A note on a confidence bound of Kuzborskij and Szepesvári

  • 2021-01-12 18:57:23
  • Omar Rivasplata
  • 0

Abstract

In an interesting recent work, Kuzborskij and Szepesv\'ari derived aconfidence bound for functions of independent random variables, which is basedon an inequality that relates concentration to squared perturbations of thechosen function. Kuzborskij and Szepesv\'ari also established thePAC-Bayes-ification of their confidence bound. Two important aspects of theirwork are that the random variables could be of unbounded range, and notnecessarily of an identical distribution. The purpose of this note is toadvertise/discuss these interesting results, with streamlined proofs. Thisexpository note is written for persons who, metaphorically speaking, enjoy the"featured movie" but prefer to skip the preview sequence.

 

Quick Read (beta)

loading the full paper ...