Last active
June 3, 2020 21:59
-
-
Save Mehdi-Amine/f79f70d22d1c64af53cd9cc6e9dcdbfb to your computer and use it in GitHub Desktop.
Implementing cross-entropy
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
import torch.nn.functional as F | |
#----------- Implementing the math -----------# | |
def cross_entropy(activations, labels): | |
return - torch.log(activations[range(labels.shape[0]), labels]).mean() | |
zs = torch.tensor([[0.1, 0.4, 0.2], [0.3, 0.9, 0.6]]) # The values of 3 output neurons for 2 instances | |
activations = softmax(zs) # = [[0.2894, 0.3907, 0.3199],[0.2397, 0.4368, 0.3236]] | |
y = torch.tensor([2,0]) # equivalent to [[0,0,1],[1,0,0]] | |
ce = cross_entropy(activations, y) | |
#----------- Using Pytorch autograd -----------# | |
torch_ce = F.cross_entropy(zs, y) | |
#----------- Comparing outputs -----------# | |
print(f"Pytorch cross-entropy: {torch_ce} \nOur cross-entropy: {ce}") | |
''' | |
Out: | |
Pytorch cross-entropy: 1.28411 | |
Our cross-entropy: 1.28411 | |
''' |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment