1.MNIST数据集
MNIST数据集是由0 到9 的手写数字图像构成的。训练图像有6 万张，测试图像有1 万张每一张图片都有对应的标签数字。因此这个测试集就可以作为验证集使用。
MNIST的图像，每张图片是包含28 像素× 28 像素的灰度图像（1 通道），各个像素的取值在0 到255 之间。每张图片都由一个28 ×28 的矩阵表示，每张图片都由一个784 维的向量表示（28*28=784）。
详细介绍参考： http://yann.lecun.com/exdb/mnist/
2.用神经网络做MNIST手写数字识别
模型结构：
模型如图所示，输入二维张量展开成一维。再经过若干次组合的，Linear层和激活函数层，最后返回。
在模型使用时，后面接到交叉熵损失函数上。所以模型的最后一层不做激活。因为本身交叉熵损失函数带有激活功能。
3.代码实现（python+pytorch）
分四个步骤：
第一步：数据集准备和加载；第二步：设计模型；第三步：构建损失函数和优化器；第四步：模型的训练和验证
因pytorc中封装了很多模块。所以我们在实现时，更多的是了解各个模块的功能，以便组合使用。
import torch
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.optim as optim
import torch.nn.functional as F
import matplotlib.pyplot as plt
batch_size = 64
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307),(0.3081)) #两个参数，平均值和标准差
])
train_dataset = datasets.MNIST(
root="../dataset/mnist/",
train= True,
download= True,
transform= transform
)
train_loader = DataLoader(train_dataset,
shuffle = True,
batch_size = batch_size)
test_dataset = datasets.MNIST(
root="../dataset/mnist/",
train=False,
download=True,
transform=transform
)
test_loder = DataLoader(test_dataset,
shuffle = True,
batch_size = batch_size)
class Net(torch.nn.Module):
def __init__(self):
super(Net,self).__init__()
self.linear1 = torch.nn.Linear(784,512)
self.linear2 = torch.nn.Linear(512,256)
self.linear3 = torch.nn.Linear(256,128)
self.linear4 = torch.nn.Linear(128,64)
self.linear5 = torch.nn.Linear(64,10)
def forward(self,x):
x = x.view(-1,784) # 改变张量形状。把输入展开成若干行，784列
x = F.leaky_relu(self.linear1(x))
x = F.leaky_relu(self.linear2(x))
x = F.leaky_relu(self.linear3(x))
x = F.leaky_relu(self.linear4(x))
return self.linear5(x) #最后一层不做激活，因为下一步输入到交叉损失函数中，交叉熵包含了激活层
model = Net()
criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum= 0.5)
def train(epoch):
total = 0
running_loss = 0.0
train_loss = 0.0 #记录每次epoch的损失
accuracy = 0 #记录每次epoch的accuracy
for batch_id, data in enumerate(train_loader,0):
inputs, target = data
optimizer.zero_grad()
# forword + backward + update
outputs = model(inputs)
loss = criterion(outputs, target)
_, predicted = torch.max(outputs.data, dim=1)
accuracy += (predicted == target).sum().item()
total += target.size(0)
loss.backward()
optimizer.step()
running_loss += loss.item()
train_loss = running_loss
#每迭代300次，求一下这三百次迭代的平均
if batch_id % 300 == 299:
print('[%d, %5d] loss: %.3f' %(epoch+1, batch_id+1, running_loss / 300))
running_loss = 0.0
print('第 %d epoch的 Accuracy on train set: %d %%, Loss on train set: %f' % (epoch + 1, 100 * accuracy / total, train_loss))
#返回acc和loss
return 1.0 * accuracy / total, train_loss
def validation(epoch):
correct = 0
total = 0
val_loss = 0.0
with torch.no_grad():
for data in test_loder:
images, target = data
outputs = model(images)
loss = criterion(outputs, target)
val_loss += loss.item()
_, predicted = torch.max(outputs.data, dim=1)
total += target.size(0)
correct += (predicted == target).sum().item()
print('第 %d epoch的 Accuracy on validation set: %d %%, Loss on validation set: %f' %(epoch+1,100*correct / total, val_loss))
#返回acc和loss
return 1.0 * correct / total, val_loss
#pytorch绘制loss和accuracy曲线
def draw_fig(list,name,name2,epoch):
# 我这里迭代了200次，所以x的取值范围为(0，200)，然后再将每次相对应的准确率以及损失率附在x上
x1 = range(1, epoch+1)
print(x1)
y1 = list
if name=="loss":
plt.cla()
plt.title('Train loss vs. epoch', fontsize=20)
plt.plot(x1, y1, '.-')
plt.xlabel('epoch', fontsize=20)
plt.ylabel('Train loss', fontsize=20)
plt.grid()
str = "./lossAndacc/"+name2+"_loss.png"
plt.savefig(str)
plt.show()
elif name =="acc":
plt.cla()
plt.title('Train accuracy vs. epoch', fontsize=20)
plt.plot(x1, y1, '.-')
plt.xlabel('epoch', fontsize=20)
plt.ylabel('Train accuracy', fontsize=20)
plt.grid()
str2 = "./lossAndacc/" + name2 + "_accuracy.png"
plt.savefig(str2)
plt.show()
def draw_in_one(list,epoch):
# x_axix，train_pn_dis这些都是长度相同的list()
# 开始画图
x_axix = [x for x in range(1, epoch+1)] #把ranage转化为list
train_acc = list[0]
train_loss = list[1]
val_acc = list[2]
val_loss = list[3]
#sub_axix = filter(lambda x: x % 200 == 0, x_axix)
plt.title('Result Analysis')
plt.plot(x_axix, train_acc, color='green', label='training accuracy')
plt.plot(x_axix, train_loss, color='red', label='training loss')
plt.plot(x_axix, val_acc, color='skyblue', label='val accuracy')
plt.plot(x_axix, val_loss, color='blue', label='val loss')
plt.legend() # 显示图例
plt.xlabel('epoch times')
plt.ylabel('rate')
plt.show()
# python 一个折线图绘制多个曲线
if __name__ == '__main__':
train_loss = []
train_acc = []
val_loss = []
val_acc = []
epoches = 10
list = []
for epoch in range(epoches):
acc1, loss1 = train(epoch)
train_loss.append(loss1)
train_acc.append(acc1)
acc2, loss2 = validation(epoch)
val_loss.append(loss2)
val_acc.append(acc2)
#四幅图分开绘制
draw_fig(train_loss, "loss","train", epoches)
draw_fig(train_acc, "acc", "train",epoches)
draw_fig(val_loss, "loss","val", epoches)
draw_fig(val_acc, "acc","val", epoches)
# 四幅图合并绘制
list.append(train_acc)
list.append(train_loss)
list.append(val_acc)
list.append(val_loss)
draw_in_one(list, epoches)
结果：
train acc
train loss
val acc
注：图的title代码中有误。读者自行更改
val loss
四幅图合并绘制
在计算这四个值时，代码可能有点小错误。导致画的图不很准确。读者发现后，自行更改吧
控制台输出内容：
E:\anaconda3\envs\pytorch\python.exe D:/PycharmProjects/pytorchProject/手写数字识别.py [1, 300] loss: 2.211 [1, 600] loss: 0.881 [1, 900] loss: 0.439 第 1 epoch的 Accuracy on train set: 65 %, Loss on train set: 14.349343 第 1 epoch的 Accuracy on validation set: 89 %, Loss on validation set: 55.763730 [2, 300] loss: 0.325 [2, 600] loss: 0.284 [2, 900] loss: 0.242 第 2 epoch的 Accuracy on train set: 91 %, Loss on train set: 8.700389 第 2 epoch的 Accuracy on validation set: 93 %, Loss on validation set: 34.062688 [3, 300] loss: 0.199 [3, 600] loss: 0.180 [3, 900] loss: 0.159 第 3 epoch的 Accuracy on train set: 94 %, Loss on train set: 5.356741 第 3 epoch的 Accuracy on validation set: 94 %, Loss on validation set: 25.656663 [4, 300] loss: 0.138 [4, 600] loss: 0.131 [4, 900] loss: 0.117 第 4 epoch的 Accuracy on train set: 96 %, Loss on train set: 4.067950 第 4 epoch的 Accuracy on validation set: 96 %, Loss on validation set: 19.429859 [5, 300] loss: 0.110 [5, 600] loss: 0.093 [5, 900] loss: 0.095 第 5 epoch的 Accuracy on train set: 97 %, Loss on train set: 3.809268 第 5 epoch的 Accuracy on validation set: 96 %, Loss on validation set: 17.569023 [6, 300] loss: 0.080 [6, 600] loss: 0.082 [6, 900] loss: 0.074 第 6 epoch的 Accuracy on train set: 97 %, Loss on train set: 3.285731 第 6 epoch的 Accuracy on validation set: 97 %, Loss on validation set: 14.668039 [7, 300] loss: 0.062 [7, 600] loss: 0.068 [7, 900] loss: 0.064 第 7 epoch的 Accuracy on train set: 98 %, Loss on train set: 2.248924 第 7 epoch的 Accuracy on validation set: 97 %, Loss on validation set: 15.119584 [8, 300] loss: 0.048 [8, 600] loss: 0.055 [8, 900] loss: 0.053 第 8 epoch的 Accuracy on train set: 98 %, Loss on train set: 2.621493 第 8 epoch的 Accuracy on validation set: 97 %, Loss on validation set: 13.119277 [9, 300] loss: 0.042 [9, 600] loss: 0.041 [9, 900] loss: 0.047 第 9 epoch的 Accuracy on train set: 98 %, Loss on train set: 1.698503 第 9 epoch的 Accuracy on validation set: 97 %, Loss on validation set: 13.277307 [10, 300] loss: 0.029 [10, 600] loss: 0.037 [10, 900] loss: 0.040 第 10 epoch的 Accuracy on train set: 98 %, Loss on train set: 1.292258 第 10 epoch的 Accuracy on validation set: 97 %, Loss on validation set: 13.084560 range(1, 11) range(1, 11) range(1, 11) range(1, 11) Process finished with exit code 0
Be First to Comment