Convergence rate analysis and error bounds for projection algorithms in convex feasibility problems (Q4664041)

The authors consider the rate of convergence of three basic projection type algorithms for solving the Convex Feasibility Problem (CFP). They give a new and simple proof, which relies on elementary convex analysis arguments, to show that Slater's condition implies Local Error Bound and thus implies linear rate of convergence. It has been demonstrated that projection algorithms for CFP can be very slow if Slater's condition is not satisfied. Moreover, without Slater's condition it is shown that one cannot bound the rate of convergence. Projection algorithms for convex feasibility problems with two sets are revisited through the use of gradient projection method applied to a convex minimization reformulation of CFM.