A geometric approach to portfolio optimization in models with transaction costs (Q1887272)

This note is aimed to show that convex analysis provides a natural language to treat multi-asset investment-consumption models with transaction costs. Moreover, convex analysis allows to replace tedious computations by appealing to elementary geometric properties. The authors are looking for a compromise between generality and ``accessibility'' of results and restrict themselves to the framework which is adequate to cover a currency market model. The main results are assertions that the Bellman function is a viscosity solution of an HJB equation and that this equation has a unique solution. It is shown, in particular, that existence of the Lyapunov function together with the monotonicity of the dual of the utility function (with respect to the partial ordering induced by the dual to the solvency cone) are the only properties needed to guarantee the existence result.