Solving moment and polynomial optimization problems on Sobolev spaces

Didier Henrion,Alessandro Rudi
2024-01-15
Abstract:Using standard tools of harmonic analysis, we state and solve the problem of moments for positive measures supported on the unit ball of a Sobolev space of multivariate periodic trigonometric functions. We describe outer and inner semidefinite approximations of the cone of Sobolev moments. They are the basic components of an infinite-dimensional moment-sums of squares hierarchy, allowing to solve numerically non-convex polynomial optimization problems on infinite-dimensional Sobolev spaces, with global convergence guarantees.
Optimization and Control
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the moment problem of measures supported on Sobolev spaces. Specifically, the author uses standard tools in harmonic analysis to define and solve the moment problem of positive measures supported on the unit ball of Sobolev spaces of multivariate periodic trigonometric functions. The paper describes the extrinsic and intrinsic semidefinite approximations of the Sobolev moment cone, which are the basic components of the infinite - dimensional moment - sum - of - squares (moment - SOS) hierarchy, allowing for the numerical solution of non - convex polynomial optimization problems on infinite - dimensional Sobolev spaces and providing global convergence guarantees. In short, the core problems of the paper are: - **How to effectively solve the moment problem on Sobolev spaces**: This involves how to numerically solve the moments of measures supported on Sobolev spaces, especially positive measures on the unit ball. - **How to use the moment - SOS hierarchy to solve non - convex polynomial optimization problems**: By constructing the extrinsic and intrinsic semidefinite approximations of Sobolev moments, an infinite - dimensional moment - SOS hierarchy can be constructed to solve non - convex polynomial optimization problems and ensure global convergence. The solutions to these problems are of great significance for expanding the application range of the moment - SOS hierarchy, especially when dealing with non - convex nonlinear optimization problems on infinite - dimensional functional spaces.