I'm a bit confused by the cross entropy loss in PyTorch.

Considering this example:

import torch
import torch.nn as nn
from torch.autograd import Variable

output = Variable(torch.FloatTensor([0,0,0,1])).view(1, -1)
target = Variable(torch.LongTensor([3]))

criterion = nn.CrossEntropyLoss()
loss = criterion(output, target)
print(loss)

I would expect the loss to be 0. But I get:

Variable containing:
 0.7437
[torch.FloatTensor of size 1]

As far as I know cross entropy can be calculated like this:

enter image description here

But shouldn't be the result then 1*log(1) = 0 ?

I tried different inputs like one-hot encodings, but this doesn't work at all, so it seems the input shape of the loss function is okay.

I would be really grateful if someone could help me out and tell me where my mistake is.

Thanks in advance!


Solution 1:

In your example you are treating output [0, 0, 0, 1] as probabilities as required by the mathematical definition of cross entropy. But PyTorch treats them as outputs, that don’t need to sum to 1, and need to be first converted into probabilities for which it uses the softmax function.

So H(p, q) becomes:

H(p, softmax(output))

Translating the output [0, 0, 0, 1] into probabilities:

softmax([0, 0, 0, 1]) = [0.1749, 0.1749, 0.1749, 0.4754]

whence:

-log(0.4754) = 0.7437

Solution 2:

Your understanding is correct but pytorch doesn't compute cross entropy in that way. Pytorch uses the following formula.

loss(x, class) = -log(exp(x[class]) / (\sum_j exp(x[j])))
               = -x[class] + log(\sum_j exp(x[j]))

Since, in your scenario, x = [0, 0, 0, 1] and class = 3, if you evaluate the above expression, you would get:

loss(x, class) = -1 + log(exp(0) + exp(0) + exp(0) + exp(1))
               = 0.7437

Pytorch considers natural logarithm.

Solution 3:

I would like to add an important note, as this often leads to confusion.

Softmax is not a loss function, nor is it really an activation function. It has a very specific task: It is used for multi-class classification to normalize the scores for the given classes. By doing so we get probabilities for each class that sum up to 1.

Softmax is combined with Cross-Entropy-Loss to calculate the loss of a model.

Unfortunately, because this combination is so common, it is often abbreviated. Some are using the term Softmax-Loss, whereas PyTorch calls it only Cross-Entropy-Loss.