STUNT: Few-shot Tabular Learning with Self-generated Tasks from Unlabeled Tables (ICLR 2023 Spotlight)

Official PyTorch implementation of "STUNT: Few-shot Tabular Learning with Self-generated Tasks from Unlabeled Tables" by Jaehyun Nam, Jihoon Tack, Kyungmin Lee, Hankook Lee, Jinwoo Shin.

TL;DR: We propose a few-shot semi-supervised tabular learning framework that meta-learns over the self-generated tasks from unlabeled tables.

1. Dependencies

conda create -n stunt python=3.8 -y
conda activate stunt

pip install torch==1.8.1+cu111 torchvision==0.9.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
pip install torchmeta tensorboardX
conda install -c conda-forge faiss-gpu

2. Dataset

Download the datasets and place at /data folder

3. Training

3.1. Training option

The options for the training method are as follows:

• <MODE>: {protonet}
• <MODEL>: {mlp}
• <DATASET>: {income}

3.2. Preparing pseudo-validation by STUNT

cd data/<DATASET>
python generate_pseudo_val.py

3.3. Training

python main.py --mode <MODE> --model <MODEL> --dataset <DATASET>

4. Evaluation

Place the labeled sample index of the test set in data/<DATASET>/index<SHOT> before evaluation.

python eval.py --data_name <DATASET> --shot_num <SHOT> --seed <SEED> --load_path <PATH>

Citation

@inproceedings{nam2023stunt,
  title={{STUNT}: Few-shot Tabular Learning with Self-generated Tasks from Unlabeled Tables},
  author={Jaehyun Nam and Jihoon Tack and Kyungmin Lee and Hankook Lee and Jinwoo Shin},
  booktitle={International Conference on Learning Representations},
  year={2023}
}

Reference

• torchmeta
• BOIL
• MetaMix
• Sparse-MAML
• SiMT