Reproduce “Towards a Better Tradeoff between Effectiveness and Efficiency in Pre-Ranking: A Learnable Feature Selection based Approach” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xu Ma, Pengjie Wang, Hui Zhao, Shaoguo Liu, Chuhan Zhao, Wei Lin, Kuang-Chih Lee, Jian Xu and 1 more
The research introduces the FSCD method, a learnable feature selection approach for pre-ranking that jointly considers feature effectiveness and computational complexity. This enables an interaction-focused model to achieve a 1.54% increase in CTR and 2.76% in RPM in a live e-commerce search system, matching the efficiency of simpler models.
Key insights
- Jointly optimizing for feature effectiveness and computational efficiency through a learnable selection mechanism (FSCD) allows for an improved pre-ranking trade-off.
- An interaction-focused pre-ranking architecture, even when utilizing a carefully selected subset of features, can yield substantial improvements in online CTR and RPM.
- FSCD demonstrates the ability to prioritize features with lower online computational complexity, yet it can still select highly effective complex features if their predictive power justifies the cost.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.