Research map / Learning theory and algorithms / Sample complexity
Lost Relatives of the Gumbel Trick
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 |
|---|---|---|---|
| Replace the Gumbel trick estimator with the Exponential trick estimator | Partition-function estimation in A* sampling · Robust Bayesian Regression | samples required for a given MSE: up to 40% fewer samples | Helped |
| Use a Weibull trick parameter between the Gumbel and Exponential tricks | Partition-function estimation with Racing-Normal · Synthetic dataset, very hard noise setting (σ = 0.1) | MSE | Helped |
| Choose nonzero α for the low-rank upper bound | Sum-unary perturbation bounds · 10 × 10 binary pairwise grid models | MSE of the log-partition-function estimator | Mixed |
Closest papers by meaning
- Stochastic Discrete Clenshaw-Curtis Quadrature (ICML 2016 · similarity 0.87)
- Sampling from Structured Log-Concave Distributions via a Soft-Threshold Dikin Walk (NeurIPS 2023 · similarity 0.87)
- Stochastic Beams and Where To Find Them: The Gumbel-Top-k Trick for Sampling Sequences Without Replacement (ICML 2019 · similarity 0.86)
- Exact Sampling with Integer Linear Programs and Random Perturbations (AAAI 2016 · similarity 0.86)
- Faster Sampling from Log-Concave Densities over Polytopes via Efficient Linear Solvers (ICLR 2024 · similarity 0.85)
- Time series analysis with Gumbel dynamics (NeurIPS 2026 · 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.