tmlr-group / SIGUA

[ICML 2020] "SIGUA: Forgetting May Make Learning with Noisy Labels More Robust"

SIGUA

ICML'20: SIGUA: Forgetting May Make Learning with Noisy Labels More Robust

======== This is the code for the paper: SIGUA: Forgetting May Make Learning with Noisy Labels More Robust

Presented at ICML'20 If you find this code useful in your research then please cite

@InProceedings{pmlr-v119-han20c,
  title = 	 {{SIGUA}: Forgetting May Make Learning with Noisy Labels More Robust},
  author =       {Han, Bo and Niu, Gang and Yu, Xingrui and Yao, Quanming and Xu, Miao and Tsang, Ivor and Sugiyama, Masashi},
  booktitle = 	 {International Conference on Machine Learning},
  pages = 	 {4006--4016},
  year = 	 {2020}
}

Setup

Istall Miniconda3, and then

conda create -f environment.yml 
conda activate pytorch1.0.0_py36

Running SIGUA on benchmark datasets (MNIST, CIFAR-10)

sh scripts/mnist_sigua_sl.sh
sh scripts/mnist_sigua_bc.sh
sh scripts/cifar10_sigua_sl.sh
sh scripts/cifar10_sigua_bc.sh

The other reproducible version

Please check the other reproducible code of SIGUA: https://github.com/yeachan-kr/pytorch-sigua