Simulation of forward-reverse stochastic representations for conditional diffusions (Q744383)

This paper develops a new approach for the simulation of conditional diffusion (diffusion bridges) expressed by the expectation of some function of a diffusion process given; it lies in some set at certain time points. The method is based on a forward-reverse estimator for a transition density. The corresponding Monte Carlo estimators are shown to have convergence based on detailed convergence analyses. Implementation issues are discussed based on truncated kernels and some numerical simulations are carried out including a stochastic volatility asset Heston model.