1: \begin{abstract}

2: The problem of finding a point in the intersection of closed sets

3: can be solved by the method of alternating projections and its variants.

4: It was shown in earlier papers that for convex sets, the strategy

5: of using quadratic programming (QP) to project onto the intersection

6: of supporting halfspaces generated earlier by the projection process

7: can lead to an algorithm that converges multiple-term superlinearly.

8:  The main contributions of this paper are to show that this strategy

9: can be effective for super-regular sets, which are structured nonconvex

10: sets introduced by Lewis, Luke and Malick. Manifolds should be approximated

11: by hyperplanes rather than halfspaces. We prove the linear convergence

12: of this strategy, followed by proving that superlinear and quadratic

13: convergence can be obtained when the problem is similar to the setting

14: of the Newton method. We also show an algorithm that converges at

15: an arbitrarily fast linear rate if halfspaces from older iterations

16: are used to construct the QP.

17: \end{abstract}

18: