Stochastic approximation algorithms for superquantiles estimation

Abstract

In this talk, we consider different two-time-scale stochastic approximation algorithms for quantile and superquantile estimation.We study the asymptotic behavior of these algorithms (almost sure convergence, quadratic strong law and law of iterated logarithm, TCL). We also derive non-asymptotic controls of the mean quadratic error of the algorithms. This talk is based on joint work with B. Bercu and S. Gadat.