Mathematical programming with system of equilibrium constraints (Q995941): Difference between revisions

The paper of Lai-Jiu Lin is located on the interface between the theoretical areas of calculus, in fact, multivalued calculus, and functional analysis, and the more applied areas of continuous optimization theory and game theory. For this purpose a unifying framework is chosen, with a special emphasis and utilization of bilevel programming. Indeed, differently from common optimization problems, two vector- or even family-valued variables \(x\) and \(y\) appear -- on a same stage, in some parallelism, but also at different stages levels, with dependencies (as with leader and follower in dynamical processes or games played, or in semi-infinite optimization). Constraints occur on the two variables \(x\) and \(y\) both jointly and separately, respectively, sometimes even separately for the components or members of \(x\) and \(y\). The dependence of \(x\) on \(y\) can be generalized by further conditions imposed on \(y\) via set-valued mappings of \(x\). This phenomenon is also known from so-called generalized semi-infinite optimization. The topological setting of the paper is given by Hausdorff spaces (sometimes, product spaces), and, especially, Hausdorff locally convex topological vector spaces. In this article, the author studies the mathematical program with a system of equilibrium constraints. This problem contains a bilevel program with a system of equilibrium constraints, a semi-infinite program with a system of equilibrium constraints, mathematical program with Nash equilibrium constraints, a mathematical program with system of mixed variational like inequality constraints. He establishes the existence theorems of a mathematical program with a system of equilibrium constraints under various assumptions. Based on the different roles and (in)dependencies of \(x\) and \(y\), there are naturally being given feasible sets and index sets (or index ``manifolds'') of different kinds and meanings. A first kind of approach to them is given in terms of convexity and generalized kinds of convexity, e.g., quasi-convexity and, furthermore, projective, intersection, upper continuity and graph properties and representations. The objective function in \(x\) and \(y\) jointly is considered to be lower semi-continuous, whereas constraint functions are often supposed to be upper semi-continuous. The exemplification of these strong existence results is given by the variety of problem classes and the assumptions made. Indeed, the relations between those kinds of problems are discussed and exploited. So the examples are of a theoretical nature. However, for sure, many important real-world motivations and examples could be given, too. Indeed, this paper can serve as a basis for future utilizations in many areas of modern science, in the areas of Operations Research and Statistical Learning.