Deep Learning_交叉熵损失函数

二分类交叉熵损失函数

PyTorch提供了两个类来计算二分类交叉熵（Binary Cross Entropy），分别是BCELoss() 和BCEWithLogitsLoss()

该损失函数应用在多标签分类任务中

import numpy as np
import math
import torch
import torch.nn as nn

def sigmoid(x):
    y=1/(1+np.exp(-x))
    return y

def bce_loss1():
    pred=np.array([1.1568,-0.6375,0.5318])
    targ=np.array([0,0,1])
    y=sigmoid(pred)
    #print(y)   [0.76075077 0.34581189 0.62990284]
    y1=-(targ*np.log(y)+(1-targ)*np.log(1-y))
    #print(y1)  [1.43024948 0.42436033 0.4621897 ]
    y2=np.mean(y1)
    print('bce_loss1:',y2)

def bce_loss2():
    input=torch.tensor([[1.1568,-0.6375,0.5318]])
    target=torch.FloatTensor([[0,0,1]])
    
    m=nn.Sigmoid()
    input1=m(input)
    
    loss=nn.BCELoss()
    rlt=loss(input1,target)
    print('bce_loss2:',rlt)

多分类交叉熵损失函数

 pytorch中提供的计算多分类的交叉熵损失函数为nn.CrossEntropyLoss()

import numpy as np
import random
import torch
import torch.nn as nn

def softmax(z):
    t = np.exp(z)
    a = np.exp(z) / np.sum(t, axis=1).reshape(-1,1)
    return a

def softmax_loss1():
    pred=np.array([[1.1568,-0.6375,0.5318]])
    targ=np.array([2])
    
    y=softmax(pred)
    #print(y)    #[[0.58771492 0.09770395 0.31458113]]
    y1=-np.log(y)
    #print(y1)   #[[0.53151328 2.32581328 1.15651328]]
    y2=y1[:,targ[0]]
    print('softmax_loss1:',y2)

def softmax_loss2():
    input=torch.tensor([[1.1568,-0.6375,0.5318]])
    targ=torch.tensor([2])
    
    loss=nn.CrossEntropyLoss()
    rlt=loss(input,targ)
    print('softmax_loss2:',rlt)