torch模拟sigmoid

收敛效果不好：

  

   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      import torch
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      t=torch.Tensor([1,1,1,1,1,1,1,1,1,1])
     
    
   

    

     

    
    

     

      x=torch.Tensor([-10,-5,-22,-0.5,1,2,3,7,8,90])
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      y= x.sigmoid()
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      print(y)
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      loss_fn = torch.nn.BCELoss()
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      loss=loss_fn(t,y)
     
    
   

    

     

    
    

     

      print(loss)
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      print(x)
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      import torch
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      bbb=torch.clamp(x,-10,10)/20+0.5
     
    
   

    

     

    
    

     

      # bbb = torch.where(x > 10, torch.full_like(x, 10), x)
     
    
   

    

     

    
    

     

      # bbb = torch.where(bbb < -10, torch.full_like(bbb, -10), bbb)
     
    
   

    

     

    
    

     
 
     
    
   

    

     

    
    

     

      print(bbb)
     
    
   

    

     

    
    

     

      loss=loss_fn(t,bbb)
     
    
   

    

     

    
    

     

      print(loss)
     
    
  
 

文章来源: blog.csdn.net，作者：网奇，版权归原作者所有，如需转载，请联系作者。

原文链接：blog.csdn.net/jacke121/article/details/94753894