Combined relaxation methods for variational inequalities (Q1586192)

The book is devoted to a new general approach to constructing solution methods for variational inequalities, which has been called combined relaxation (CR) approach. This approach is based on combining, modifying and generalizing ideas contained in various relaxation methods. Each combined relaxation method has a two-level structure, i.e., a descent direction and a stepsize at each iteration are computed by finite relaxation procedures. These parameters enables one to determine a hyperplane separating the current iterate and the solution set. Varying the rules of choosing the parameters and auxiliary procedures, one can obtain a number of algorithms within this framework. The CR allows one to construct methods both for single-valued and for multi-valued variational inequalities, including nonlinearly constrained problems. The other essential feature of all CR methods is that they are convergent under very weak assumptions. These assumptions are weaker than the pseudomonotonicity of the underlying mapping. Various rates of convergence results have been established for most CR methods. In general, they correspond to linear convergence. The book can be viewed as an attempt to present the existing combined relaxation methods as a whole. In Chapters 1 and 2, the author describes various CR methods for variational inequalities with single-valued and multivalued mappings, respectively. The first sections of both the chapters contain the corresponding theoretical background. Chapter 3 is devoted to applications of CR methods. In Chapter 4 necessary auxiliary results are presented.