New results in importance sampling [of false alarm statistics]

Importance sampling is a technique which can significantly reduce the number of Monte Carlos necessary to accurately estimate the probability of low-probability of occurrence events (e.g., the probability of false alarm P/sub F/ associated with a given detection scheme). A new technique called the Chernoff Importance Sampling Method is introduced. It is shown that the number of required Monte Carlos can be reduced by a factor of a Chernoff-like bound on P/sub F/. In addition, techniques for choosing the multiplying factor of the distorted variance method (the most common method used in importance sampling) are presented.