aiqbench analysis
This paper proposes PRISM, a method for optimizing over a manifold that is known only through feasible samples. It replaces Riemannian gradient steps with iterative sampling from an objective-aware proximal distribution, and it claims descent, feasibility, and stationarity guarantees. The work sits closer to optimization and control theory than to mainstream ML, but its core is a learned sampling-based procedure, so it is most relevant to readers interested in data-driven optimization. The building-control and real-world experiments are the practical test of whether the guarantees hold outside synthetic settings.
Riemannian Optimization on Unknown Manifolds via Stochastic Proximal Refinement
arXiv:2610.41379v1 Announce Type: cross Abstract: We study Riemannian optimization on an unknown manifold accessed only through feasible samples. The central challenge is to turn feasible samples into optimization steps, i.e., the samples are drawn from the unknown feasible manifold, but do not directly provide the tangent spaces, projections, or retractions required by classic Riemannian methods. Moreover, regions with low objective values may be sparsely represented in the data, so directly sampling from a distribution fitted to these observations can require many draws to find better feasible solutions. We introduce Proximal Refinement on Implicit Smooth Manifolds (PRISM), which turns feasible-sample access into an iterative optimization procedure by repeatedly sampling an objective-aware proximal distribution conditioned on the current iterate. Theoretically, we show that PRISM recovers the effect of Riemannian-gradient refinement, leading to provable objective descent, feasibility (manifold adherence), and convergence toward first-order stationarity. We demonstrate the effectiveness of our approach through synthetic experiments illustrating its geometric properties and applications to data-driven building control and diverse real world optimization tasks, highlighting its potential as a general framework for data-driven optimization across scientific and engineering domains.
Abstract:We study Riemannian optimization on an unknown manifold accessed only through feasible samples. The central challenge is to turn feasible samples into optimization steps, i.e., the samples are drawn from the unknown feasible manifold, but do not directly provide the tangent spaces, projections, or retractions required by classic Riemannian methods. Moreover, regions with low objective values may be sparsely represented in the data, so directly sampling from a distribution fitted to these observations can require many draws to find better feasible solutions. We introduce Proximal Refinement on Implicit Smooth Manifolds (PRISM), which turns feasible-sample access into an iterative optimization procedure by repeatedly sampling an objective-aware proximal distribution conditioned on the current iterate. Theoretically, we show that PRISM recovers the effect of Riemannian-gradient refinement, leading to provable objective descent, feasibility (manifold adherence), and convergence toward first-order stationarity. We demonstrate the effectiveness of our approach through synthetic experiments illustrating its geometric properties and applications to data-driven building control and diverse real world optimization tasks, highlighting its potential as a general framework for data-driven optimization across scientific and engineering domains.
Systems and Control (eess.SY); Machine Learning (cs.LG); Optimization and Control (math.OC)
arXiv:2610.41379 [eess.SY]
(or arXiv:2610.41379v1 [eess.SY] for this version)
https://doi.org/10.48550/arXiv.2610.41379 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 279 words
Related coverage
Iso-Riemannian Optimization on Learned Data Manifolds
arXiv · Aug 19, 2026
Decentralized Online Riemannian Optimization for Strongly Geodesically Convex Functions
arXiv · Jul 23, 2026
Riemannian Optimization for Hadamard Products of Low-Rank Matrices
arXiv · Sep 1, 2026
Gradient based Bilevel for Inverse Optimal Control, a Riemannian approach
arXiv · Jun 10, 2026