Conference Paper (https://iclr.cc/)

Title: TDR-CL: TARGETED DOUBLY ROBUST COLLABORATIVE LEARNING FOR DEBIASED RECOMMENDATIONS

Please refer to the following jupyter notebooks in the "semi-synthetic" folder for ml-100k and ml-1m semi-synthetic experiments.

-- completion.ipynb: complete the entire user-item matrix.
-- convert.ipynb: generate six predicted matrices.
-- synthetic.ipynb: get the final results and comparision on four baselines.

Please refer to the following jupyter notebooks in the "real world" folder for real world experiments.

-- coat_mf.ipynb: comparison on baselines with MF (Matrix Factorization) backbone on COAT dataset.
-- yahoo_mf.ipynb: comparison on baselines with MF (Matrix Factorization) backbone on YAHOO dataset.

Our experimental environment is shown below:

sklearn version: 1.0.2
pytorch version: 1.9.0 + cu111
numpy version: 1.21.2
tensorflow version: 1.5.0

We follow the previous studies, which are shown below:

```
@inproceedings{MRDR,
  author    = {Siyuan Guo and
               Lixin Zou and
               Yiding Liu and
               Wenwen Ye and
               Suqi Cheng and
               Shuaiqiang Wang and
               Hechang Chen and
               Dawei Yin and
               Yi Chang},
  title     = {Enhanced Doubly Robust Learning for Debiasing Post-Click Conversion
               Rate Estimation},
  booktitle = {SIGIR}, 
  year      = {2021}
}
```

```
@inproceedings{wang2020information,
  title={Information Theoretic Counterfactual Learning from Missing-Not-At-Random Feedback},
  author={Wang, Zifeng and Chen, Xi and Wen, Rui and Huang, Shao-Lun and Kuruoglu, Ercan E and Zheng, Yefeng},
  booktitle={Neural Information Processing Systems (NeurIPS)},
  year={2020}
}
```

