Loading...
Close
XCurveXCurve
Sidebar

Search tutorials

Get Started

Documentation

Example Projects

XCurve.Losses.AUTKC

We build the model, the loss function, and the optimizer. By default, we use the resnet18 provided by Pytorch. The following class supports the computation of StandardAUTKCLoss: 

  CLASS StandardAUTKCLoss(surrogate='Sq', K=5, epoch_to_paced=5)     [SOURCE]
Parameters:
surrogate (str)

This parameter specifies the surrogate loss, whose options include Sq, Exp, Logit, and Hinge;

K (int)

This parameter specifies the hyperparameter K for the AUTKC loss;

epoch_to_paced (int)

This parameter specifies the number of warm-up epoch for training. By default, we use the CE loss as the warm-up loss.

Example: 
  batch_size, num_class = 32, 10
  loss_fn = StandardAUTKCLoss(surrogate='Sq', K=3, epoch_to_paced=0)
  preds = torch.randn(batch_size, num_class, requires_grad=True)
  targets = torch.empty(batch_size, dtype=torch.long).random_(1)
  loss = loss_fn(preds, targets)
  loss.backward()