Research map / Learning theory and algorithms / Prediction and decision losses
One for All: Simultaneous Metric and Preference Learning over Multiple Users
Gregory Canal, Blake Mason, Ramya Korlakai Vinayak, Robert D Nowak
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 |
|---|---|---|---|
| use the jointly constrained Nuclear full method instead of the reported baseline methods | Simulated data (d=10, r=1, n=100, K=10) · held-out test set | prediction accuracy | Improved |
| use Nuclear split instead of the other baseline methods | Simulated data (d=10, r=1, n=100, K=10) · held-out test set | prediction accuracy | Improved |
| replace Nuclear full's joint subspace constraint with separate nuclear-norm constraints (ablation) | Nuclear full · Simulated data (d=10, r=1, n=100, K=10) | prediction accuracy | Hurt |
| constrain the metric and user-vector matrix nuclear norms separately (ablation) | Nuclear full · Simulated data (d=10, r=1, n=100, K=10) | prediction accuracy | Improved |
| remove the nuclear-norm and user-vector constraints, enforcing only metric positive semidefiniteness (ablation) | Nuclear full · Simulated data (d=10, r=1, n=100, K=10) | prediction accuracy | Hurt |
| fit a separate metric and ideal point for each user instead of sharing a metric | Nuclear full · Simulated data (d=10, r=1, n=100, K=10) | prediction accuracy | Hurt |
| add a shared learned metric rather than fixing the metric to the identity | Color preference model · Color preference dataset, held-out test set | prediction accuracy | Improved |
| learn one common metric rather than a separate metric for each respondent | Color preference model · Color preference dataset, fewer than 20 pairs per user | prediction accuracy | Improved |
| learn a separate metric for each respondent rather than sharing one | Color preference model · Color preference dataset, after many queries | prediction accuracy | Improved |
Closest papers by meaning
- Leveraging Sparsity for Sample-Efficient Preference Learning: A Theoretical Perspective (ICML 2025 · similarity 0.86)
- Learning Multi-Dimensional Human Preference for Text-to-Image Generation (CVPR 2024 · similarity 0.85)
- Active preference learning for ordering items in- and out-of-sample (NeurIPS 2024 · similarity 0.85)
- Preference Learning with Response Time: Robust Losses and Guarantees (NeurIPS 2025 · similarity 0.84)
- Preference learning made easy: Everything should be understood through win rate (ICML 2025 · similarity 0.84)
- Geometric-Averaged Preference Optimization for Soft Preference Labels (NeurIPS 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.
