TenIPS: Inverse Propensity Sampling for Tensor Completion

This GitHub repository has the code for the paper [1] at AISTATS 2021. Please refer to the corresponding Jupyter notebook for each of the experiments.

The video in the semi-synthetic experiment is available for download at https://drive.google.com/file/d/1HX6-motNPz_xZnkTJ32ZetRj_VUkI9XZ/view?usp=sharing.

Acknowledgments for experiments

The authors would like to thank https://github.com/OsmanMalik/tucker-tensorsketch for the video used in the semi-synthetic experiment, thank https://github.com/georgehc/mnar_mc for the code on 1-bit matrix completion and weighted soft-impute, and thank several StackOverflow posts for some helper functions.

References

[1] Chengrun Yang, Lijun Ding, Ziyang Wu, Madeleine Udell. TenIPS: Inverse Propensity Sampling for Tensor Completion. AISTATS 2021.