PyTorch 多层神经网络 (DNN) 训练 HIV 患者状态预测模型

本代码使用 PyTorch 构建多层神经网络模型，通过训练多个模型，最终实现对 HIV 患者状态的预测。代码示例展示了如何读取数据、定义模型、训练模型以及解决训练过程中出现的错误。

import torch
import torch.nn as nn
import torch.optim as optim
import pandas as pd

# 读取 Excel 表格
data = pd.read_excel('C:\Users\lenovo\Desktop\HIV\DNN神经网络测试\data1.xlsx')
x = data.iloc[:, 1:].values  # 取除第一列以外的所有列，即基因的表达量
y = data.iloc[:, 0].values  # 取第一列，即患者状态标志 state

# 定义第一个模型，输出为 8 分类
class Model1(nn.Module):
    def __init__(self):
        super(Model1, self).__init__()
        self.fc1 = nn.Linear(16, 128)
        self.fc2 = nn.Linear(128, 64)
        self.fc3 = nn.Linear(64, 32)
        self.fc4 = nn.Linear(32, 16)
        self.fc5 = nn.Linear(16, 8)
        self.dropout = nn.Dropout(0.2)

    def forward(self, x):
        x = nn.functional.relu(self.fc1(x))
        x = self.dropout(x)
        x = nn.functional.relu(self.fc2(x))
        x = self.dropout(x)
        x = nn.functional.relu(self.fc3(x))
        x = self.dropout(x)
        x = nn.functional.relu(self.fc4(x))
        x = self.dropout(x)
        x = self.fc5(x)
        return x

# 定义第二个模型，输入为第一个模型的 8 分类输出，输出为 4 分类
class Model2(nn.Module):
    def __init__(self):
        super(Model2, self).__init__()
        self.fc1 = nn.Linear(8, 16)
        self.fc2 = nn.Linear(16, 8)
        self.fc3 = nn.Linear(8, 4)
        self.dropout = nn.Dropout(0.2)

    def forward(self, x):
        x = nn.functional.relu(self.fc1(x))
        x = self.dropout(x)
        x = nn.functional.relu(self.fc2(x))
        x = self.dropout(x)
        x = self.fc3(x)
        return x

# 定义第三个模型，第三个模型为二分类模型，输入为第二个模型的 4 分类输出
class Model3(nn.Module):
    def __init__(self):
        super(Model3, self).__init__()
        self.fc1 = nn.Linear(4, 8)
        self.fc2 = nn.Linear(8, 1)
        self.dropout = nn.Dropout(0.2)

    def forward(self, x):
        x = nn.functional.relu(self.fc1(x))
        x = self.dropout(x)
        x = torch.sigmoid(self.fc2(x))
        return x

# 将数据转换为张量
x = torch.tensor(x).float()
y = torch.tensor(y).float().squeeze(0)

# 定义损失函数和优化器
criterion1 = nn.CrossEntropyLoss()
criterion2 = nn.CrossEntropyLoss()
criterion3 = nn.BCELoss()
model1 = Model1()
optimizer1 = optim.Adam(model1.parameters(), lr=0.001)

# 训练第一个模型
for epoch in range(100):
    optimizer1.zero_grad()
    output1 = model1(x)
    loss1 = criterion1(output1, y.long())
    loss1.backward()
    optimizer1.step()
    acc1 = (output1.argmax(dim=1) == y).float().mean()
    print("Epoch [{}/{}], Loss: {:.4f}, Accuracy: {:.4f}".format(epoch+1, 100, loss1.item(), acc1.item()))


# 训练第二个模型
model2 = Model2()
output1 = model1(x)
output2 = model2(output1)
optimizer2 = optim.Adam(model2.parameters(), lr=0.001)
for epoch in range(100):
    optimizer2.zero_grad()
    loss2 = criterion2(output2, output1.argmax(dim=1).long())
    loss2.backward(retain_graph=True)  # 设置 retain_graph 参数为 True
    optimizer2.step()
    acc2 = (output2.argmax(dim=1) == output1.argmax(dim=1)).float().mean()
    print("Epoch [{}/{}], Loss: {:.4f}, Accuracy: {:.4f}".format(epoch+1, 100, loss2.item(), acc2.item()))

# 训练第三个模型
model3 = Model3()
output3 = model3(output2)
optimizer3 = optim.Adam(model3.parameters(), lr=0.001)
for epoch in range(100):
    optimizer3.zero_grad()
    loss3 = criterion3(output3, y)
    loss3.backward()
    optimizer3.step()
    acc3 = ((output3 > 0.5) == y).float().mean()
    print("Epoch [{}/{}], Loss: {:.4f}, Accuracy: {:.4f}".format(epoch+1, 100, loss3.item(), acc3.item()))

解决错误：

在训练第二个模型时，出现了错误：Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward.

这是因为在计算梯度时，计算图被释放了，但在后续的计算中仍然试图访问已经释放的计算图，导致出现错误。

解决方法：

在每次计算梯度时，设置 retain_graph 参数为 True，表示保留计算图，以便后续计算使用。修改后的代码如上所示。

总结：

本代码展示了使用 PyTorch 构建多层神经网络模型并训练预测 HIV 患者状态的步骤。代码还解决了训练过程中常见的错误，并解释了如何通过设置 retain_graph 参数来解决该错误。