Emanuel Olariu

We present an alternative approach to the problem of estimatingprobabilities of rare events and for optimization problems using the classof Rényi divergences of order α > 1. The general procedure wedescribe does not involve any specific family of distributions, the onlyrestriction is that the search space consists of product form probabilitydensity functions. We discuss an algorithm for estimation of probability ofrare events and a version for continuous optimization. The results ofnumerical experimentation with these algorithms carried in the last sectionsupport their performances.

Full Document (PDF)

Bibtex

@TechReport{isRd,
    author = "Emanuel Olariu",
    title = "{Importance Sampling using {R'{e}nyi} divergence}",
    institution = "``Al.I.Cuza'' University of Ia{c s}i, Faculty of Computer Science",
    year = "2012",
    number = "TR 12-02",
    url = "https://publications.info.uaic.ro/technical-reports/archive/tr12-02-2012-importance-sampling-using-rnyi-divergence/"
}