Research map / Label-efficient and robust learning / Recognition and long-tail
Simplify and Robustify Negative Sampling for Implicit Collaborative Filtering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| memory-based variance-aware negative sampling (SRNS) vs best baseline | SRNS · Movielens-1m / Pinterest / Ecommerce | NDCG@1: 0.1846 -> 0.1933 (ML-1m); 0.2690 -> 0.2891 (Pinterest); 0.1357 -> 0.1471 (Ecom) | Helped |
| SRNS vs uniform sampling | SRNS · Movielens-1m | NDCG@1: 0.1744 -> 0.1933 | Helped |
| add variance-based criterion over difficulty-only sampling (ablation) | SRNS · ML-100k and Ecom-toy synthetic noise | NDCG | Helped |
| increase memory size S1 with difficulty-only sampling (ablation) | SRNS · Movielens-1m/Pinterest/Ecommerce | NDCG@1 | Hurt |
| add variance-based sampling at large memory size (ablation) | SRNS · Movielens-1m/Pinterest | NDCG@1 | Helped |
| variance-based sampling on Ecom (ablation) | SRNS · Ecommerce | NDCG@1 | No clear effect |
| linearly increasing alpha (warm-start) vs flat or decreasing (ablation) | SRNS · ML-100k and Ecom-toy synthetic | NDCG | Helped |
| score-based memory sampling vs random selection from memory (ablation) | SRNS · real datasets | NDCG@1 | Helped |
| MLP scoring function instead of GMF (ablation) | SRNS · ML-1m and Ecom | NDCG | No clear effect |
| sampling-based hard negatives vs non-sampling ENMF | RNS-AS · Pinterest/Ecom | NDCG@1: 0.2594 -> 0.2690 (RNS-AS, Pinterest) | Helped |
Closest papers by meaning
- Adaptive Hardness Negative Sampling for Collaborative Filtering (AAAI 2024 · similarity 0.92)
- Relaxing the Accurate Imputation Assumption in Doubly Robust Learning for Debiased Collaborative Filtering (ICML 2024 · similarity 0.88)
- Unbiased Recommender Learning from Implicit Feedback via Weakly Supervised Learning (ICML 2025 · similarity 0.87)
- Neighborhood-Aware Negative Sampling for Student Knowledge and Behavior Modeling (AAAI 2025 · similarity 0.87)
- ImplicitSLIM and How it Improves Embedding-based Collaborative Filtering (ICLR 2024 · similarity 0.86)
- TriSampler: A Better Negative Sampling Principle for Dense Retrieval (AAAI 2024 · similarity 0.84)
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.