Minimizing measures of risk by saddle point conditions (Q984899)

The authors start with a short but informative overview of risk measures and optimization problems with such measures. Following and extending an original idea of \textit{A. Balbás, R. Balbás} and \textit{S. Mayoral} [Eur. J. Oper. Res. 192, No.~2, 603--620 (2009; Zbl 1157.91350)] they get an equivalent optimization problem that is on the one hand differentiable (and often linear) but on the other hand for the dual problem (which has no duality gap under the considered conditions) one must deal with complicated problems in Banach spaces of measures. To overcome this difficulty they prove a mean value theorem, which simplifies the complications for the dual problem. As a consequence they can characterize the optimal solutions by saddle points using the subgradients for the dual problem of the risk measure to be optimized. Four applications taken from actuarial and financial mathematics are added.