anyuexuan / PCL

Code for "Learning to Learn from Corrupted Data for Few-Shot Learning" in IJCAI 2023.

Code for PCL

Code for "Learning to Learn from Corrupted Data for Few-Shot Learning" in IJCAI 2023.

If you use the code in this repo for your work, please cite the following bib entries:

@inproceedings{An2023PCL,
  author    = {Yuexuan An and
               Xingyu Zhao and
               Hui Xue},
  title     = {Learning to Learn from Corrupted Data for Few-Shot Learning},
  booktitle = {Proceedings of the 32nd International Joint Conference on Artificial Intelligence, {IJCAI} 2023, Macao, China, 19-25 August 2023},
  pages     = {},
  year      = {2023},
}

Requirements

• Python >= 3.6
• PyTorch (GPU version) >= 1.5
• NumPy >= 1.13.3
• Scikit-learn >= 0.20

Getting started

• Change directory to ./filelists/cifar
• Download CIFAR-FS
• run python make.py in the terminal

Running the scripts

To train and test the PCL model in the terminal, use:

$ python run_PCL.py --dataset cifar --algorithm PCL_matchingnet --tao 2.0 --noise_type feature --noise_rate 0.2 --train_n_way 5 --test_n_way 5 --n_shot 5 --model_name Conv4 --device cuda:0

Acknowledgment

Our project references the codes and datasets in the following repo and papers.

CloserLookFewShot

Oriol Vinyals, Charles Blundell, Tim Lillicrap, Koray Kavukcuoglu, Daan Wierstra. Matching Networks for One Shot Learning. NIPS 2016: 3630-3638.