A projection method for least-squares solutions to overdetermined systems of linear inequalities (Q1821512)

This paper is the third in a series of papers of the author [Appl. Math. Optimization 10, 247-265 (1983; Zbl 0524.90072); Linear Algebra Appl. 65, 45-62 (1985; Zbl 0564.65044)]. Let \(<x,u_ i>\leq b_ i\), \(i=1,...,n\) be a system of linear inequalities and \(C_ i=\{x;<x,u_ i>\leq b_ i\}\), \(C=C_ 1\cap C_ 2\cap...\cap C_ n\). In paper I the author introduced a primal-dual algorithm for finding a point \(x\in C\). In paper II the author proved that int(C)\(\neq \emptyset\) implies, for some k, \(x_ k=x_{k+1}=..\). where \(x_ k\in C\) \((x_ k\) generated by the algorithm). In this paper the author obtains elegant results for the cases 1) \(int(C)=\emptyset\) but \(C\neq \emptyset\), 2) \(C=\emptyset\). The sequence \((x_ k)\) generated by algorithm converges to a solution and \(dist(x_ k,X)\leq c\mu^ k\) for some \(c>0\) and \(0\leq \mu \leq 1\) (convergence at a linear rate to the set X of least-squares solutions). An extensive bibliography is provided.