Research map / Learning theory and algorithms / Prediction and decision losses
Indexable Probabilistic Matrix Factorization for Maximum Inner Product Search
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 |
|---|---|---|---|
| constrain item vectors to unit hypersphere via Von Mises-Fisher prior (IPMF vs BPMF) | BPMF · Netflix test, D=20 | RMSE: 0.9059 -> 0.9067 | No clear effect |
| constrain item vectors to unit hypersphere via Von Mises-Fisher prior (IPMF vs BPMF) | BPMF · Yahoo! Music test, D=20 | RMSE: 24.902 -> 25.063 | No clear effect |
| constrain item vectors to unit hypersphere via Von Mises-Fisher prior (IPMF vs BPMF) | BPMF · Yahoo! Music, D=100 | training time (h): 10.42 -> 31.45 | Hurt |
| constrain item vectors to unit hypersphere via Von Mises-Fisher prior (IPMF vs BPMF) | BPMF · Netflix, D=20 | training time (h): 0.91 -> 1.04 | Hurt |
| native indexable IPMF instead of BPMF with extra-dimension post-processing, with priority-search k-means tree | MIPS retrieval · Netflix, D=20, Precision@10 of 0.6 | retrieval speedup: around 50% faster | Helped |
| native indexable IPMF instead of BPMF+ post-processing | MIPS retrieval · Yahoo! Music, D=20, Precision@10 of 0.6 | retrieval speedup: at least 50% faster | Helped |
| use IPMF instead of BPMF+ with randomized kd-trees and LSH (ablation) | ANNS structures · Netflix, D=20 | speedup at fixed Precision@10 | Helped |
| IPMF vs BPMF on rarely rated items | BPMF · Netflix/Yahoo item groups, D=20 | RMSE | Hurt |
Closest papers by meaning
- Recommendations with Sparse Comparison Data: Provably Fast Convergence for Nonconvex Matrix Factorization (ICML 2025 · similarity 0.88)
- Ordinal Non-negative Matrix Factorization for Recommendation (ICML 2020 · similarity 0.88)
- Probabilistic Matrix Factorization for Automated Machine Learning (NeurIPS 2018 · similarity 0.87)
- Gamma-Poisson Dynamic Matrix Factorization Embedded with Metadata Influence (NeurIPS 2018 · similarity 0.87)
- Coupled Poisson Factorization Integrated With User/Item Metadata for Modeling Popular and Sparse Ratings in Scalable Recommendation (AAAI 2018 · similarity 0.87)
- Confidence-Aware Matrix Factorization for Recommender Systems (AAAI 2018 · similarity 0.87)
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.