The optimal parameters of candidate measures established by importance sampling are usually obtained by minimizing the quadratic criterion. Efficient importance sampling (EIS) implements such quadratic criterion by a generic procedure known as generalized least squares or its modification as weighted least squares. Firstly, EIS is extended by changing the optimal criterion from quadratic criterion to p-order (P >= 1) relative centered moment criterion. Secondly, two methods are developed to get the optimal parameters of candidate measures. The first one is also solving a weighted least squares...
