Efficient Stochastic Optimisation via Sequential Monte Carlo

James Cuin, Davide Carbone, Yanbo Tang, O. Deniz Akyildiz
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:22115-22159, 2026.

Abstract

The problem of optimising functions with intractable gradients frequently arises in machine learning and statistics, ranging from maximum marginal likelihood estimation procedures to fine-tuning of generative models. Stochastic approximation methods for this class of problems typically require inner sampling loops to obtain (biased) stochastic gradient estimates, which rapidly becomes computationally expensive. In this work, we develop sequential Monte Carlo (SMC) samplers for optimisation of functions with intractable gradients. Our approach replaces expensive inner sampling methods with efficient SMC approximations, which can result in significant computational gains. We establish convergence results for the basic recursions defined by our methodology which SMC samplers approximate. We demonstrate the effectiveness of our approach on the reward-tuning of energy-based models within various settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cuin26a, title = {Efficient Stochastic Optimisation via Sequential {M}onte {C}arlo}, author = {Cuin, James and Carbone, Davide and Tang, Yanbo and Akyildiz, O. Deniz}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {22115--22159}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/cuin26a/cuin26a.pdf}, url = {https://proceedings.mlr.press/v306/cuin26a.html}, abstract = {The problem of optimising functions with intractable gradients frequently arises in machine learning and statistics, ranging from maximum marginal likelihood estimation procedures to fine-tuning of generative models. Stochastic approximation methods for this class of problems typically require inner sampling loops to obtain (biased) stochastic gradient estimates, which rapidly becomes computationally expensive. In this work, we develop sequential Monte Carlo (SMC) samplers for optimisation of functions with intractable gradients. Our approach replaces expensive inner sampling methods with efficient SMC approximations, which can result in significant computational gains. We establish convergence results for the basic recursions defined by our methodology which SMC samplers approximate. We demonstrate the effectiveness of our approach on the reward-tuning of energy-based models within various settings.} }
Endnote
%0 Conference Paper %T Efficient Stochastic Optimisation via Sequential Monte Carlo %A James Cuin %A Davide Carbone %A Yanbo Tang %A O. Deniz Akyildiz %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-cuin26a %I PMLR %P 22115--22159 %U https://proceedings.mlr.press/v306/cuin26a.html %V 306 %X The problem of optimising functions with intractable gradients frequently arises in machine learning and statistics, ranging from maximum marginal likelihood estimation procedures to fine-tuning of generative models. Stochastic approximation methods for this class of problems typically require inner sampling loops to obtain (biased) stochastic gradient estimates, which rapidly becomes computationally expensive. In this work, we develop sequential Monte Carlo (SMC) samplers for optimisation of functions with intractable gradients. Our approach replaces expensive inner sampling methods with efficient SMC approximations, which can result in significant computational gains. We establish convergence results for the basic recursions defined by our methodology which SMC samplers approximate. We demonstrate the effectiveness of our approach on the reward-tuning of energy-based models within various settings.
APA
Cuin, J., Carbone, D., Tang, Y. & Akyildiz, O.D.. (2026). Efficient Stochastic Optimisation via Sequential Monte Carlo. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:22115-22159 Available from https://proceedings.mlr.press/v306/cuin26a.html.

Related Material