Tuning Sequential Monte Carlo Samplers via Greedy Incremental Divergence Minimization
Abstract
The performance of sequential Monte Carlo (SMC) samplers heavily depends on the tuning of the Markov kernels used in the path proposal. For SMC samplers with unadjusted Markov kernels, standard tuning objectives, such as the Metropolis-Hastings acceptance rate or the expected-squared jump distance, are no longer applicable. While stochastic gradient-based end-to-end optimization has been explored for tuning SMC samplers, they often incur excessive training costs, even for tuning just the kernel step sizes. In this work, we propose a general adaptation framework for tuning the Markov kernels in SMC samplers by minimizing the incremental Kullback-Leibler (KL) divergence between the proposal and target paths. For step size tuning, we provide a gradient-and tuning-free algorithm that is generally applicable for kernels such as Langevin Monte Carlo (LMC). We further demonstrate the utility of our approach by providing a tailored scheme for tuning kinetic LMC used in SMC samplers. Our implementations are able to obtain a full schedule of tuned parameters at the cost of a few vanilla SMC runs, which is a fraction of gradient-based approaches.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
3 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Practical and Asymptotically Exact Conditional Sampling in Diffusion Models cites Diffusion Probabilistic Modeling of Protein Backbones in 3D for the motif-scaffolding problem
- Efficient Stochastic Optimisation via Sequential Monte Carlo cites Learning Latent Variable Models via Jarzynski-adjusted Langevin Algorithm
- Asymptotically exact variational flows via involutive MCMC kernels cites Annealed Flow Transport Monte Carlo
- Asymptotically exact variational flows via involutive MCMC kernels cites Continual Repeated Annealed Flow Transport Monte Carlo
- Continual Repeated Annealed Flow Transport Monte Carlo cites Annealed Flow Transport Monte Carlo
Cited by top-tier papers3
- Learning Latent Variable Models via Jarzynski-adjusted Langevin AlgorithmJames Cuin, Davide Carbone, O. Deniz AkyildizNeurIPS 2025 · 4 citations
- Asymptotically exact variational flows via involutive MCMC kernelsZuheng Xu, Trevor CampbellNeurIPS 2025 · 2 citations
- Efficient Stochastic Optimisation via Sequential Monte CarloJames Cuin, Davide Carbone, Yanbo Tang, O. AkyildizICML 2026 · 1 citation
Builds on13
- Practical and Asymptotically Exact Conditional Sampling in Diffusion ModelsLuhuan Wu, Brian L. Trippe, Christian A. Naesseth, David M. Blei et al.NeurIPS 2023 · 276 citations
- Annealed Flow Transport Monte CarloMichael Arbel, Alexander G. de G. Matthews, Arnaud DoucetICML 2021 · 99 citations
- Diffusion Probabilistic Modeling of Protein Backbones in 3D for the motif-scaffolding problemBrian L. Trippe, Jason Yim, Doug Tischer, David Baker et al.ICLR 2023 · 96 citations
- Continual Repeated Annealed Flow Transport Monte CarloAlexander G. de G. Matthews, Michael Arbel, Danilo Jimenez Rezende, Arnaud DoucetICML 2022 · 69 citations
- Probabilistic Inference in Language Models via Twisted Sequential Monte CarloStephen Zhao, Rob Brekelmans, Alireza Makhzani, Roger Baker GrosseICML 2024 · 61 citations
Related papers
- Entropy-based adaptive Hamiltonian Monte CarloMarcel Hirt, Michalis K. Titsias, Petros DellaportasNeurIPS 2021 · 11 citations
- A connection between Tempering and Entropic Mirror DescentNicolas Chopin, Francesca R. Crucinio, Anna KorbaICML 2024 · 22 citations
- Test-time Alignment of Diffusion Models without Reward Over-optimizationSunwoo Kim, Minkyu Kim, Dongmin ParkICLR 2025
- AutoStep: Locally adaptive involutive MCMCTiange Liu, Nikola Surjanovic, Miguel Biron-Lattes, Alexandre Bouchard-Côté et al.ICML 2025
- Sequential Controlled Langevin DiffusionsJunhua Chen, Lorenz Richter, Julius Berner, Denis Blessing et al.ICLR 2025
