Subsmooth semi-infinite and infinite optimization problems (Q715070): Difference between revisions

The article of Xi Yin Zheng and Kung Fu Ng is a very valuable contribution to the field of nonlinear and continuous optimization with a great promise for generalized frameworks of the usual constrained programs by allowing (i) an infinite number of inequality constraints to be given, or, including the objective function now also, (ii) to be defined over an infinite dimensional space. Those optimization problems are called (i) semi-infinite and (ii) infinite, respectively. They play an increasing role in the representation, treatment and solution of problems from (i) approximation of real-world phenomena, e.g. through Chebychev approximation, in design, manufacturing, robotics, computational biology, etc., and (ii) calculus of variations and data mining, etc., respectively. Both (i) and (ii) are useful in optimal control and, e.g. via the new ``infinite kernel learning'' in machine learning. On the other hand, the article also contributes to nonsmooth optimization; in fact, it lies in the definition of problems from (i) and (ii), that max- or sup-type functions occur, which are nonsmooth usually. Of course, such nonsmooth functions are still rather well-structured and -behaving ones, compared with even more general nonsmooth functions. In fact, the authors apply the concept of subsmoothness (introduced for sets and function families as well) to rigorously represent and study the problems from (i) and (ii). This study is done in a wide functional-analytic setting and with great rigor. Here, a particular interest of the authors exists in strong isolated and strong unique local minimizers; they use the notion of a sharp minimum in their related investigations. First, the authors consider subsmoothness for a function family and present formulas of the subdifferential of the pointwise supremum of a such a family. Then, they consider subsmooth infinite and semi-infinite optimization problems. Especially, they are able to state several dual and primal characterizations for a point to be a sharp minimum of a weak sharp minimum for such optimization problems. The five sections of the article are as follows: 1. Introduction, 2. Preliminaries, 3. Subsmoothness for a function family, 4. Subsmooth infinite optimization problem, and 5. Subsmooth semi-infinite optimization problem. Further deep results and also methods may be expected in the future, initiated and fostered by this research paper. Such advances could then support and stimulate additional advances in science, especially, in data mining and statistics, in engineering, economics and social-political decision making, in finance and OR, in healthcare and medicine, and, herewith, to improvements in the living conditions of the peoples on earth.