Research map / Learning theory and algorithms / Prediction and decision losses
Learning from Comparisons and Choices
Sahand Negahban, Sewoong Oh, Kiran K. Thekumparampil, Jiaming Xu
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 nuclear-norm regularized MNL estimator instead of unregularized and rank-1 Plackett-Luce on Food100 | nucnorm MNL · Food100 (test data) | mean test log-likelihood | Improved |
| add nuclear-norm regularization (vs fullrank lambda=0) on Food100 training data (ablation) | nucnorm MNL · Food100 (training data) | mean training log-likelihood | Hurt |
| solve G groups of items jointly instead of separately (ablation) | nuclear norm MNL estimator · Simulation, d=360, r=4, n=2^14, varying G | L-RMSE and RMSE | Improved |
| use empirical Laplacian instead of the true Laplacian in the L-nuc norm (ablation) | nuclear norm MNL estimator · Simulation, various sampling graphs | L-RMSE | No clear effect |
| use k-wise ranking likelihood instead of pairwise rank breaking (ablation) | MNL estimator · Simulation, d=50, r=3 and 12 | RMSE | Improved |
| use line or barbell sampling graph instead of complete graph with i.i.d. Theta (ablation) | nuclear norm MNL estimator · Simulation, d=300, r=4, i.i.d. Theta | RMSE | No clear effect |
| use line or barbell sampling graph instead of complete graph with topology-biased Theta (ablation) | nuclear norm MNL estimator · Simulation, barbell biased and line biased Theta | Frobenius RMSE | Hurt |
| vary regularization lambda over a broad range (ablation) | nuclear norm MNL estimator · Simulation, k-wise comparisons | RMSE: lambda in [sqrt(log d/(k d^2)), 2^8 sqrt(log d/(k d^2))] | No clear effect |
Closest papers by meaning
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.87)
- Active preference learning for ordering items in- and out-of-sample (NeurIPS 2024 · similarity 0.87)
- Recommendations with Sparse Comparison Data: Provably Fast Convergence for Nonconvex Matrix Factorization (ICML 2025 · similarity 0.87)
- Learning Multiple User Distributions for Recommendation via Guided Conditional Diffusion (AAAI 2025 · similarity 0.86)
- Preference learning made easy: Everything should be understood through win rate (ICML 2025 · similarity 0.86)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.86)
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.
