Understanding Machine Learning: From Theory to Algorithms

(Jeff_L) #1

420 Technical Lemmas


Proof For alli= 0, 1 , 2 ,...denoteti=a(i+


log(b)). Sincetiis monotonically
increasing we have that

E[|X−x′|]≤a


log(b) +

∑∞

i=1

tiP[|X−x′|> ti− 1 ].

Using the assumption in the lemma we have

∑∞

i=1

tiP[|X−x′|> ti− 1 ]≤ 2 ab

∑∞

i=1

(i+


log(b))e−(i−1+


log(b))^2

≤ 2 ab

∫∞

1+


log(b)

xe−(x−1)

2
dx

= 2ab

∫∞


log(b)

(y+ 1)e−y

2
dy

≤ 4 ab

∫∞


log(b)

ye−y

2
dy

= 2ab

[

−e−y

2 ]∞


log(b)
= 2ab/b= 2a.

Combining the preceding inequalities we conclude our proof.

lemmaA.5 Letm,dbe two positive integers such thatd≤m− 2. Then,

∑d

k=0

(

m
k

)


(em
d

)d
.

Proof We prove the claim by induction. Ford= 1 the left-hand side equals
1 +mwhile the right-hand side equalsem; hence the claim is true. Assume that
the claim holds fordand let us prove it ford+ 1. By the induction assumption
we have

∑d+1

k=0

(

m
k

)


(em
d

)d
+

(

m
d+ 1

)

=

(em
d

)d(
1 +

(

d
em

)d
m(m−1)(m−2)···(m−d)
(d+ 1)d!

)


(em
d

)d(
1 +

(

d
e

)d
(m−d)
(d+ 1)d!

)

.
Free download pdf