This is the official pytorch implementation of AutoDebias, a debiasing method for recommendation system. AutoDebias is proposed in the paper:
AutoDebias: Learning to Debias for Recommendation
by Jiawei Chen, Hande Dong, Yang Qiu, Xiangnan He, Xin Xin, Liang Chen, Guli Lin and Keping Yang
Published at SIGIR 2021.
AutoDebias is an automatic debiasing method for recommendation system based on meta learning, exploiting a small amout of uniform data to learn de-biasing parameters and using these parameters to guide the learning of the recommendation model.
The code runs well under python 3.8.5. The required packages are as follows:
- pytorch == 1.4.0
- numpy == 1.19.1
- scipy == 1.5.2
- pandas == 1.1.3
- cppimport == 20.8.4.2
We use two public datasets (Yahoo!R3 and Coat) and a synthetic dataset (Simulation).
- user.txt: biased data collected by normal policy of recommendation platform. For Yahoo!R3 and Coat, each line is user ID, item ID, rating of the user to the item. For Simulation, each line is user ID, item ID, position of the item, binary rating of the user to the item.
- random.txt: unbiased data collected by stochastic policy where items are assigned to users randomly. Each line in the file is user ID, item ID, rating of the user to the item.
Explicit feedback
- For dataset Yahoo!R3:
python train_explicit.py --dataset yahooR3
- For dataset Coat:
python train_explicit.py --dataset coat
Implicit feedback
- For dataset Yahoo!R3:
python train_implicit.py --dataset yahooR3
- For dataset Coat:
python train_implicit.py --dataset coat
Feedback on list recommendation
- For dataset Simulation:
python train_list.py --dataset simulation
Please contact [email protected] or [email protected] if you have any questions about the code and paper.