Research map / Optimization / Optimizers
Zeroth-Order Non-Log-Concave Sampling with Variance Reduction and Applications to Inverse Problems
M. Berk Sahin, Behzad Sharif, Abolfazl Hashemi
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 variance-reduced ZO estimator with O(1) evaluations per iteration instead of O(d) batch | ZO-LMC · non-log-concave sampling theory | batch size / evaluations per iteration | Improved |
| lower large-batch probability p to 0.3 (ablation) | ZO-APMC · synthetic bimodal Gaussian mixture inverse problem | relative Fisher information | Hurt |
| vary p and b at fixed per-iteration cost pb = 10 (ablation) | ZO-APMC · synthetic inverse problem | relative Fisher information: FI below 0.01 | No clear effect |
| use ZO-APMC instead of gradient-based APMC | ZO-APMC · MRI reconstruction | PSNR (dB): 36.55 -> 35.29 | Mixed |
| use ZO-APMC instead of best black-box baseline DPG | ZO-APMC · MRI reconstruction | PSNR (dB): 32.17 -> 35.29 | Improved |
| use ZO-APMC instead of Forward-GSG | ZO-APMC · black-hole imaging | closure-phase chi2: 6.77 -> 5.42 | Improved |
| use ZO-APMC instead of DPG on MRI | ZO-APMC · MRI reconstruction | sample SD: 2.69e-2 -> 2.99e-2 | Hurt |
| use ZO-APMC instead of EnKG/DPG on Navier-Stokes | ZO-APMC · Navier-Stokes | NRMSE | No clear effect |
| increase batch size b or probability p (ablation) | ZO-APMC · MRI/black-hole | sample SD | Improved |
Closest papers by meaning
- Zeroth-Order Sampling Methods for Non-Log-Concave Distributions: Alleviating Metastability by Denoising Diffusion (NeurIPS 2024 · similarity 0.90)
- From stability of Langevin diffusion to convergence of proximal MCMC for non-log-concave sampling (NeurIPS 2025 · similarity 0.88)
- Non-asymptotic Error Bounds in $\mathcal{W}_2$-Distance with Sqrt(d) Dimension Dependence and First Order Convergence for Langevin Monte Carlo beyond Log-Concavity (ICML 2025 · similarity 0.87)
- Fast Non-Log-Concave Sampling under Nonconvex Equality and Inequality Constraints with Landing (NeurIPS 2025 · similarity 0.87)
- Provable Posterior Sampling with Denoising Oracles via Tilted Transport (NeurIPS 2024 · similarity 0.87)
- Preconditioned Langevin Dynamics with Score-based Generative Models for Infinite-Dimensional Linear Bayesian Inverse Problems (NeurIPS 2025 · similarity 0.87)
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.
