.Linear):
                nn.init.normal_(m.weight, 0, 0.01)
                nn.init.constant_(m.bias, 0)

这样，在每个全连接层后面都会添加一个dropout操作，用于防止过拟合。