pairlab / CBS

Curriculum by Smoothing (NeurIPS 2020)
MIT License
39 stars 4 forks source link

Curriculum by Smoothing (NeurIPS 2020)

The official PyTorch implementation of Curriculum by Smoothing (NeurIPS 2020, Spotlight).

For any questions regarding the codebase, please send a message at: samarth.sinha@mail.utoronto.ca

USAGE: Simply use the code by running:

python3 main.py --dataset <DATASET> --alg <MODEL> --data <PATH_TO_DATA>

For example, to train a ResNet on CIFAR10 and the data is saved in ./data/, we can run:

python3 main.py --dataset cifar10 --alg res --data ./data/

For new expeiments it is important to tune the following hyperparameters:

--std --std_factor --epoch

LINK: https://arxiv.org/abs/2003.01367

This codebase has experiments for image classification and transfer learning.

If you use this codebase or find this repository helpful then please cite our paper:

@article{sinha2020curriculum,
  title={Curriculum By Smoothing},
  author={Sinha, Samarth and Garg, Animesh and Larochelle, Hugo},
  journal={Advances in Neural Information Processing Systems},
  volume={33},
  year={2020}
}