FairFS: Addressing Deep Feature Selection Biases for Recommender System Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates. Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates ISBN: 979-8-4007-2307-0
Xianquan Wang, Zhaocheng Du, Jieming Zhu, Qinglin Jia, Zhenhua Dong, Kai Zhang · 2026
Read FairFS: Addressing Deep Feature Selection Biases for Recommender System Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates. Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates ISBN: 979-8-4007-2307-0 by Xianquan Wang, Zhaocheng Du, Jieming Zhu, Qinglin Jia, Zhenhua Dong, Kai Zhang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20001.txt.
Description
Recommender systems choose features from thousands of candidates, and judging which features matter most decides both online quality and computing cost. A 2026 arXiv paper by Xianquan Wang and five co-authors diagnoses three biases that skew that judgment in deep learning systems.
Gate-based and sensitivity-based selection methods can end up leaning on only part of the model layers, samples, or gradients, producing layer bias, baseline bias, and approximation bias. The proposed method, FairFS, regularizes importance estimates across all nonlinear layers, introduces a smooth baseline feature near the classifier boundary, and aggregates approximations to ease the other two biases. Tests show FairFS curbs all three biases and leads rival feature selection methods.
Catalog record
| CREATOR | Xianquan Wang, Zhaocheng Du, Jieming Zhu, Qinglin Jia, Zhenhua Dong, Kai Zhang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20001