AI Paper Atlas

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

Boosting Factorization Machines via Saliency-Guided Mixup

Chenwang Wu, Defu Lian, Yong Ge, Min Zhou, Enhong Chen, Dacheng Tao

TPAMI 2024 · Label-efficient and robust learning › Recognition and long-tail · approach: Recommendation representation learning · 7 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
double raw samples by copying (CopyFM) (ablation)FM · PetAUC: 0.6776 -> 0.6739Hurt
add mixup with nearest-neighbor samples (MixFM)FM · FrappeLogLoss: 0.2047 -> 0.1537Improved
add mixup with nearest-neighbor samples (MixFM)FM · PetAUC: 0.6776 -> 0.6846Improved
add saliency-guided mixup (SMFM) instead of random-neighbor mixupMixFM · PetAUC: 0.6846 -> 0.7023Improved
add saliency-guided mixup (SMFM) instead of random-neighbor mixupMixFM · FrappeLogLoss: 0.1537 -> 0.1668Hurt
apply saliency-guided mixup to NFMNFM · MovieLensAUC: 0.9403 -> 0.9529Improved
apply saliency-guided mixup to DCNDCN · FrappeAUC: 0.9753 -> 0.9792Improved
apply mixup to DCNDCN · PetAUC: 0.7077 -> 0.7093No clear effect
use Maxup instead of saliency-guided mixupFM · FrappeLogLoss: 0.1668 -> 0.3374Hurt
increase number of mixed samples n' (ablation)MixFM/MixNFM · five datasetsAUCMixed
increase candidate neighbors p beyond 8 (ablation)SMNFM · MovieLensAUCHurt
increase embedding size d (ablation)MixFM/SMFM · Frappe, MovieLens vs Amazon datasetsAUCMixed

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