深度学习作业——使用AE和TSNE对MINIST数据集进行可视化

废话不多，直接上源码

# 使用AE和TSNE对MNIST进行可视化
import torchvision
from torchvision import transforms
import torch
from torch import nn
import matplotlib.pyplot as plt

batch_size = 30
train_dataset = torchvision.datasets.MNIST(root='.data',
                                           train=True,
                                           transform=transforms.ToTensor(),
                                           download=True)
test_dataset = torchvision.datasets.MNIST(root='.data',
                                           train=False,
                                           transform=transforms.ToTensor())
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size,
                                          shuffle=False)
# 定义AE
class DeepAutoEncoder(nn.Module):
    def __int__(self):
        super(DeepAutoEncoder, self).__init__()
        self.encoder = nn.Sequential(
            nn.Conv2d(in_channels=1, out_channels=3, kernel_size=2, stride=1, padding=0, bias=False),
            nn.BatchNorm2d(3),
            nn.LeakyReLU(0.2, inplace=True),

            nn.Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=2, padding=0, bias=False),
            nn.BatchNorm2d(6),
            nn.LeakyReLU(0.2, inplace=True),

            nn.Conv2d(in_channels=6, out_channels=6, kernel_size=3, stride=2, padding=0, bias=False),
            nn.BatchNorm2d(6),
            nn.LeakyReLU(0.2, inplace=True),

            nn.Conv2d(in_channels=6, out_channels=6, kernel_size=4, stride=2, padding=0, bias=False),
            nn.BatchNorm2d(6),
            nn.LeakyReLU(0.2, inplace=True),
        )
        self.decoder = nn.Sequential(
            nn.ConvTranspose2d(in_channels=6, out_channels=6, kernel_size=4, stride=2, bias=False),
            nn.BatchNorm2d(6),
            nn.ReLU(True),

            nn.ConvTranspose2d(in_channels=6, out_channels=6, kernel_size=3, stride=2, bias=False),
            nn.BatchNorm2d(6),
            nn.ReLU(True),

            nn.ConvTranspose2d(in_channels=6, out_channels=3, kernel_size=3, stride=2, bias=False),
            nn.BatchNorm2d(3),
            nn.ReLU(True),

            nn.ConvTranspose2d(in_channels=3, out_channels=1, kernel_size=2, stride=1, bias=False),
            nn.BatchNorm2d(1),

        )
    def forward(self, x):
        x = self.encoder(x)
        print(x.shape)
        x = self.decoder(x)
        return x
AE = DeepAutoEncoder().cuda()
criterion = nn.L1Loss()
learning_rate = 0.001
optimizer = torch.optim.Adam(AE.parameters(), lr=learning_rate)
list_loss = []
num_epoch = 10
total_step = len(train_loader)
for epoch in range(num_epoch):
    totalloss = 0
    for i, (images,_) in enumerate(train_loader):
        images = images.cuda()
        batch_size = images.size(0)
        AE.train()
        fake_image = AE(images)
        loss = criterion(images, fake_image)
        optimizer.zero_grad() # 优化器梯度清零
        optimizer.step()
        totalloss = totalloss + loss.item()
    list_loss.append(totalloss / (i+1))
fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(13, 7))
axes.plot(list_loss, 'k--')
T_images = torch.stack((train_dataset[i][0] for i in range(16)))
show()

有用的话，一键三连哦，码字不易！