English

Improved Hoeffding's Lemma and Hoeffding's Tail Bounds

Probability 2020-12-08 v1 Signal Processing

Abstract

The purpose of this letter is to improve Hoeffding's lemma and consequently Hoeffding's tail bounds. The improvement pertains to left skewed zero mean random variables X[a,b]X\in[a,b], where a<0a<0 and a>b-a>b. The proof of Hoeffding's improved lemma uses Taylor's expansion, the convexity of exp(sx),sR\exp(sx), s\in {\bf R} and an unnoticed observation since Hoeffding's publication in 1963 that for a>b-a>b the maximum of the intermediate function τ(1τ)\tau(1-\tau) appearing in Hoeffding's proof is attained at an endpoint rather than at τ=0.5\tau=0.5 as in the case b>ab>-a. Using Hoeffding's improved lemma we obtain one sided and two sided tail bounds for P(Snt)P(S_n\ge t) and P(Snt)P(|S_n|\ge t), respectively, where Sn=i=1nXiS_n=\sum_{i=1}^nX_i and the Xi[ai,bi],i=1,...,nX_i\in[a_i,b_i],i=1,...,n are independent zero mean random variables (not necessarily identically distributed). It is interesting to note that we could also improve Hoeffding's two sided bound for all {Xi:aibi,i=1,...,n}\{X_i: a_i\ne b_i,i=1,...,n\}. This is so because here the one sided bound should be increased by P(Snt)P(-S_n\ge t), wherein the left skewed intervals become right skewed and vice versa.

Keywords

Cite

@article{arxiv.2012.03535,
  title  = {Improved Hoeffding's Lemma and Hoeffding's Tail Bounds},
  author = {David Hertz},
  journal= {arXiv preprint arXiv:2012.03535},
  year   = {2020}
}
R2 v1 2026-06-23T20:46:26.275Z