Scribd
Upload
6 views2 pages

2020 Strasbourg Exercises

The document is an exercise sheet for a Probability Master Class focusing on Random Euclidean Bipartite Matching Problems. It presents a series of problems related to matching costs, Gaussian random variables, bounded differences, order statistics, and Poisson random variables. Each problem requires mathematical proofs and derivations based on the provided definitions and hints.

Uploaded by

crosrol8
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views2 pages

2020 Strasbourg Exercises

The document is an exercise sheet for a Probability Master Class focusing on Random Euclidean Bipartite Matching Problems. It presents a series of problems related to matching costs, Gaussian random variables, bounded differences, order statistics, and Poisson random variables. Each problem requires mathematical proofs and derivations based on the provided definitions and hints.

Uploaded by

crosrol8
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Probability Master Class 2020

Exercise Sheet on Random Euclidean Bipartite Matching Problems


Please refer to the lecture notes if some definitions are not clear.

Let d ≥ 1 and consider i.i.d. uniform random variables (Xi )∞ ∞


i=1 , (Yi )i=1 with values in [0, 1]
d

(if not otherwise stated). For n ≥ 1, p > 0 write


n
X
Bp,n := minn |Xi − Yσ(i) |p
σ∈S
i=1
n
X
Mp,n = min |Xσ(i) − Xσ(n+i) |p ,
σ∈S 2n
i=1

for the minimum bipartite and monopartite matching costs (S k denotes the set of permutations
over {1, . . . , k}).

Problem 1
Show that E [Mp,n ] ≤ E [Bp,n ].

Problem 2
Show that there exists c(d, p) > 0 such that, for every m ≥ 1, x ∈ [0, 1]d .
 
c(d, p)
E min |Xi − x|p ≤ .
i=1,...,m mp/d
(Hint: follow the proof of the converse inequality and try to bound from above instead of
below: where should be located x ∈ [0, 1]d so that [0, 1]d \ B(x, t1/p ) is maximized?)

Problem 3
Show that for some c(d, p) > 0 one has E [Mp,n ] ≥ c(d, p)n1−p/d .

Problem 4
Let Z be a Gaussian random variable with mean m and variance σ 2 . Then
r2
 
2σ
P (|Z − m| > r) ≤ √ exp − 2 for every r > 0.
2πr 2σ
(Hint: reduce to a standard case first, m = 0, σ 2 = 1, and then compute by a change of
variables Z ∞ Z ∞ Z ∞
2 2 2 2
e−x /2 dx = e−r /2 e−x /2 e−xr dx ≤ e−r /2 e−xr dx.
r 0 0

Problem 5
Let E be a set, m ≥ 1 and say that f : E m → R has bounded differences if there exists
(di )m
i=1 such that

for every i ∈ {1, . . . , m} there exists di ≥ 0 such that


|f (x1 , . . . , xi−1 , xi , xi+1 , . . . , xm ) − f (x1 , . . . , xi−1 , x0i , xi+1 , . . . , xm )| ≤ di
for every x1 , . . . , xi−1 , xi , xi+1 , . . . , xm , x0i ∈ E.
Let (fu )u∈U be a family of non-negative functions on E m having bounded differences (and
(di )di=1 does not depend on u ∈ U ). Then, inf u∈U fu also has bounded differences.

Pag. 1
Problem 6
Recall the definition of order statistics for a set of n different points (zi )ni=1 ⊆ R:

z(1) = min {zi } , z(k+1) = min zi : zi > z(k) .

Recall that we introduce the matching via order statistics between (xi )ni=1 , (yi )ni=1 as the
permutation such that n on  n
(xi , yσ† (i) ) = (x(i) , y(i) ) i=1 .
i=1

Give an explicit example showing that σ † is not necessarily an optimal matching if the
distance is raised to power p with 0 < p < 1.

Problem 7
Find an expression for E [B4,n ] and show that the following limit exists:

lim nE [B4,n ] .
n→∞

Problem 8
Show that a Poisson random variable Z with parameter λ has mean λ and variance λ and
that, for α ∈ (0, 1], λ ≥ 1,
E [Z α ] ≥ c(α)λα ,

where c(α) > 0 depends on α only. (Hint: use the inequality Z α ≥ λα − |Z − λ|α )

Pag. 2

You might also like