Abstract
We introduce a game formulation for stochastic target problems under controlled loss, and prove a weak version of the Geometric Dynamic Programming principle which allows to provide a Pde characterization of the associated viability set. We will show how this approach allows one to provide a direct PDE characterization of prices of contingent claims computed under an expected loss constraints, in situations where some parameters (e.g. volatility) are unknown/uncertain.
[PDF] Slides of the presentation.