PyTorch深度学习（B站刘二大爷）第九讲 多分类问题

知识点

Numpy中的交叉熵

import numpy as np
y=np.array([1,0,0])
z=np.array([0.2,0.1,-0.1])
y_pred = np.exp(z)/np.exp(z).sum()
loss = (-y* np.log(y_pred)).sum()
print(loss)

Pytorch中的交叉熵

import torch
y=torch.LongTensor([0]) #标签分类
z=torch.Tensor([[0.2,0.1,-0.1]])
criterion = torch.nn.CrossEntropyLoss()
loss = criterion(z,y)
print(loss)

import torch
criterion = torch.nn.CrossEntropyLoss()
Y = torch.LongTensor([2,0,1]) #大小顺序为 第三个 第一个 第二个

Y_pred1 = torch.Tensor([[0.1,0.2,0.9], #第三个最大 符合2
                        [1.1,0.1,0.2], #第一个最大 符合0
                        [0.2,2.1,0.1]])#第二个最大 符合1
Y_pred2 = torch.Tensor([[0.8,0.2,0.3], #第一个最大 符合1
                        [0.2,0.3,0.5], #第三个最大 符合2
                        [0.2,0.2,0.5]])#第三个最大 符合2

l1=criterion(Y_pred1,Y)
l2=criterion(Y_pred2,Y)

print('Batch Loss1 = ',l1.data,'\n Batch Loss2=',l2.data)

NLLLOSS

https://pytorch.org/docs/stable/generated/torch.nn.NLLLoss.html#torch.nn.NLLLoss

m = torch.nn.LogSoftmax(dim=1)
loss = torch.nn.NLLLoss()
# input is of size N x C = 3 x 5
input = torch.randn(3, 5, requires_grad=True)
# each element in target has to have 0 <= value < C
target = torch.tensor([1, 0, 4])
output = loss(m(input), target)
output.backward()
# 2D loss example (used, for example, with image inputs)
N, C = 5, 4
loss = torch.nn.NLLLoss()
# input is of size N x C x height x width
data = torch.randn(N, 16, 10, 10)
conv = torch.nn.Conv2d(16, C, (3, 3))
m = torch.nn.LogSoftmax(dim=1)
# each element in target has to have 0 <= value < C
target = torch.empty(N, 8, 8, dtype=torch.long).random_(0, C)
output = loss(m(conv(data)), target)
output.backward()

CrossenTropy

https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss

# Example of target with class indices
loss = torch.nn.CrossEntropyLoss()
input = torch.randn(3, 5, requires_grad=True)
target = torch.empty(3, dtype=torch.long).random_(5)
output = loss(input, target)
output.backward()
# Example of target with class probabilities
input = torch.randn(3, 5, requires_grad=True)
target = torch.randn(3, 5).softmax(dim=1)
output = loss(input, target)
output.backward()

MNIST 数据集多分类

导入库

import torch
from torchvision import transforms #图像处理
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F #为了使用激活函数
import torch.optim as optim

数据准备

batch_size =64
transform = transforms.Compose([ #把[]里的东西构成pipline一样处理
    transforms.ToTensor(), #将PILlow图像转为Tensor 图像是{0,1,2,...,255} 转为[0,1] ，把维度28*28变成1*28*28；
    transforms.Normalize((0.1307,),(0.3081,)) #均值 标准差  变成0-1分布；
])

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_loader = DataLoader(test_dataset,
                        shuffle = False,
                        batch_size=batch_size)

模型设计

class Net(torch.nn.Module):
    def __init__(self):
        super(Net,self).__init__()
        self.l1 = torch.nn.Linear(784,512)
        self.l2 = torch.nn.Linear(512,256)
        self.l3 = torch.nn.Linear(256,128)
        self.l4 = torch.nn.Linear(128,64)
        self.l5 = torch.nn.Linear(64,10)
    
    def forward(self,x):    
        x = x.view(-1,784) #784个元素 view改变张量的形状，改成2阶的 -1自动计算N
        x = F.relu(self.l1(x))
        x = F.relu(self.l2(x))
        x = F.relu(self.l3(x))
        x = F.relu(self.l4(x))
        return self.l5(x) # 最后一层不做激活，不进行非线性变换
model = Net()

损失函数和优化器

criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(),lr = 0.01 ,momentum=0.5) #冲量，冲过鞍点和局部最优

训练和测试

def train(epoch): # 每轮循环封装成函数  epoch决定循环多少次   
    running_loss = 0.0
    for batch_idx , data in enumerate(train_loader,0):
        inputs,target = data #输入数据
        optimizer.zero_grad() #优化器归零
        
        #正向+反向+更新
        outputs = model(inputs)  # 获得模型预测结果(64, 10)
        loss = criterion(outputs , target) # 交叉熵代价函数outputs(64,10),target（64）
        loss.backward() 
        optimizer.step()
        
        running_loss += loss.item() #累计的损失
        if batch_idx %300 ==299: # 每300次打印一次平均损失，因为idx是从0开始的，所以%299，而不是300
            print('[%d,%5d] loss:%.3f'% (epoch+1 ,batch_idx+1 , running_loss/300))
            running_loss = 0.0

def test():
    correct =0 # 所有预测正确的样本数
    total = 0 # 所有样本的数量
    with torch.no_grad(): #test里不需要计算梯度
        for data in test_loader:
            images , labels = data 
            outputs= model(images) # 获得预测值
            _,predicted = torch.max(outputs.data , dim=1) #沿着dim=1 （维度1=每一行）找最大值. <_>是最大值（因为用不到，所以这样写），<predicted>是最大值的索引
            total += labels.size(0) # 累加每批次的样本数，以获得一个测试周期所有的样本数 ;labels.size(0) 第0个元素，(N,1)  
            correct += (predicted == labels).sum().item() #predicted==labels 布尔认证，相等就是真(1)，不相等就是假(0)
    print('测试集准确率：%d %%' %(100*correct / total))

if __name__ =='__main__':
    for epoch in range(10):
        train(epoch)
        test()