PET

This is the official implementation of PET (Personalized View Weighting with Data Enhancement Two-Pronged Contrast) (CIKM 2024 Short Paper Track) [Paper Link]

---

Supplementary Document

For additional data anlysis results and details about loss function, you can check here. Supplementary Document

Datasets

We use three widely used datasets for bundle recommendation, iFashion, NetEase and Youshu. For the iFashion dataset, please unzip data.zip in the same folder.

---

How to run PET

• Installation

cd PET
pip install -r requirements.txt

• iFashion

python train.py -d iFashion -g [gpu_id]

• NetEase

python train.py -d NetEase -g [gpu_id]

• Youshu

python train.py -d Youshu -g [gpu_id]   

---

Citation

We appreciate your interest in our work. If our research contributes to your projects, please consider citing our paper:

@inproceedings{kim2024towards,
  title={Towards Better Utilization of Multiple Views for Bundle Recommendation},
  author={Kim, Kyungho and Kim, Sunwoo and Lee, Geon and Shin, Kijung},
  booktitle={CIKM},
  year={2024}
}

---

Acknowledgement

This code is implemented based on the open source code from the paper CrossCBR : Cross-view Contrastive Learning for Bundle Recommendation (KDD '22).