Constraint qualifications and optimality conditions for optimization problems with cardinality constraints (Q344949)

Optimization problems with cardinality constraints are problems with inequality and equality constraints where additionally the number of nonzero components of the solution is bounded. These problems can equivalently be formulated as mixed-integer optimization problems. Topic of the article is the investigation of problem-tailored modifications of different constraint qualifications and stationarity conditions for the linear relaxation of this mixed-integer problem. To formulate the stationarity conditions, a tightened nonlinear optimization problem for the optimization problem with cardinality constraints is used. A local minimum of the linear relaxation is an S-stationary point if the weakest of the formulated constraint qualifications is satisfied or if the constraints are linear. If the inequality constraints include nonnegativity constraints for the variables, a special case of a mathematical program with complementarity constraints (MPCC) arises. For this problem, the four different stationarity concepts for MPCCs (W-, C-, M-, S-stationarity) reduce to only two, and the MPCC-LICQ is likely to be violated at a solution.