Convergence criteria for generalized gradient methods of solving locally Lipschitz feasibility problems (Q1803652): Difference between revisions

The authors prove two theorems giving sufficient conditions for the convergence of some iterative algorithms based on the idea of a generalization of the gradient method, applied to two subsequent problems: a) Compute solutions of a system of inequalities \(f_ i(x)\leq 0\), (\(i\in I\)), where \(I\) is a finite set and, for each \(i\in I\), \(f_ i\) is a continuous real functional on \(\mathbb{R}^ n\). If all functionals \(f_ i\), (\(i\in I\)) are locally Lipschitz ones (resp. convex ones), this problem is called a locally Lipschitz (resp., convex) feasibility problem. The so-called synchronized maximal-functions reduction (SMFR) algorithms, being the matter of the first theorem, may be used to solve the problem. b) Compute solutions of the particular feasibility problem of the form \(d^ 2_{Q_ i}(x)\leq 0\), (\(i\in I\)), where \(Q_ i\), (\(i\in I\)) is a family of closed convex subsets of \(\mathbb{R}^ n\) which common intersection has nonempty interior, \(d^ 2_{Q_ i}(x)\) denotes the square of the distance of the point \(x\) to the subset \(Q_ i\). This problem is called a convex intersection problem and the so-called synchronized maximal- distance reduction (SMDR) methods (of the block-iterative projection type) may be used to solve it (this is the matter of the second theorem). The authors also discuss some computational aspects of the SMFR and SMDR algorithms and give a numerical example.