自编码器提取数据集特征代码

以下是使用PyTorch实现自编码器提取数据集特征的示例代码：

import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from torchvision.datasets import MNIST
from torchvision.transforms import ToTensor
from tqdm import tqdm

# 定义自编码器模型
class Autoencoder(nn.Module):
    def __init__(self):
        super(Autoencoder, self).__init__()
        
        # 编码器
        self.encoder = nn.Sequential(
            nn.Conv2d(1, 16, 3, stride=3, padding=1),  # 16x10x10
            nn.ReLU(inplace=True),
            nn.Conv2d(16, 8, 3, stride=2, padding=1),  # 8x5x5
            nn.ReLU(inplace=True),
            nn.Conv2d(8, 4, 2, stride=1, padding=0),   # 4x4x4
            nn.ReLU(inplace=True),
        )
        
        # 解码器
        self.decoder = nn.Sequential(
            nn.ConvTranspose2d(4, 8, 2, stride=1, padding=0),  # 8x5x5
            nn.ReLU(inplace=True),
            nn.ConvTranspose2d(8, 16, 3, stride=2, padding=1, output_padding=1),  # 16x10x10
            nn.ReLU(inplace=True),
            nn.ConvTranspose2d(16, 1, 3, stride=3, padding=1),  # 1x28x28
            nn.Sigmoid(),
        )
    
    def forward(self, x):
        x = self.encoder(x)
        x = self.decoder(x)
        return x

# 加载MNIST数据集
train_dataset = MNIST(root='data/', train=True, transform=ToTensor(), download=True)
train_loader = DataLoader(train_dataset, batch_size=128, shuffle=True)

# 初始化自编码器模型
model = Autoencoder()

# 定义优化器
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)

# 训练自编码器
num_epochs = 10
for epoch in range(num_epochs):
    for data in tqdm(train_loader):
        img, _ = data
        optimizer.zero_grad()
        output = model(img)
        loss = nn.MSELoss()(output, img)
        loss.backward()
        optimizer.step()

# 提取特征
feature_extractor = nn.Sequential(model.encoder, nn.Flatten())
features = []
with torch.no_grad():
    for data in tqdm(train_loader):
        img, _ = data
        feature = feature_extractor(img)
        features.append(feature)
features = torch.cat(features, dim=0)

这里使用了一个简单的自编码器作为特征提取器，将MNIST手写数字数据集的图像转换为4x4x4的特征向量。训练自编码器时使用均方误差损失函数，优化器为Adam。最后使用训练好的自编码器的编码器部分来提取特征，将特征向量展平为一维向量。