Skip to content

Latest commit

 

History

History
76 lines (59 loc) · 3.13 KB

README.md

File metadata and controls

76 lines (59 loc) · 3.13 KB

Self-Adaptive Training for Selective Classification

This repository contains the PyTorch implementation of Selective Classification used in the

Self-adaptive training significantly improves the generalization of deep networks under noise and enhances the self-supervised representation learning. It also advances the state-of-the-art on learning with noisy label, adversarial training and the linear evaluation on the learned representation.

News

  • 2021.10: We have released the code for Selective Classification.
  • 2021.01: We have released the journal version of Self-Adaptive Training, which is a unified algorithm for both the supervised and self-supervised learning. Code for self-supervised learning will be available soon.
  • 2020.09: Our work has been accepted at NeurIPS'2020.

Requirements

  • Python >= 3.6
  • PyTorch >= 1.0
  • CUDA
  • Numpy

Usage

Training and evaluating Selective Classification based on SAT

The main.py contains training and evaluation functions in standard training setting.

Runnable scripts

  • Training and evaluation using the default parameters

    We provide our training scripts in directory train.sh. For a concrete example, we can use the command as below to train the default model (i.e., VGG16-BN) on CIFAR10 dataset:

    $ bash train.sh
  • Additional arguments

    • arch: the architecture of backbone model, e.g., vgg16_bn
    • dataset: the trainiing dataset, e.g., cifar10
    • loss: the loss function for training, e.g., sat
    • sat-momentum: the momentum term of our approach

Results on CIFAR10, SVHN and Dogs&Cats datasets under various coverage rate

Self-Adaptive Training vs. prior state-of-the-arts on the Selective Classification at vaiours coverage rate.

Reference

For technical details, please check the conference version or the journal version of our paper.

@inproceedings{huang2020self,
  title={Self-Adaptive Training: beyond Empirical Risk Minimization},
  author={Huang, Lang and Zhang, Chao and Zhang, Hongyang},
  booktitle={Advances in Neural Information Processing Systems},
  volume={33},
  year={2020}
}

@article{huang2021self,
  title={Self-Adaptive Training: Bridging the Supervised and Self-Supervised Learning},
  author={Huang, Lang and Zhang, Chao and Zhang, Hongyang},
  journal={arXiv preprint arXiv:2101.08732},
  year={2021}
}

Acknowledgement

This code is based on:

We thank the authors for sharing their code.

Contact

If you have any question about this code, feel free to open an issue or contact [email protected].