A note on the existence of nonsmooth nonconvex optimization problems (Q481039)

The note presents a class of minimization problems which do not allow to guarantee the existence of a solution, since the cost is neither convex nor coercive. The authors provide a class of problems and corresponding conditions that imply the existence of their minimizers. The general setting has the form \[ \min_{x\in X}G(x), \] where \(G\) is an extended, real-valued functional on a Banach space \(X\). The mapping \(G\) is not necessarily convex or smooth. The existence of solutions is obtained by considering the family of problems \[ \min_{x\in X}G(x)=G(x)+\frac{\epsilon}2|x|^2,\;\epsilon>0,\,\epsilon\to 0. \] Some applications to sparsity-constrained optimization and to unilateral problems from mechanics are also given.