Research map / Learning theory and algorithms / Kernels and regression theory
Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Optimize the stochastic majority vote's expected risk directly rather than a proxy oracle bound | StocMV with Dirichlet weights · two-moons | test error | Helped |
| Use direct expected-risk optimization instead of FO or SO | StocMV · real benchmarks | test error | No clear effect |
| Use direct expected-risk optimization rather than the baselines' oracle-bound objectives | StocMV · real benchmarks | PAC-Bayesian bound | Helped |
| Increase the Monte Carlo draw count (ablation) | StocMV-MC · two-moons | test error | Helped |
| Use MC risk optimization instead of exact risk optimization (ablation) | StocMV · two-moons with n > 10^2 | test error | No clear effect |
| Increase the number of base predictors (ablation) | StocMV exact and MC · two-moons | test error | Hurt |
| Add input noise (ablation) | StocMV exact · two-moons | test error | Hurt |
| Use a sparser-favoring prior (ablation) | StocMV · Protein | test error | Helped |
| Use a prior encouraging concentrated posteriors near the simplex mean (ablation) | StocMV · MNIST | test error | Helped |
| Remove PAC-Bayes bound regularization and optimize empirical risk only (ablation) | StocMV exact · synthetic dataset | KL divergence to prior | Hurt |
| Use stronger base voters by increasing decision-tree depth (ablation) | StocMV and baselines · voter-strength experiments | test error | Helped |
| Use weaker base voters (ablation) | Second Order · voter-strength experiments | test error | Hurt |
Closest papers by meaning
- On Margins and Generalisation for Voting Classifiers (NeurIPS 2022 · similarity 0.92)
- Second Order PAC-Bayesian Bounds for the Weighted Majority Vote (NeurIPS 2020 · similarity 0.91)
- Exact power indices for plurality-voting ensembles (NeurIPS 2026 · similarity 0.85)
- How many classifiers do we need? (NeurIPS 2024 · similarity 0.85)
- On Rate-Optimal Partitioning Classification from Observable and from Privatised Data (NeurIPS 2026 · similarity 0.84)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (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.