Joint Optimization Framework for Learning with Noisy Labels

This repository contains the code for the paper Joint Optimization Framework for Learning with Noisy Labels.

Requirements

• Python 3.6
• Chainer 4.0.0
• CuPy 4.0.0
• ChainerCV 0.9.0

Training

To train the network on the Symmmetric Noise CIFAR-10 dataset (noise rate = 0.7):

$ python first_step_train.py --gpu 0 --out first_sn07 --learnrate 0.08 --alpha 1.2 --beta 0.8 --percent 0.7
$ python second_step_train.py --gpu 0 --out second_sn07 --label first_sn07

To train the network on the Asymmmetric Noise CIFAR-10 dataset (noise rate = 0.4):

$ python first_step_train.py --gpu 0 --out first_an04 --learnrate 0.03 --alpha 0.8 --beta 0.4 --percent 0.4 --asym
$ python second_step_train.py --gpu 0 --out second_an04 --label first_an04

Citation

@inproceedings{tanaka2018joint,
    title = {Joint Optimization Framework for Learning with Noisy Labels},
    author = {Tanaka, Daiki and Ikami, Daiki and Yamasaki, Toshihiko and Aizawa, Kiyoharu},
    booktitle = {CVPR},
    year = {2018}
}