基于协同信息的自编码器

import torch
import torch.nn as nn
import torch.optim as optim

class Encoder(nn.Module):
    def __init__(self, input_dim, hidden_dim, output_dim):
        super(Encoder, self).__init__()
        self.fc1 = nn.Linear(input_dim, hidden_dim)
        self.fc2 = nn.Linear(hidden_dim, output_dim)

    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

class Decoder(nn.Module):
    def __init__(self, input_dim, hidden_dim, output_dim):
        super(Decoder, self).__init__()
        self.fc1 = nn.Linear(input_dim, hidden_dim)
        self.fc2 = nn.Linear(hidden_dim, output_dim)

    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

class AutoEncoder(nn.Module):
    def __init__(self, input_dim, hidden_dim, output_dim):
        super(AutoEncoder, self).__init__()
        self.encoder = Encoder(input_dim, hidden_dim, output_dim)
        self.decoder = Decoder(output_dim, hidden_dim, input_dim)

    def forward(self, x):
        x = self.encoder(x)
        x = self.decoder(x)
        return x

# 超参数设置
input_dim = 4096
hidden_dim = 1024
output_dim = 128
learning_rate = 0.001
epochs = 100
batch_size = 32

# 初始化模型、损失函数和优化器
model = AutoEncoder(input_dim, hidden_dim, output_dim)
criterion = nn.MSELoss()
optimizer = optim.Adam(model.parameters(), lr=learning_rate)

# 构造共现性分数矩阵，用于生成对比学习的正负样本
co_occurrence_score = ...

# 训练
for epoch in range(epochs):
    for batch in range(0, len(embeddings), batch_size):
        # 获取batch数据
        batch_embeddings = embeddings[batch:batch + batch_size]

        # 生成对比学习正负样本
        positive_samples, negative_samples = generate_samples(batch_embeddings, co_occurrence_score)

        # 前向传播
        outputs = model(positive_samples)

        # 计算损失
        loss = criterion(outputs, positive_samples) + criterion(model(negative_samples), negative_samples)

        # 反向传播和优化
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

    print(f'Epoch [{epoch + 1}/{epochs}], Loss: {loss.item():.4f}')