AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Meta-Wrapper: Differentiable Wrapping Operator for User Interest Selection in CTR Prediction

Tianwei Cao, Qianqian Xu, Zhiyong Yang, Qingming Huang

TPAMI 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Recommendation representation learning · 13 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add attentive feature selector (C1) to the base predictor (ablation)Base model · five CTR datasetsAUCImproved
add the auxiliary feature-selection loss (C2) to the attentive selector (ablation)M1 · five CTR datasetsAUCImproved
add the GDmax wrapper with N=1 (C3) without the second term (ablation)M2 · five CTR datasetsAUCHurt
add the second meta-learned term (C4) with N=1 (ablation)M3 · five CTR datasetsAUCImproved
use second-order crossing models over high-order ones (Wide&Deep, PNN, DeepFM vs xDeepFM, FGCNN)feature-crossing baselines · ElectronicsAUC: 0.7979 vs 0.7800Improved
replace general feature selection (AutoFIS) with attention (DIN)user-behavior CTR models · BooksAUC: 0.7789 -> 0.7818Improved
replace DIN with Meta-WrapperDIN · BooksAUC: 0.7818 -> 0.8237Improved
add attentive feature selector to base model on noisy synthetic dataBase · synthetic noisy dataAUCHurt
add L2 weight decay or dropout to the feature selector modelFS · synthetic noisy dataAUCNo clear effect
replace FS with Meta-WrapperFS · synthetic noisy dataAUCImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy