Skip to content

ll490187880/SoftmaxWithLoss-OHEM

 
 

Repository files navigation

SoftmaxWithLoss-OHEM

SoftmaxWithLoss+OHEM Main idea

  1. Choosing those samples with top loss
  2. Dont backward loss for ignored samples API description
  3. use_use_hard_mining: if it is false, it is a traditional SoftmaxWithLoss
  4. batch_size: how many samples are taken into consideration
  5. hard_ratio: the ratio of hard samples (top most samples in loss) of batch_size, if it is zero, it is just the softmax loss function

About

SoftmaxWithLoss+OHEM

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 60.1%
  • Cuda 39.9%