Research map / Learning theory and algorithms / Prediction and decision losses
Fast Algorithm for Generalized Multinomial Models with Ranking Data
Jiaqi Gu, Guosheng Yin
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 |
|---|---|---|---|
| Markov chain algorithm vs MM and Bayesian weaver | Markov chain algorithm · NASCAR | iterations: 14 -> 7 | Improved |
| Markov chain algorithm vs MM on running time | Markov chain algorithm · NASCAR | running time (s): 0.01 -> 0.07 | Mixed |
| fit Bradley-Terry instead of Plackett-Luce on full-ranking sushi | MM · Sushi full ranking | iterations: 9 -> 35 | Hurt |
| fit Bradley-Terry instead of Plackett-Luce on full-ranking sushi | Markov chain · Sushi full ranking | iterations: 6 -> 5 | No clear effect |
| increase number of basic cells d from 16 to 128 (small union incompleteness) | MM · Synthetic | avg iterations: 18.0 -> 6.0 | Improved |
| increase d from 16 to 128 (small union incompleteness) | Markov chain · Synthetic | avg iterations: 6.3 -> 6.0 | No clear effect |
| increase d from 16 to 128 (composite cell incompleteness) | Markov chain · Synthetic | avg iterations: 275.6 -> 342.1 | Hurt |
| lower |C|/d (stronger small union incompleteness) | MM and Bayesian weaver · Synthetic | iterations | Hurt |
| lower |C|/d (stronger small union incompleteness) | Markov chain · Synthetic | iterations | No clear effect |
| more basic cells per candidate set (composite incompleteness) | all three algorithms · Synthetic d=128 | iterations | Hurt |
| increase sample size n | all three algorithms · Synthetic cases II and III | iterations | No clear effect |
| fewer distinct likelihood components (64, 36, 22) | MM and Bayesian weaver · Synthetic case II | iterations | Hurt |
Closest papers by meaning
- Fast and Stable Maximum Likelihood Estimation for Incomplete Multinomial Models (ICML 2019 · similarity 0.90)
- A Model Selection Approach for Clustering a Multinomial Sequence with Non-Negative Factorization (TPAMI 2016 · similarity 0.86)
- Bayesian Non-Parametric Clustering of Ranking Data (TPAMI 2016 · similarity 0.86)
- Learning Rich Rankings (NeurIPS 2020 · similarity 0.86)
- Mallows ranking models: maximum likelihood estimate and regeneration (ICML 2019 · similarity 0.86)
- Generalized Results for the Existence and Consistency of the MLE in the Bradley-Terry-Luce Model (ICML 2022 · 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.
