1.使用PyTorch实现LeNet网络 2.使用实验7的 Fashion MNIST 对LeNet进行训练和测试,优化算法采用SGD或Adam,torch.optim.SGD 或 torch.optim.Adam 3.分别绘制训练和测试的损失函数曲线和分类正确率曲线 4.调节BatchSize、学习率,并依据测试损失曲线的拐点确定最佳模型,保存该模型 5.使用测试集测试所保存模型的性能,以混淆矩阵展示
时间: 2023-06-16 16:04:09 浏览: 239
(python源码)(LeNet网络)使用PyTorch框架来搭建LeNet网络实现分类
好的,这是一个深度学习的任务,需要使用PyTorch实现LeNet网络并对Fashion MNIST数据集进行训练和测试,我来帮您一步一步完成这个任务。
首先,我们需要导入必要的包:torch、torchvision、numpy和matplotlib。
```python
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import numpy as np
import matplotlib.pyplot as plt
```
接下来,我们定义LeNet网络的结构。
```python
class LeNet(nn.Module):
def __init__(self):
super(LeNet, self).__init__()
self.conv1 = nn.Conv2d(1, 6, 5)
self.pool1 = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 5)
self.pool2 = nn.MaxPool2d(2, 2)
self.fc1 = nn.Linear(16 * 4 * 4, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = self.pool1(torch.relu(self.conv1(x)))
x = self.pool2(torch.relu(self.conv2(x)))
x = x.view(-1, 16 * 4 * 4)
x = torch.relu(self.fc1(x))
x = torch.relu(self.fc2(x))
x = self.fc3(x)
return x
```
这里我们定义了一个包含2个卷积层和3个全连接层的LeNet网络,其中第一个卷积层有6个5x5的卷积核,第二个卷积层有16个5x5的卷积核。每个卷积层后面都跟了一个2x2的最大池化层,然后是3个全连接层,分别有120、84和10个神经元。
接下来,我们加载Fashion MNIST数据集,并将其划分为训练集和验证集。
```python
transform = torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize((0.1307,), (0.3081,))
])
train_dataset = torchvision.datasets.FashionMNIST(root='./data', train=True, download=True, transform=transform)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=64, shuffle=True)
val_dataset = torchvision.datasets.FashionMNIST(root='./data', train=False, download=True, transform=transform)
val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=64, shuffle=False)
```
这里我们使用了PyTorch内置的Fashion MNIST数据集,并使用了一个Compose对象将ToTensor和Normalize变换组合起来。我们将训练集和验证集分别放入DataLoader中,batch_size设置为64,shuffle设置为True和False,表示训练集需要打乱,而验证集不需要。
接下来,我们定义优化算法和损失函数。
```python
net = LeNet()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.01)
```
这里我们使用了SGD优化算法和交叉熵损失函数,学习率设置为0.01。
接下来,我们开始训练模型。
```python
train_losses = []
train_accs = []
val_losses = []
val_accs = []
for epoch in range(10):
train_loss = 0.0
train_acc = 0.0
val_loss = 0.0
val_acc = 0.0
net.train()
for i, (inputs, labels) in enumerate(train_loader):
optimizer.zero_grad()
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
train_acc += (predicted == labels).sum().item()
net.eval()
with torch.no_grad():
for inputs, labels in val_loader:
outputs = net(inputs)
loss = criterion(outputs, labels)
val_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
val_acc += (predicted == labels).sum().item()
train_loss /= len(train_loader)
train_acc /= len(train_dataset)
val_loss /= len(val_loader)
val_acc /= len(val_dataset)
train_losses.append(train_loss)
train_accs.append(train_acc)
val_losses.append(val_loss)
val_accs.append(val_acc)
print('Epoch %d: train_loss=%.4f train_acc=%.4f val_loss=%.4f val_acc=%.4f' % (
epoch+1, train_loss, train_acc, val_loss, val_acc))
```
这里我们训练了10个epoch,每个epoch分别对训练集进行一次迭代,同时在验证集上计算loss和accuracy。在每个epoch结束时,我们将训练集和验证集的loss和accuracy记录下来。
最后,我们绘制训练和验证的损失函数曲线和分类正确率曲线。
```python
fig, ax = plt.subplots(1, 2, figsize=(12, 6))
ax[0].plot(train_losses, label='train')
ax[0].plot(val_losses, label='val')
ax[0].set_xlabel('epoch')
ax[0].set_ylabel('loss')
ax[0].set_title('Training and validation loss')
ax[0].legend()
ax[1].plot(train_accs, label='train')
ax[1].plot(val_accs, label='val')
ax[1].set_xlabel('epoch')
ax[1].set_ylabel('accuracy')
ax[1].set_title('Training and validation accuracy')
ax[1].legend()
plt.show()
```
这里我们使用了matplotlib库来绘制图形,包括训练和验证的损失函数曲线和分类正确率曲线。
接下来,我们调节BatchSize和学习率,并依据测试损失曲线的拐点确定最佳模型,并保存该模型。
```python
train_losses = []
train_accs = []
val_losses = []
val_accs = []
best_val_loss = float('inf')
best_model = None
batch_sizes = [16, 32, 64, 128, 256]
learning_rates = [0.001, 0.01, 0.1, 1]
for batch_size in batch_sizes:
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=batch_size, shuffle=False)
for learning_rate in learning_rates:
net = LeNet()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=learning_rate)
for epoch in range(10):
train_loss = 0.0
train_acc = 0.0
val_loss = 0.0
val_acc = 0.0
net.train()
for i, (inputs, labels) in enumerate(train_loader):
optimizer.zero_grad()
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
train_acc += (predicted == labels).sum().item()
net.eval()
with torch.no_grad():
for inputs, labels in val_loader:
outputs = net(inputs)
loss = criterion(outputs, labels)
val_loss += loss.item()
_, predicted = torch.max(outputs.data, 1)
val_acc += (predicted == labels).sum().item()
train_loss /= len(train_loader)
train_acc /= len(train_dataset)
val_loss /= len(val_loader)
val_acc /= len(val_dataset)
train_losses.append(train_loss)
train_accs.append(train_acc)
val_losses.append(val_loss)
val_accs.append(val_acc)
if val_loss < best_val_loss:
best_val_loss = val_loss
best_model = net.state_dict()
print('BatchSize=%d LearningRate=%.3f Epoch %d: train_loss=%.4f train_acc=%.4f val_loss=%.4f val_acc=%.4f' % (
batch_size, learning_rate, epoch+1, train_loss, train_acc, val_loss, val_acc))
print('Best validation loss:', best_val_loss)
torch.save(best_model, 'best_model.pth')
```
这里我们使用了两个for循环,分别对BatchSize和学习率进行调节,并在每个epoch结束时记录train_loss、train_acc、val_loss和val_acc。在每次更新最佳模型时,我们将模型的参数保存下来。
最后,我们使用测试集测试所保存模型的性能,并以混淆矩阵展示。
```python
test_dataset = torchvision.datasets.FashionMNIST(root='./data', train=False, download=True, transform=transform)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=64, shuffle=False)
net = LeNet()
net.load_state_dict(torch.load('best_model.pth'))
net.eval()
with torch.no_grad():
all_predicted = []
all_labels = []
for inputs, labels in test_loader:
outputs = net(inputs)
_, predicted = torch.max(outputs.data, 1)
all_predicted.extend(predicted.numpy())
all_labels.extend(labels.numpy())
all_predicted = np.array(all_predicted)
all_labels = np.array(all_labels)
confusion_matrix = np.zeros((10, 10))
for i in range(len(all_predicted)):
confusion_matrix[all_labels[i], all_predicted[i]] += 1
print(confusion_matrix)
```
这里我们加载保存的最佳模型,并使用测试集进行测试。使用numpy库创建混淆矩阵,将模型预测结果与真实标签进行比对,并将结果输出。
阅读全文