In order to dealf with large scale, high dimension and rare event problem, we apply the efficient importance sampling method for computing Value--at--Risk of a portfolio. Firstly, building on Delta--Gamma approximations to changes in portfolio value and by replacing the initial sampler by an auxiliary parametric importance sampler, the optimal choices for the sampling parameters are given by solving a nonlinear Generalized Least Squares problem. Secondly, under the assumption of density kernels within the exponential family of distributions, the problem is further simplified to solving a linea...
