什么是loss?
1、计算实际输出的和目标之间的差距
2、为我们更新输出停工一定的依据(反向传播)grad
L1Loss和MSELoss
举个栗子
x:1,2,3 y:1,2,5
计算方式:
L1loss =(0+0+2)/3=0.6
MSE = (0+0+2^2)/3=1.333
import torch
from torch.nn import L1Loss
from torch import nn
#定义一个数组 dtype不能是int要转换为浮点数
inputs = torch.tensor([1,2,3],dtype=torch.float32)
targets = torch.tensor([1,2,5],dtype=torch.float32)
#1个样本 1通道 高宽1*3
inputs = torch.reshape(inputs,(1,1,1,3))
targets = torch.reshape(targets,(1,1,1,3))
loss =L1Loss()
result = loss(inputs,targets)
print(result)
loss_mse = nn.MSELoss()
result_mse =loss_mse(inputs,targets)
print(result_mse)
交叉熵
常用于分类的损失函数
例子:
有一个分类问题
cat,dog,fog
(0 , 1 , 2)one-hot编码
output [0.1,0.2,0.3]
如果target是1 表示dog类,dog的class,也是output,就会等于0.2
代入公式Loss(x.class) = -0.2+log(exp(0.1)+exp(0.2)+exp(0.3))
#交叉熵
input_data = torch.tensor([0.1,0.2,0.3])
target_data = torch.tensor([1])
#batch_size=1 有3个类别
input_data = torch.reshape(input_data,(1,3))
loss_cross = nn.CrossEntropyLoss()
result_cross = loss_cross(input_data,target_data)
print(result_cross)
反向传播
通过生成的grad来调整找到最少的loss
优化器
应用在前一篇博客的网络结构中
import torch.optim
import torchvision.datasets
from torch import nn
from torch.nn import Sequential, Conv2d, MaxPool2d, Flatten, Linear
from torch.utils.data import DataLoader
dataset = torchvision.datasets.CIFAR10("./dataset",train=False,transform=torchvision.transforms.ToTensor(),download=True)
dataloader = DataLoader(dataset,batch_size=1)
#网络结构
class Test(nn.Module):
def __init__(self):
super(Test, self).__init__()
self.module1 = Sequential(
Conv2d(3,32,5,padding=2),
MaxPool2d(2),
Conv2d(32, 32, 5, padding=2),
MaxPool2d(2),
Conv2d(32, 64, 5, padding=2),
MaxPool2d(2),
Flatten(),
Linear(1024,64),
Linear(64,10)
)
def forward(self,x):
x = self.module1(x)
return x
loss = nn.CrossEntropyLoss()
test = Test()
optim = torch.optim.SGD(test.parameters(),lr=0.01)
#迭代20次
for epoch in range(20):
running_loss =0.0
for data in dataloader:
imgs,targets = data
outputs = test(imgs)
result_loss = loss(outputs,targets)
optim.zero_grad()#清除grad
result_loss.backward()
optim.step()
print(result_loss)
running_loss = running_loss +result_loss
print(running_loss)