Research map / Optimization / Optimizers
Stein Variational Gradient Descent Without Gradient
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 |
|---|---|---|---|
| Use an over-dispersed surrogate with its mean matched to the target (ablation) | GF-SVGD · 2D multivariate Gaussian | MMD: σρ ≈ 3σ | Helped |
| Use an over-dispersed surrogate with its mean matched to the target (ablation) | GF-SVGD · 2D multivariate Gaussian | mean square error of variance estimation: σρ ≈ 3σ | Mixed |
| Set the surrogate to the improper constant function ρ(x) = 1 (ablation) | GF-SVGD · 2D multivariate Gaussian | MMD | Hurt |
| Set the surrogate to the improper constant function ρ(x) = 1 (ablation) | GF-SVGD · 2D multivariate Gaussian | mean square error of variance estimation | Hurt |
| Add simulated annealing and construct a surrogate at each intermediate distribution (ablation) | GF-SVGD · 25-dimensional GMM | MMD | Helped |
| Use annealed GF-SVGD instead of gradient-free annealed importance sampling | AGF-SVGD · 25-dimensional GMM | MMD | Helped |
| Use annealed GF-SVGD instead of standard GF-SVGD (ablation) | AGF-SVGD · 25-dimensional GMM | MMD | Helped |
| Use annealed GF-SVGD instead of gradient-based SVGD | AGF-SVGD · 25-dimensional GMM | MMD | Hurt |
| Use GF-SVGD instead of standard importance sampling with the same surrogate | GF-SVGD · 2D multivariate Gaussian | MMD | Helped |
| Choose the surrogate variance to be approximately three times the target variance (ablation) | GF-SVGD · 2D multivariate Gaussian | MMD: σρ ≈ 3σ | Helped |
| Use AGF-SVGD instead of KAMH or KHMC | AGF-SVGD · Gauss-Bernoulli RBM | MMD | Helped |
| Use AGF-SVGD instead of KAMH or KHMC | AGF-SVGD · Gauss-Bernoulli RBM | mean square error of mean and variance estimation | Helped |
Closest papers by meaning
- Stein Variational Gradient Descent With Matrix-Valued Kernels (NeurIPS 2019 · similarity 0.94)
- Stein Variational Gradient Descent as Moment Matching (NeurIPS 2018 · similarity 0.93)
- A Stein variational Newton method (NeurIPS 2018 · similarity 0.93)
- A Non-Asymptotic Analysis for Stein Variational Gradient Descent (NeurIPS 2020 · similarity 0.92)
- Towards Understanding the Dynamics of Gaussian-Stein Variational Gradient Descent (NeurIPS 2023 · similarity 0.92)
- ELBOing Stein: Variational Bayes with Stein Mixture Inference (ICLR 2025 · similarity 0.90)
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.