Please contact DaXu5180@gmail.com or Ruanchuanwei@gmail.com for questions.
The feedback data of recommender systems are often subject to what was exposed to the users; however, most learning and evaluation methods do not account for the underlying exposure mechanism. We first show in theory that applying supervised learning to detect user preferences may end up with inconsistent results in the absence of exposure information. The counterfactual propensity-weighting approach from causal inference can account for the exposure mechanism; nevertheless, the partial-observation nature of the feedback data can cause identifiability issues. We propose a principled solution by introducing a minimax empirical risk formulation. We show that the relaxation of the dual problem can be converted to an adversarial game between two recommendation models, where the opponent of the candidate model characterizes the underlying exposure mechanism. We provide learning bounds and conduct extensive simulation studies to illustrate and justify the proposed approach over a broad range of recommendation settings, which shed insights on the various benefits of the proposed approach.
Install the package in edit mode.
conda create -n acgan
source activate acgan
pip install -e .
#log will be saved here
mkdir -p log
Inside each folder in data, there is a Pyhton script containing the preprocessing logics. Please down the data from the links below and place them in the same folder where the script locates.
-
MoivesLens-1M data: http://files.grouplens.org/datasets/movielens/ml-1m.zip
-
LastFM: http://files.grouplens.org/datasets/hetrec2011/hetrec2011-lastfm-2k.zip
-
Goodread: https://drive.google.com/uc?id=1roQnVtWxVE1tbiXyabrotdZyUY7FA82W'
- Real-world data
Example of running experiemnts on Goodread data.
SIM_PATH='./data/simulations'
DATA_PATH='./data/books'
python train_on_real.py --data_path $DATA_PATH --prefix books_real
- Simulation
Example of running experiemnts on MovieLens-1M data.
SIM_PATH='./data/simulations'
DATA_PATH='./data/ml-1m'
python robust_simulation.py \
--sim_path $SIM_PATH \
--data_path $DATA_PATH --prefix ml_1m_sim
python train_on_simulation.py \
--sim_path $SIM_PATH \
--data_path $DATA_PATH --cuda_idx 0 --prefix ml_1m_sim --models mlp acgan
We use a different random split and negative sampling hence the metrics reported in the paper differ from what reported in Neural Collaborative Filtering. Our implementation is able to achieve similar performance using the data and evaluation methods provided in https://github.com/hexiangnan/neural_collaborative_filtering.
To validate the performance, please extract data/ncf_data.tar.gz and then run NCF_validation.py.
@article{xu2020adversarial,
title={Adversarial Counterfactual Learning and Evaluation for Recommender System},
author={Xu, Da and Ruan, Chuanwei and Korpeoglu, Evren and Kumar, Sushant and Achan, Kannan},
journal={Advances in Neural Information Processing Systems},
volume={33},
year={2020}
}