You are on page 1of 3

HW Problem 5

Christopher
July 21, 2018

1 P RIVACY ON AN n X n GRID
1.1 P RIVACY ON A 4 X 4 GRID
1.1.1 C LOSED FORM SOLUTION

Rather than consider the 16 squares on the grid, consider the 24 shared edges. Each edge
is adjacent to exactly 2 squares. When privacy is not guaranteed, it is because, for at least
one edge, the adjacent squares contain arrows which both point at that edge. Clearly, the
probability of privacy is identically distributed for all edges. From simple counting, we have:

1
P (NonPrivacy on edge i ) = P (n i ) = (1.1)
16
We note that:

P (Privacy on the grid) = 1 − P (NonPrivacy on the grid)


24
[ (1.2)
= 1−P( ni )
i =1

Using the identity P (a∪b) = P (a)+P (b)−P (a∩b), this decomposes simply into, for iid events,
à !µ ¶
X24
i +1 24 1 i
(−1) (1.3)
i =1 i 16

Here we note some specificities of our problem. Namely, the events n i are not iid. Since each
element of {n i } uniquely specifies the contents of two squares in the grid, there are subsets of
(n j , n k ) ∈ {n i } such that P (n j ∩n k ) = 0. Call the set of these subsets F . In particular, all subsets

1
of {n i } with more than 8 members are elements of F . Applying this reasoning to the above we
have:
(Ã ! )µ ¶
X 8
i +1 24 1 i
P (NonPrivacy on the grid) = (−1) − f (i ) (1.4)
i =1 i 16
where f (i ) = number of subsets of {n i } of size i ∈ F (1.5)

It is relatively simple, albeit tedious, to enumerate the values of f (i ) members using a script
and conclude.

f (1) = 0
f (2) = 52
f (3) = 980
f (4) = 8033
f (5) = 39116
f (6) = 132446
f (7) = 345552
f (8) = 735435

From this we can compute the probability as

P (Privacy on the grid) ≈ 1 − .84342 (1.6)


≈ .15658 (1.7)

1.1.2 C OMPUTATIONAL A PPROXIMATION

As a form of validation, we construct a monte carlo simulation of the grid. Running the sim-
ulation 5 times, and generating 10,000 grids each time, results in the following estimates:

.1589, .1626, .1521, 1568, .1553

These estimates support the closed form solution detailed above.

1.2 A DDENDUM : P(S UCCESS ) IN LARGE DIMENSIONS


For an i x i -dimensional grid, call S i the probability that the grid will guarantee privacy. We
will prove that this sequence converges to 0. That is to say

lim S d = 0 (1.8)
d →+∞

First, we note that n i x i -dimensional grid contains at least m = bi /2c pairs of disjoint, adja-
cent squares. To see this consider the pairs {(n 1,1 , n 1,2 ), (n 2,3 , n 2,4 ), . . . , (n m,2m−1 , n m,2m )}. Call

2
P (t i ) the probability that a given pair of squares contains arrows that do not point to each
other. Clearly, these probabilities are i.i.d for all pairs of squares. Further

15
P (t d ) = P (t ) = ∀d < m
16
We further note that
m
\
P( t d ) ≥ P (i x i grid guarantees privacy)
d =1

As the events are i.i.d:


m
\ m
\
lim P ( t d ) = lim P ( td )
i →+∞ d =1 d →+∞ d =1
µ ¶d
15
= lim =0
d →+∞ 16
=⇒ lim P (i x i grid guarantees privacy) = 0
i →+∞

You might also like