Robust Importane Sampling for Bounded Equivocal Systems

摘要

Importance samplingis a fast estimation technique based on a biased form of Monte Carlo simulation. Biasing techniquesin the literature assume that the system under consideration is fully and accurately characterized. WHile this can be a reasonable for some problems, in prctice it is frequently the case that the systems of interest are significantly differnet form the models employed in importance sampling simulation design. Because of this, importance sampling bisaing strategies derived with respectot these erroneous system models might not noly lead to unrealistic simulation performnace, but also to potentially biased estimators.