VolMinNet

Provably End-to-end Label-noise Learning without Anchor Points

Xuefeng Li, Tongliang Liu, Bo Han, Gang Niu, Masashi Sugiyama.

PyTorch implementation

Dependencies

we implement our methods by PyTorch on NVIDIA Tesla V100. The environment is as bellow:

• PyTorch, version >= 1.7.1
• CUDA, version >= 11.1

Install PyTorch and Torchvision (Pip3):

pip3 install torch torchvision

Experiments

We verify the effectiveness of VolMinNet on three synthetic noisy datasets (MNIST, CIFAR-10, CIFAR-100), and one real-world noisy dataset (clothing1M). And We provide datasets (the images and labels have been processed to .npy format).

To run the code:

python3 main.py --dataset <-dataset-> --noise_type <-noise type-> --noise_rate <-noise rate-> --save_dir <-path of the directory->

Here is an example:

python3 main.py --dataset mnist --noise_type flip --noise_rate 0.45 --save_dir tmp

The statistics will be saved in the specified directory (tmp in this example).