|
abstract:
We study a class of stochastic dynamic programming problems which arise in nonlinear option pricing problems (e.g. due to early exercise features, credit value adjustment, or model risk). This class of problems also appears in time discretization schemes for (reflected) backward stochastic differential equations and fully nonlinear second order parabolic PDEs. Generalizing the primal-dual methodology, which is popular in Bermudan option pricing, we design several Monte-Carlo algorithms for the construction of confidence intervals for the value of the dynamic program. |
