BCELoss in PyTorch

Buy Me a Coffee☕

*Memos:

• My post explains L1 Loss(MAE), L2 Loss(MSE), Huber Loss, BCE and Cross Entropy Loss.
• My post explains L1Loss() and MSELoss().
• My post explains HuberLoss().
• My post explains CrossEntropyLoss().

BCELoss() can get the 0D or more D tensor of the zero or more values(float) computed by BCE Loss from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

• The 1st argument for initialization is weight(Optional-Default:None-Type:tensor of int, float or bool). If not given, it's 1.
• There is reduction argument for initialization(Optional-Default:'mean'-Type:str). *'none', 'mean' or 'sum' can be selected.
• There are size_average and reduce argument for initialization but they are deprecated.
• The 1st argument is input(Required-Type:tensor of float). *It must be 0<=x<=1.
• The 2nd argument is target(Required-Type:tensor of float). *It must be 0<=y<=1.
• input and target must be the same size otherwise there is error.
• The empty 1D or more D input and target tensor with reduction='mean' return nan.
• The empty 1D or more D input and target tensor with reduction='sum' return 0..

import torch
from torch import nn

tensor1 = torch.tensor([0.4, 0.8, 0.6, 0.3, 0.0, 0.5])
tensor2 = torch.tensor([0.2, 0.9, 0.4, 0.1, 0.8, 0.5])
                   # -w(y*logx+(1-y)*log(1-x)))
                   # -1(0.2*log0.4+(1-0.2)*log(1-0.4))
                   # ↓↓↓↓↓↓
                   # 0.5919+0.3618+0.7541+0.4414+80.0+0.6931 = 82.8423
                   # 82.8423 / 6 = 13.8071
bceloss = nn.BCELoss()
bceloss(input=tensor1, target=tensor2)
# tensor(7.2500)

bceloss
# BCELoss()

print(bceloss.weight)
# None

bceloss.reduction
# 'mean'

bceloss = nn.BCELoss(weight=None, reduction='mean')
bceloss(input=tensor1, target=tensor2)
# tensor(13.8071)

bceloss = nn.BCELoss(reduction='sum')
bceloss(input=tensor1, target=tensor2)
# tensor(82.8423)

bceloss = nn.BCELoss(reduction='none')
bceloss(input=tensor1, target=tensor2)
# tensor([0.5919, 0.3618, 0.7541, 0.4414, 80.0000, 0.6931])

bceloss = nn.BCELoss(weight=torch.tensor([0., 1., 2., 3., 4., 5.]))
bceloss(input=tensor1, target=tensor2)
# tensor(54.4433)

bceloss = nn.BCELoss(weight=torch.tensor([0.]))
bceloss(input=tensor1, target=tensor2)
# tensor(0.)

bceloss = nn.BCELoss(weight=torch.tensor([0, 1, 2, 3, 4, 5]))
bceloss(input=tensor1, target=tensor2)
# tensor(54.4433)

bceloss = nn.BCELoss(weight=torch.tensor([0]))
bceloss(input=tensor1, target=tensor2)
# tensor(0.)

bceloss = nn.BCELoss(
              weight=torch.tensor([True, False, True, False, True, False])
          )
bceloss(input=tensor1, target=tensor2)
# tensor(13.5577)

bceloss = nn.BCELoss(weight=torch.tensor([False]))
bceloss(input=tensor1, target=tensor2)
# tensor(0.)

tensor1 = torch.tensor([[0.4, 0.8, 0.6], [0.3, 0.0, 0.5]])
tensor2 = torch.tensor([[0.2, 0.9, 0.4], [0.1, 0.8, 0.5]])

bceloss = nn.BCELoss()
bceloss(input=tensor1, target=tensor2)
# tensor(13.8071)

tensor1 = torch.tensor([[[0.4], [0.8], [0.6]], [[0.3], [0.0], [0.5]]])
tensor2 = torch.tensor([[[0.2], [0.9], [0.4]], [[0.1], [0.8], [0.5]]])

bceloss = nn.BCELoss()
bceloss(input=tensor1, target=tensor2)
# tensor(13.8071)

tensor1 = torch.tensor([])
tensor2 = torch.tensor([])

bceloss = nn.BCELoss(reduction='mean')
bceloss(input=tensor1, target=tensor2)
# tensor(nan)

bceloss = nn.BCELoss(reduction='sum')
bceloss(input=tensor1, target=tensor2)
# tensor(0.)