Random sampling with removal (Q2207595)

Randomized algorithms are applied to the solution of constrained optimization problems in abstract spaces. Approximate solutions are defined by replacing the set of constraints by random samples of constraints. A variant of this procedure is obtained by removing a certain number of constraints from the sample according to fixed rules. The tradeoff between improving the value of the objective function at decreasing the accuracy of the approximate solution occurring from removing constraints is studied. A main goal of the present article is to apply the random sampling method with removal to chance-constrained optimization problems. The models and methods are presented in abstract setting for discrete problems and with some extensions to continuous cases.