The \(\varepsilon\)-strategy in variational analysis: illustration with the closed convexification of a function (Q2275342): Difference between revisions

The paper is devoted to analysing the possible relationship between the minimizers of a general variational (or optimization) problem (actually the \(\varepsilon\)-minimizers) and those of a relaxed version of it (by relaxation the authors mean here closed-convexification of the objective function). Problems of interest are those which do not have solutions necessarily but do have approximate solutions (or solutions within \(\varepsilon >0\)). The authors present how to recover all the (exact) minimizers of the relaxed version of the original problem in terms of the \(\varepsilon\)-minimizers of the original problem. Section 1 presents what the authors call the \(\varepsilon\)-strategy in mathematics, and show how to benefit from \(\varepsilon\)-perturbations of a given problem. Section 2 is devoted to the links between a general function \(J\) and its closed convex hull \(\overline{\text{co}}J\) in which it gives ways to derive \(\operatorname{argmin} (\overline{\text{co}}J)\) from the \(\varepsilon\)-\(\operatorname{argmin}(J)\)'s. Section 3 is devoted to some applications of results of the mentioned affirmation to some Hilbertian approximation-minimization problems.