Two-stage stochastic standard quadratic optimization

• First treatment of standard quadratic optimization problems under uncertainty with two-stage stochastic approach. Efficient and tractable upper lower bounds for this NP-hard problem. Modern first-order method (pairwise Frank-Wolfe) enhanced copositive techniques aim at scalability bounds. Two-stage decision are characterized by the property that in stage one part has to be made before relevant data observed rest can two after available. It is generally assumed while missing not known, their probability distribution known. In paper, we consider a special form such where objective function first second decisions goal minimize expected cost function. We assume must lie simplex, but do require convexity uncertain objective. well known programs NP-complete thus belong most complex problems. A viable attack find original complicated problem solving easier approximate describe approximations show exploiting properties, good necessarily optimal solutions found. also present possible applications, as selecting invited speakers conferences generate community high coherence. Finally, systematic numerical results provided.