DebiasedSR_DRO

This is a PyTorch implementation for the paper accepted by WSDM 2024 (oral presentation):

Jiyuan Yang, Yue Ding, Yidan Wang, Pengjie Ren, Zhumin Chen, Fei Cai, Rui Zhang, Jun Ma, Zhaochun Ren, Xin Xin. Debiasing Sequential Recommenders through Distributionally Robust Optimization over System Exposure.

Overview

In this paper, we propose to debias sequential recommenders through Distributionally Robust Optimization (DRO) over system exposure data. The key idea is to utilize DRO to optimize the worst-case error over an uncertainty set to safeguard the model against distributional discrepancy caused by the exposure bias. The main challenge to apply DRO for exposure debiasing in sequential recommendation lies in how to construct the uncertainty set and avoid the overestimation of user preference on biased samples. Moreover, since the test set could also be affected by the exposure bias, how to evaluate the debiasing effect is also an open question. To this end, we first introduce an exposure simulator trained upon the system exposure data to calculate the exposure distribution, which is then regarded as the nominal distribution to construct the uncertainty set of DRO. Then, we introduce a penalty to items with high exposure probability to avoid the overestimation of user preference for biased samples. Finally, we design a debiased self-normalized inverse propensity score (SNIPS) evaluator for evaluating the debiasing effect on the biased offline test set. We conduct extensive experiments on two real-world datasets to verify the effectiveness of the proposed methods. Experimental results demonstrate the superior exposure debiasing performance of the proposed methods.

Requirements

torch==1.1.0
numpy==1.19.1
scipy==1.5.2
tqdm==4.48.2

Datasets

The default dataset setting of code is ZhihuRec and run following code to preprocess the ZhihuRec dataset:

python data/ZhihuRec/data_process_ZhihuRec.py

For the Tenrec dataset, the licence to acquire data is needed, and more details can be found in this link. After that, put the 'QB-video.csv' into 'data/Tenrec' and run following code to preprocess the Tenrec dataset:

python data/Tenrec/data_process_Tenrec.py

Recommender

To reproduce results of recomender,

python run_full.py

Exposure simulator and Evaluation simulator

Run following code to access the Exposure Simulator and Evaluation Simulator, respectively:

python exposure.py

python evaluation.py

Slurm script

Besides, we provide the slurm execute script to run our code.

More details about usage of slurm can be found in this link: https://slurm.schedmd.com/documentation.html

[Note: please modify "conda activate envname" to your environment]

Reference

If you find our codes and datasets useful for your research, please cite:

@inproceedings{Yang2024DebiasingSR,
  title={Debiasing Sequential Recommenders through Distributionally Robust Optimization over System Exposure},
  author={Jiyuan Yang and Yue Ding and Yidan Wang and Pengjie Ren and Zhumin Chen and Fei-yang Cai and Jun Ma and Rui Zhang and Zhaochun Ren and Xin Xin},
  year={2024},
  booktitle={Proceedings of the Seventeen ACM International Conference on Web Search and Data Mining}
}

The code is implemented based on S3-Rec.