Abstract
In this talk, we propose an importance-sampling based method to obtain an unbiased simulator to evaluate expectations involving random variables whose probability density functions are unknown while their Fourier transforms have an explicit form. We give a general principle about how to choose appropriate importance samplers under different models. Compared with the existing methods, our method avoids time-consuming numerical Fourier inversion and can be applied effectively to high dimensional financial applications such as option pricing and sensitivity estimation under Heston stochastic volatility model, high dimensional affine jump-diffusion model, and various Levy processes.
This is joint work with N. Chen and Y. Wang.