Research map / Optimization / Optimizers
Stochastic Bouncy Particle Sampler
Ari Pakman, Dar Gilboa, David Carlson, Liam Paninski
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 |
|---|---|---|---|
| Use SBPS with an adaptive local-regression thinning bound instead of optimally tuned stochastic-gradient samplers | Bayesian logistic regression · Bayesian logistic regression | mixing speed | Improved |
| Use SBPS instead of exact-bound subsampling samplers | Bayesian logistic regression · Bayesian logistic regression | mixing speed | Improved |
| Add diagonal adaptive preconditioning | SBPS · Bayesian logistic regression | convergence to posterior mode | Improved |
| Estimate expectations along continuous trajectory segments rather than using discrete samples | SBPS · Continuous trajectory sampling, when r/b is less than 1 | expectation-estimation error | Improved |
| Use SBPS rather than SGLD with a selected step size | MNIST posterior sampling · MNIST | mixing | No clear effect |
| Use SBPS rather than large-step SGLD or mSGNHT | Highly non-Gaussian hyperboloid posterior · Hyperboloid posterior | posterior exploration | Improved |
| Use zero-mean independent noisy gradients | BPS · BPS with noisy gradients | invariant-distribution bias | No clear effect |
| Use noisy rather than exact gradients | BPS · BPS with noisy gradients | mixing speed per iteration | Hurt |
| Replace the local linear predictive model with Gaussian process regression (ablation) | SBPS · Examples considered | mixing speed | No clear effect |
| Use a small, controlled bias through approximate thinning rather than exact conservative bounds | SBPS · Bayesian logistic regression | mixing speed | Improved |
| Estimate expectations by continuous trajectory integration rather than discrete sampling | SBPS · Continuous trajectory sampling, when r/b is approximately 1 | expectation-estimation error | No clear effect |
| Use a poor SGLD step size rather than an appropriately chosen one | SGLD · MNIST posterior sampling | mixing | Mixed |
Closest papers by meaning
- Analysis of high-dimensional Continuous Time Markov Chains using the Local Bouncy Particle Sampler (JMLR 2021 · similarity 0.88)
- The Boomerang Sampler (ICML 2020 · similarity 0.88)
- Stochastic Localization via Iterative Posterior Sampling (ICML 2024 · similarity 0.86)
- To discretize continually: Mean shift interacting particle systems for Bayesian inference (NeurIPS 2026 · similarity 0.86)
- Sampling from Binary Quadratic Distributions via Stochastic Localization (ICML 2025 · similarity 0.86)
- Provable Posterior Sampling with Denoising Oracles via Tilted Transport (NeurIPS 2024 · similarity 0.86)
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.
