Finite Convergence of Circumcentered-Reflection Method on Closed Polyhedral Cones in Euclidean Spaces (2412.00512v1)
Abstract: The Circumcentered Reflection Method (CRM) is a recently developed projection method for solving convex feasibility problems. It offers preferable convergence properties compared to classic methods such as the Douglas-Rachford and the alternating projections method. In this study, our first main theorem establishes that CRM can identify a feasible point in the intersection of two closed convex cones in (\mathbb{R}2) from any starting point in the Euclidean plane. We then apply this theorem to intersections of two polyhedral sets in (\mathbb{R}2) and two wedge-like sets in (\mathbb{R}n), proving that CRM converges to a point in the intersection from any initial position finitely. Additionally, we introduce a modified technique based on CRM, called the Sphere-Centered Reflection Method. With the help of this technique, we demonstrate that CRM can locate a feasible point in finitely many iterations in the intersection of two proper polyhedral cones in (\mathbb{R}3) when the initial point lies in a subset of the complement of the intersection's polar cone. Lastly, we provide an example illustrating that finite convergence may fail for the intersection of two proper polyhedral cones in (\mathbb{R}3) if the initial guess is outside the designated set.