Summary: | Discarding constraints in scenario optimization, a technique known as the sampling-and-discarding scheme, allows the
decision maker to trade feasibility to performance. Recently, a removal scheme with a less conservative bound on the constraint
violation probability of the final decision has been proposed. In this letter, we further contribute to the theoretical properties
of such a scheme by extending the number of discarded scenarios to be arbitrary, as opposed to an integer multiple of the
dimension of the decision space. There are two facets to the results of this paper. On the one hand, our feasibility guarantees
outperform the standard “sampling-and-discarding” bound in the literature. On the other hand, we highlight an inherent
property of the discarding mechanism, namely, the fact that removing a number of scenarios that is not an integer multiple
of the dimension of the decision space is likely to introduce additional conservatism.
|