Cross-Dataset Propensity Estimation for Debiasing Recommender Systems
Information Retrieval
2022-12-29 v1 Machine Learning
Methodology
Abstract
Datasets for training recommender systems are often subject to distribution shift induced by users' and recommenders' selection biases. In this paper, we study the impact of selection bias on datasets with different quantization. We then leverage two differently quantized datasets from different source distributions to mitigate distribution shift by applying the inverse probability scoring method from causal inference. Empirically, our approach gains significant performance improvement over single-dataset methods and alternative ways of combining two datasets.
Cite
@article{arxiv.2212.13892,
title = {Cross-Dataset Propensity Estimation for Debiasing Recommender Systems},
author = {Fengyu Li and Sarah Dean},
journal= {arXiv preprint arXiv:2212.13892},
year = {2022}
}
Comments
In Workshop on Distribution Shifts, 36th Conference on Neural Information Processing Systems (NeurIPS 2022)