Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Permuted and Augmented Stick-Breaking Bayesian Multinomial Regression
Quan Zhang, Mingyuan Zhou
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 |
|---|---|---|---|
| add larger capacity (K=5,T=3 instead of K=T=1) to MSR (ablation) | MSR · square(3) | error rate (%): 57.14 -> 0 | Improved |
| add larger capacity (K=5,T=3 instead of K=T=1) to MSR (ablation) | MSR · satimage(6) | error rate (%): 16.65 -> 12.00 | Improved |
| stack a second MSR on transformed covariates (DT-MSR) (ablation) | MSR K=5 T=3 · vehicle(4) | error rate (%): 15.75 -> 14.96 | Improved |
| stack a second MSR on transformed covariates (DT-MSR) (ablation) | MSR K=5 T=3 · segment(7) | error rate (%): 5.63 -> 7.65 | Hurt |
| use paSB-MLR with permuted category-stick mapping instead of L2-MLR | multinomial logistic regression · square(3) | error rate (%): 62.29 -> 59.52 | Improved |
| use paSB-robit instead of paSB-MLR | paSB regression · satimage(6) | error rate (%): 20.65 -> 17.25 | Improved |
| use paSB-robit instead of paSB-MLR | paSB regression · iris(3) | error rate (%): 4.00 -> 5.33 | Hurt |
| add MH update of category-stick mapping (paSB instead of aSB) (ablation) | MSR K=5 T=3 · satimage log-likelihood | log-likelihood: -1203.82 | Improved |
| add MH update of category-stick mapping (parSB instead of arSB) (ablation) | MSR K=5 T=3 · satimage log-likelihood | log-likelihood: -1350.21 | Improved |
| add mapping inference (paSB-MSR instead of aSB-MSR with random fixed mapping) (ablation) | MSR K=T=4 · square101 | log-likelihood: far below -4000 -> larger | Improved |
| use paSB-MLR instead of Bayes MLR Gibbs sampler | multinomial logistic regression · iris testing | median ESS: 156.25 -> 174.16 | Improved |
| use paSB-MLR instead of Bayes MLR for the 10% ESS quantile | multinomial logistic regression · square training | 10% quantile ESS: 411.46 -> 194.54 | Hurt |
Closest papers by meaning
- A Class of Conjugate Priors for Multinomial Probit Models which Includes the Multivariate Normal One (JMLR 2022 · similarity 0.85)
- Bayesian Dyadic Trees and Histograms for Regression (NeurIPS 2017 · similarity 0.84)
- Bayesian Additive Distribution Regression (NeurIPS 2026 · similarity 0.84)
- Being Bayesian about Categorical Probability (ICML 2020 · similarity 0.84)
- Efficient Analytic Uncertainty Quantification for Multimodal Regression (NeurIPS 2026 · similarity 0.83)
- SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning (ICML 2024 · similarity 0.82)
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.
