2012 年， AlexNet 横空出世。这个模型的名字来源于论⽂第一作者的姓名 Alex Krizhevsky。AlexNet 使⽤了 8 层卷积神经⽹络，并以很⼤的优势赢得了 ImageNet 2012 图像识别挑战赛冠军。

AlexNet网络结构

Alexnet模型由5个卷积层和3个池化Pooling 层 ，其中还有3个全连接层构成。AlexNet 跟 LeNet 结构类似，但使⽤了更多的卷积层和更⼤的参数空间来拟合⼤规模数据集 ImageNet。它是浅层神经⽹络和深度神经⽹络的分界线。

Crepe

Crepe

AlexNet代码实现

import torch 
import torch.nn as nn
import torch.nn.functional as F

class AlexNet(nn.Module):
    def __init__(self, p):
        super(AlexNet, self).__init__()
        # AlexNet模型由五个卷积层，三个池化层，三个全连接层构成
        self.conv1 = nn.Conv2d(1, 96, kernel_size = (5, 5), stride = 1)
        self.pool1 = nn.MaxPool2d(kernel_size=(2, 2), stride = 2) # 使用池化层减小输入的高和宽
        self.conv2 = nn.Conv2d(96, 256, kernel_size = (5, 5), stride = 1, padding = 2)
        self.pool2 = nn.MaxPool2d(kernel_size = (2, 2), stride = 2)
        self.conv3 = nn.Conv2d(256, 384,kernel_size = (3, 3), stride = 1, padding = 1)
        self.conv4 = nn.Conv2d(384, 384, kernel_size = (3, 3), stride = 1, padding = 1)
        self.conv5 = nn.Conv2d(384, 256, kernel_size = (3, 3), stride = 1, padding = 1)
        # 连续三个卷积层，使用更小的卷积窗口，除最后一个卷积层外，都进一步增大了输出通道数
        self.pool3 = nn.MaxPool2d(kernel_size = (2, 2), stride = 2)
        # 全连接层
        self.dropout = nn.Dropout(p = p)
        self.fc1 = nn.Linear(256*3*3, 1024)
        self.fc2 = nn.Linear(1024, 1024)
        self.fc3 = nn.Linear(1024, 10)

    def forward(self, x):
        x = self.conv1(x)
        x = self.pool1(F.relu(x))
        x = self.conv2(x)
        x = self.pool2(F.relu(x))
        x = self.conv3(x)
        x = F.relu(x)
        x = self.conv4(x)
        x = F.relu(x)
        x = self.conv5(x)
        x = self.pool3(F.relu(x))
        x = torch.flatten(x, 1)
        x = self.fc1(F.relu(x))
        x = self.dropout(x)
        x = self.fc2(F.relu(x))
        x = self.dropout(x)
        x = self.fc3(F.relu(x))

        return x