Research map / Learning theory and algorithms / Prediction and decision losses
Probabilistic preference learning with the Mallows rank model
Valeria Vitelli, Øystein Sørensen, Marta Crispino, Arnoldo Frigessi, Elja Arjas
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 |
|---|---|---|---|
| increase the number of importance samples K from 10^2 to 10^6 (ablation) | IS partition function · Footrule, n=75 | max relative error: 152.036 -> 0.056 | Improved |
| increase the number of importance samples K from 10^2 to 10^6 (ablation) | IS partition function · Footrule, n=100 | max relative error: 67.487 -> 0.045 | Improved |
| increase K from 10^6 to 10^8 (ablation) | IS partition function · Footrule, n=100 | max relative error: 0.045 -> 0.004 | Improved |
| use IS approximation with K=10^4 instead of exact partition function (ablation) | Bayesian Mallows MCMC · Footrule, n=20 | posterior of alpha and rho | No clear effect |
| use asymptotic approximation Z_lim instead of exact or IS partition function (ablation) | Bayesian Mallows MCMC · Footrule, n=50 | posterior of alpha | Hurt |
| use approximate partition function (IS or asymptotic) (ablation) | Bayesian Mallows MCMC · Footrule, n=50 and n=100 | marginal posterior of rho | No clear effect |
| vary exponential prior rate lambda for alpha (0.1, 1, 10) (ablation) | Bayesian Mallows MCMC · Footrule, n=20 | posterior of rho | No clear effect |
| increase number of assessors N from 20 to 100 (ablation) | Bayesian Mallows MCMC · Footrule, n=20 | posterior precision | Improved |
| increase alpha_true (more concentrated assessors) | Bayesian Mallows MCMC · Footrule, n=50 and n=100 | posterior accuracy | Improved |
| choose number of clusters C equal to true value 3 | Mallows mixture model · Simulated pairwise data N=200, n=15 | within-cluster distance | Improved |
Closest papers by meaning
- Pseudo-Mallows for Efficient Probabilistic Preference Learning (ICML 2026 · similarity 0.91)
- Concentric mixtures of Mallows models for top-$k$ rankings: sampling and identifiability (ICML 2021 · similarity 0.89)
- On A Mallows-type Model For (Ranked) Choices (NeurIPS 2022 · similarity 0.89)
- Conditional independence and graphical models for rankings (NeurIPS 2026 · similarity 0.88)
- Generalized Top-k Mallows Model for Ranked Choices (NeurIPS 2025 · similarity 0.88)
- Preference Diffusion for Recommendation (ICLR 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.
