AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Towards Hybrid-grained Feature Interaction Selection for Deep Sparse Network

Fuyuan Lyu, Xing Tang, Dugang Liu, Chen Ma, Weihong Luo, Liang Chen, xiuqiang He, Xue Liu

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Long-context KV and memory attention

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 value-level selection on top of field-level (hybrid granularity) (ablation)OptFeature-f · AvazuAUC: 0.7920 -> 0.7925Improved
Add field-level selection on top of value-level (hybrid) (ablation)OptFeature-v · KDD12AUC: 0.7981 -> 0.7982Improved
Use value-level instead of field-level selection (ablation)OptFeature · CriteoAUC: 0.8115 -> 0.8116No clear effect
Use value-level instead of field-level selection (ablation)OptFeature · KDD12AUC: 0.7978 -> 0.7981Improved
Use OptFeature selection algorithm instead of AutoFIS at field levelOptFeature-f · CriteoAUC: 0.8089 -> 0.8115Improved
Use sparsification-based selection instead of PROFIT progressive selectionOptFeature-f · CriteoLogloss: 0.4406 -> 0.4404Improved
Use outer product instead of inner product for interactions (ablation)OptFeature · AvazuAUC: 0.7925 -> 0.7899Hurt
Use outer product instead of inner product for interactions (ablation)OptFeature · KDD12AUC: 0.7982 -> 0.7961Hurt
Use outer product instead of inner product for interactions (ablation)OptFeature · CriteoAUC: 0.8116 -> 0.8115No clear effect
Increase selection dimension d-hat (ablation)OptFeature · CriteoAUCNo clear effect
Increase selection dimension d-hat (ablation)OptFeature · Avazu and KDD12AUCImproved
Add third-order interactions to second-order (ablation)OptFeature-f · KDD12AUCImproved

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