A Second-Order Approach to Learning with Instance-Dependent Label Noise (CVPR'21 oral)

This code is a PyTorch implementation of the paper "A Second-Order Approach to Learning with Instance-Dependent Label Noise" accepted by CVPR 2021 as oral presentation.

Prerequisites

Python 3.6.9

PyTorch 1.4.0

Torchvision 0.5.0

Instructions

Run the code:

CIFAR10:

python run_exptPRLD_C10_CAL.py

CIFAR100:

python run_exptPRLD_C100_CAL.py

The following changes also apply to CIFAR100.

Run the code step-by-step

Step-1: Construct $\hat D$:

Modify Lines 27-34 of run_exptPRLD_C10_CAL.py as:

#-------------- customized parameters --------------#
noise_rate = 0.6 # noise rates = 0.2, 0.4, 0.6

lossfunc = "crossentropy"  # use this lossfunc for constructing D
# lossfunc = "crossentropy_CAL" # use this lossfunc for CAL

gpu_idx = "0"   # Choose one GPU index
#---------------------------------------------------#

Step-2: Train CAL:

Modify Lines 27-34 of run_exptPRLD_C10_CAL.py as:

#-------------- customized parameters --------------#
noise_rate = 0.6 # noise rates = 0.2, 0.4, 0.6

# lossfunc = "crossentropy"  # use this lossfunc for constructing D
lossfunc = "crossentropy_CAL" # use this lossfunc for CAL

gpu_idx = "0"   # Choose one GPU index
#---------------------------------------------------#

Citation

If you find this code useful, please cite the following paper:

@inproceedings{zhu2021second,
  title={A second-order approach to learning with instance-dependent label noise},
  author={Zhu, Zhaowei and Liu, Tongliang and Liu, Yang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={10113--10123},
  year={2021}
}

Corresponding authors:

Dr. Zhaowei Zhu: zwzhu@ucsc.edu

Prof. Yang Liu: yangliu@ucsc.edu