cnn
卷积神经网络CNN的结构一般包含这几个层：
输入层：用于数据的输入
卷积层：使用卷积核进行特征提取和特征映射
激励层：由于卷积也是一种线性运算，因此需要增加非线性映射
池化层：进行下采样，对特征图稀疏处理，减少数据运算量。
全连接层：通常在CNN的尾部进行重新拟合，减少特征信息的损失
输出层：用于输出结果
导包
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.autograd import Variable
换了一台电脑，这台只装了pytorch，没有装Torchvision，重新安装Torchvision
conda install torchvision
D:\document\ML\camp>conda install torchvision
Fetching package metadata ...............
PackageNotFoundError: Packages missing in current channels:
- torchvision
找不到相应的package
换官网上安装命令conda install pytorch-cpu torchvision-cpu -c pytorch，报http请求错误
D:\document\ML\camp>conda install pytorch-cpu torchvision-cpu -c pytorch
Fetching package metadata ...
CondaHTTPError: HTTP 000 CONNECTION FAILED for url <https://conda.anaconda.org/pytorch/win-64/repodata.json>
Elapsed: -
删掉中间的pytorch
D:\document\ML\camp>conda install torchvision-cpu -c pytorch
Fetching package metadata .................
Solving package specifications: .
Package plan for installation in environment C:\ProgramData\Anaconda3:
The following NEW packages will be INSTALLED:
torchvision-cpu: 0.2.2-py_3 pytorch
Proceed ([y]/n)? y
torchvision-cp 100% |###############################| Time: 0:00:01 23.94 kB/s
装好了
加载数据
# MNIST Dataset
train_dataset = datasets.MNIST(root='./data/',
train=True,
transform=transforms.ToTensor(),
download=True)
test_dataset = datasets.MNIST(root='./data/',
train=False,
transform=transforms.ToTensor())
# Data Loader (Input Pipeline)
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)
构造CNN
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
# 输入1通道，输出10通道，kernel 5*5
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
self.mp = nn.MaxPool2d(2)
# fully connect
self.fc = nn.Linear(320, 10)
def forward(self, x):
# in_size = 64
in_size = x.size(0) # one batch
# x: 64*10*12*12
x = F.relu(self.mp(self.conv1(x)))
# x: 64*20*4*4
x = F.relu(self.mp(self.conv2(x)))
# x: 64*320
x = x.view(in_size, -1) # flatten the tensor
# x: 64*10
x = self.fc(x)
return F.log_softmax(x)
model = Net()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
训练并计算损失
def train(epoch):
for batch_idx, (data, target) in enumerate(train_loader):
data, target = Variable(data), Variable(target)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.step()
if batch_idx % 200 == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
def loss():
test_loss = 0
correct = 0
for data, target in test_loader:
data, target = Variable(data, volatile=True), Variable(target)
output = model(data)
# sum up batch loss
test_loss += F.nll_loss(output, target, size_average=False).item()
# get the index of the max log-probability
pred = output.data.max(1, keepdim=True)[1]
correct += pred.eq(target.data.view_as(pred)).cpu().sum()
test_loss /= len(test_loader.dataset)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / len(test_loader.dataset)))
for epoch in range(1, 10):
train(epoch)
loss()
结果
最后得到的准确率为98%， loss: 0.0507
Train Epoch: 1 [0/60000 (0%)] Loss: 2.320769
Train Epoch: 1 [12800/60000 (21%)] Loss: 0.519038
Train Epoch: 1 [25600/60000 (43%)] Loss: 0.335463
Train Epoch: 1 [38400/60000 (64%)] Loss: 0.266052
Train Epoch: 1 [51200/60000 (85%)] Loss: 0.192065
Test set: Average loss: 0.1718, Accuracy: 9496/10000 (94%)
Train Epoch: 2 [0/60000 (0%)] Loss: 0.196151
Train Epoch: 2 [12800/60000 (21%)] Loss: 0.117710
Train Epoch: 2 [25600/60000 (43%)] Loss: 0.112761
Train Epoch: 2 [38400/60000 (64%)] Loss: 0.215878
Train Epoch: 2 [51200/60000 (85%)] Loss: 0.158413
Test set: Average loss: 0.1082, Accuracy: 9683/10000 (96%)
Train Epoch: 3 [0/60000 (0%)] Loss: 0.140935
Train Epoch: 3 [12800/60000 (21%)] Loss: 0.135737
Train Epoch: 3 [25600/60000 (43%)] Loss: 0.092188
Train Epoch: 3 [38400/60000 (64%)] Loss: 0.084430
Train Epoch: 3 [51200/60000 (85%)] Loss: 0.077458
Test set: Average loss: 0.0812, Accuracy: 9763/10000 (97%)
Train Epoch: 4 [0/60000 (0%)] Loss: 0.064760
Train Epoch: 4 [12800/60000 (21%)] Loss: 0.151439
Train Epoch: 4 [25600/60000 (43%)] Loss: 0.113604
Train Epoch: 4 [38400/60000 (64%)] Loss: 0.061374
Train Epoch: 4 [51200/60000 (85%)] Loss: 0.036855
Test set: Average loss: 0.0656, Accuracy: 9807/10000 (98%)
Train Epoch: 5 [0/60000 (0%)] Loss: 0.159602
Train Epoch: 5 [12800/60000 (21%)] Loss: 0.054143
Train Epoch: 5 [25600/60000 (43%)] Loss: 0.112333
Train Epoch: 5 [38400/60000 (64%)] Loss: 0.163274
Train Epoch: 5 [51200/60000 (85%)] Loss: 0.067363
Test set: Average loss: 0.0667, Accuracy: 9796/10000 (97%)
Train Epoch: 6 [0/60000 (0%)] Loss: 0.092683
Train Epoch: 6 [12800/60000 (21%)] Loss: 0.111712
Train Epoch: 6 [25600/60000 (43%)] Loss: 0.053559
Train Epoch: 6 [38400/60000 (64%)] Loss: 0.033269
Train Epoch: 6 [51200/60000 (85%)] Loss: 0.048830
Test set: Average loss: 0.0587, Accuracy: 9822/10000 (98%)
Train Epoch: 7 [0/60000 (0%)] Loss: 0.043137
Train Epoch: 7 [12800/60000 (21%)] Loss: 0.034103
Train Epoch: 7 [25600/60000 (43%)] Loss: 0.072622
Train Epoch: 7 [38400/60000 (64%)] Loss: 0.066607
Train Epoch: 7 [51200/60000 (85%)] Loss: 0.041002
Test set: Average loss: 0.0544, Accuracy: 9840/10000 (98%)
Train Epoch: 8 [0/60000 (0%)] Loss: 0.009038
Train Epoch: 8 [12800/60000 (21%)] Loss: 0.059637
Train Epoch: 8 [25600/60000 (43%)] Loss: 0.012170
Train Epoch: 8 [38400/60000 (64%)] Loss: 0.018512
Train Epoch: 8 [51200/60000 (85%)] Loss: 0.049446
Test set: Average loss: 0.0475, Accuracy: 9846/10000 (98%)
Train Epoch: 9 [0/60000 (0%)] Loss: 0.016563
Train Epoch: 9 [12800/60000 (21%)] Loss: 0.068632
Train Epoch: 9 [25600/60000 (43%)] Loss: 0.027334
Train Epoch: 9 [38400/60000 (64%)] Loss: 0.032353
Train Epoch: 9 [51200/60000 (85%)] Loss: 0.021624
Test set: Average loss: 0.0507, Accuracy: 9840/10000 (98%)
Process finished with exit code 0
359
