屏蔽ip地址访问网站,响应式外贸网站价格,旅游业网站建设方案特点,要点营销网站神经网络-损失函数与反向传播 官网损失函数L1Loss MAE 平均MSELoss 平方差CROSSENTROPYLOSS 交叉熵损失注意code 反向传播在debug中的显示code B站小土堆pytorch视频学习
官网
https://pytorch.org/docs/stable/nn.html#loss-functions
损失函数 L1Loss MAE 平均 import to… 神经网络-损失函数与反向传播 官网损失函数L1Loss MAE 平均MSELoss 平方差CROSSENTROPYLOSS 交叉熵损失注意code 反向传播在debug中的显示code B站小土堆pytorch视频学习
官网
https://pytorch.org/docs/stable/nn.html#loss-functions
损失函数 L1Loss MAE 平均 import torchinput torch.tensor([1, 2, 3], dtypefloat)
# target torch.tensor([1, 2, 5], dtypefloat)
target torch.tensor([[[[1, 2, 5]]]], dtypefloat) # shape [1, 1, 1, 3]
input torch.reshape(input, (1,1,1,3))
# target torch.reshape(target, (1,1,1,3))
print(input.shape)
print(target.shape)loss1 torch.nn.L1Loss()
loss2 torch.nn.L1Loss(reductionsum)
result1 loss1(input, target)
print(result1) # tensor(0.6667, dtypetorch.float64)
result2 loss2(input, target)
print(result2) # tensor(2., dtypetorch.float64)MSELoss 平方差 import torchinput torch.tensor([1, 2, 3], dtypefloat)
# target torch.tensor([1, 2, 5], dtypefloat)
target torch.tensor([[[[1, 2, 5]]]], dtypefloat) # shape [1, 1, 1, 3]
input torch.reshape(input, (1,1,1,3))
# target torch.reshape(target, (1,1,1,3))
print(input.shape)
print(target.shape)loss_mse torch.nn.MSELoss(reductionmean)
result_mse loss_mse(input, target)
print(result_mse) # tensor(1.3333, dtypetorch.float64)
loss_mse2 torch.nn.MSELoss(reductionsum)
result_mse2 loss_mse2(input, target)
print(result_mse2) # tensor(4., dtypetorch.float64)CROSSENTROPYLOSS 交叉熵损失
https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss 在神经网络中默认log是以e为底的所以也可以写成ln
注意
根据需求选择对应的loss函数注意loss函数的输入输出shape
code
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWritertest_set torchvision.datasets.CIFAR10(./dataset, trainFalse, transformtorchvision.transforms.ToTensor(),downloadTrue)dataloader DataLoader(test_set, batch_size1)class MySeq(nn.Module):def __init__(self):super(MySeq, self).__init__()self.model1 Sequential(Conv2d(3, 32, kernel_size5, stride1, padding2),MaxPool2d(2),Conv2d(32, 32, kernel_size5, stride1, padding2),MaxPool2d(2),Conv2d(32, 64, kernel_size5, stride1, padding2),MaxPool2d(2),Flatten(),Linear(1024, 64),Linear(64, 10))def forward(self, x):x self.model1(x)return xloss nn.CrossEntropyLoss()
myseq MySeq()
print(myseq)
for data in dataloader:imgs, targets dataprint(imgs.shape)output myseq(imgs)result loss(output, targets)print(result)
反向传播
在debug中的显示
显示在网络结构中每一层的保护属性中都有weight属性梯度属性在weitht属性里面 先找模型结构 在找每一层 在找weight权重梯度在weight权重里面 code
核心代码result_loss.backward() # 要在最后获取 backward函数要挂在通过loss函数计算后的结果上。
# 模型定义、数据加载 同上个代码
for data in dataloader:imgs, targets dataprint(imgs.shape)output myseq(imgs)result_loss loss(output, targets)result_loss.backward() # 要在最后获取print(result_loss)print(result_loss.grad)
