Hoeffding’s Inequality:
Suppose x1,x2,……xN are independent random variables, and xi∈[ai,bi], i=1,2,……,N; Xˉ is the expirical mean of x1,x2,……xN, namely, Xˉ=N1∑i=1Nxi, then for any t>0, the following inequality holds:
P[Xˉ−E(Xˉ)≥t]≤exp(−∑i=1N(bi−ai)22N2t2)
P[E(Xˉ)−Xˉ≥t]≤exp(−∑i=1N(bi−ai)22N2t2)
Hoeffding’s lemma:
Suppose x is an random variable, and x∈[a,b], and E(x)=0, then for any t>0,the following inequality holds:
E(etx)≤exp8t2(b−a)2
We prove the lemma first:
Obviously, f(x)=etx is a convex function, so for any α∈[0,1], we have:
f(αx1+(1−α)x2)≤αf(x1)+(1−α)f(x2)

Let α=b−ab−x,∀x∈[a,b], then αx1+(1−α)x2=x, so we have:
etx≤b−ab−xeta+b−ax−aetb,∀x∈[a,b]
take expectations on both sides, and because of E(x)=0, there is:
E(etx)≤b−abeta−b−aaetb
Try to simplify it.
Let p=−b−aa, then right=(1−p)e−t(b−a)p+pe−t(b−a)(p−1),
let h=t(b−a), then right=(1−p)e−hp+peh(1−p)=e−hp(1−p+peh), note the function L(h)=−hp+ln(1−p+peh),h>0, now we need to prove L(h)≤8t2(b−a)2=8h2.
Give two ways to evident this inequality above:
-
Construct function, g(h)=L(h)−8h2=−hp+ln(1−p+peh)−8h2,h>0
Obviously, g(0)=0,g′(0)=0,g′′(h)=[(1−p)+(peh)]2(1−p)peh−41≤0, according to the monotonicity, g(h)≤0, namely L(h)≤8h2=8t2(b−a)2.
-
Apply taylor expansion to function g(h), g(h)=0+0+2!g′′(0)(h−0)2+o(h2)=([(1−p)+(peh)]2(1−p)peh−41)h2+o(h2)≤0, namely L(h)≤8h2=8t2(b−a)2.
So the hoeffding’s lemma is true.
To be continue……
reference:
https://www.cnblogs.com/kolmogorov/p/9518867.html
http://dict.cnki.net/
李航《统计学习方法》 北京:清华大学出版社,2019