Scribd
Upload
20 views1 page

LSR

This Python class defines a loss function called LSRLoss for use in neural networks. The loss function calculates the softmax cross entropy loss between predicted outputs and one-hot encoded target classes, and returns the mean loss value.

Uploaded by

manyam29523
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
20 views1 page

LSR

This Python class defines a loss function called LSRLoss for use in neural networks. The loss function calculates the softmax cross entropy loss between predicted outputs and one-hot encoded target classes, and returns the mean loss value.

Uploaded by

manyam29523
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 1

from __future__ import absolute_import

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

class LSRLoss(nn.Module):
def __init__(self, epsilon=0.1):
super(LSRLoss, self).__init__()
self.epsilon = epsilon

def forward(self, inputs, targets):


num_class = inputs.size()[1]
targets = self._class_to_one_hot(targets.data.cpu(), num_class)
targets = Variable(targets.cuda())
outputs = torch.nn.LogSoftmax()(inputs)
loss = - (targets * outputs)
loss = loss.sum(dim=1)
loss = loss.mean(dim=0)
return loss

def _class_to_one_hot(self, targets, num_class):


targets = torch.unsqueeze(targets, 1)
targets_onehot = torch.FloatTensor(targets.size()[0], num_class)
targets_onehot.zero_()
targets_onehot.scatter_(1, targets, 1 - self.epsilon)
targets_onehot.add_(self.epsilon / num_class)
return targets_onehot

You might also like