Research map / Learning theory and algorithms / Prediction and decision losses
Probabilistic Ensemble of Collaborative Filters
Zhiyu Min, Dahua Lin
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace a single WMF filter with progressive PECF | WMF · MovieLens | Recall@50: 0.377 -> 0.460 | Improved |
| replace a single WMF filter with progressive PECF | WMF · CiteULike | Recall@50: 0.388 -> 0.446 | Improved |
| replace a single WMF filter with progressive PECF | WMF · Netflix | Recall@50: 0.142 -> 0.258 | Improved |
| use progressive complementary component construction instead of L₂Boost | collaborative-filter ensemble · MovieLens, CiteULike, and Netflix | Recall@50: MovieLens: 0.418 -> 0.460; CiteULike: 0.416 -> 0.446; Netflix: 0.178 -> 0.258 | Improved |
| add components progressively, focusing on poorly predicted entries (ablation) | PECF · MovieLens | Recall@50 | Improved |
| increase latent dimension from 50 to 80 (ablation) | PECF · MovieLens | Recall@50: 0.460 -> 0.457 | No clear effect |
| reduce latent dimension from 50 to 30 (ablation) | PECF · MovieLens | Recall@50: 0.460 -> 0.453 | No clear effect |
| reduce reweighting bandwidth σ from 1 to 0.1 at ν=10 (ablation) | PECF · MovieLens | Recall@50: 0.460 -> 0.450 | Hurt |
| reduce new-component assignment tendency ν from 10 to 1 at σ=1 (ablation) | PECF · MovieLens | Recall@50: 0.460 -> 0.435 | Hurt |
| set ν=1 and σ=0.1 instead of ν=10 and σ=1 (ablation) | PECF · MovieLens | Recall@50: 0.460 -> 0.431 | Hurt |
| replace dynamic α with fixed α=0.2 (ablation) | PECF · MovieLens | Recall | Hurt |
| replace dynamic α with fixed α=0.4 (ablation) | PECF · MovieLens | Recall | Hurt |
Closest papers by meaning
- The Sample Complexity of Online One-Class Collaborative Filtering (ICML 2017 · similarity 0.87)
- Markov Random Fields for Collaborative Filtering (NeurIPS 2019 · similarity 0.87)
- On the Generalizability and Predictability of Recommender Systems (NeurIPS 2022 · similarity 0.87)
- From Pairwise to Ranking: Climbing the Ladder to Ideal Collaborative Filtering with Pseudo-Ranking (AAAI 2025 · similarity 0.85)
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.85)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.85)
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.
