Unbiased Optimal Stopping via the MUSE

We propose a new unbiased estimator for estimating the utility of optimal stopping problem. The MUSE, short Multilevel Unbiased Stopping Estimator, constructs Monte Carlo (MLMC) at every stage problem in backward recursive way. In contrast to traditional sequential methods, MUSE can be implemented parallel. prove has finite variance, computational complexity, and achieves $\epsilon$-accuracy with $O(1/\epsilon^2)$ cost under mild conditions. demonstrate empirically an option pricing involving high-dimensional input use many parallel processors.