Transfer and Share: Semi-Supervised Learning from Long-Tailed Data

by Tong Wei, Qian-Yu Liu, Jiang-Xin Shi, Wei-Wei Tu and Lan-Zhe Guo at SEU, NJU, and 4Paradigm Inc.

Preprint

This repository contains an official re-implementation of TRAS from the authors. Further information please contact Tong Wei and Qian-Yu Liu.

This repo has TRAS on Wide-ResNet 28.

Citation

If you find our work inspiring or use our codebase in your research, please consider giving a star ⭐ and a citation.

@article{wei2022transfer,
  title={Transfer and Share: Semi-Supervised Learning from Long-Tailed Data},
  author={Wei, Tong and Liu, Qian-Yu and Shi, Jiang-Xin and Tu, Wei-Wei and Guo, Lan-Zhe},
  journal={Machine Learning},
  year={2022}
}

Transfer and Share: Semi-Supervised Learning from Long-Tailed Data

Requirements

Python 3.8.8
PyTorch 1.9.1+cu111
torchvision
randAugment (Pytorch re-implementation: https://github.com/ildoonet/pytorch-randaugment)
pandas
numpy

Dataset

The code will download data automatically with the dataloader. We use data the same way as ABC.

dataset
├── fix_cifar10.py
├── fix_cifar100.py
└── fix_svhn.py

Training and Evaluation Instructions

Please check out run.sh for the script to run our TRAS.

python train.py --gpu 0 --label_ratio 20 --num_max 1000 --imb_ratio 100 \
--epoch 500 --val-iteration 500 --manualSeed 0 --dataset cifar10  \
--out out

How to get support from us?

If you have any general questions, feel free to email us at weit at seu.edu.cn and liuqy at lamda.nju.edu.cn. If you have code or implementation-related questions, please feel free to send emails to us or open an issue in this codebase (We recommend that you open an issue in this codebase, because your questions may help others).