Math Work 12345
Math Work 12345
Math Work 12345
Example 1.1. Let {0, 1} 3 denote the set of all {0, 1}-sequences of length 3:
{0, 1}3 := { ω = (ω1 , ω2 , ω3 ) | ωi ∈ {0, 1}, 1 ≤ i ≤ 3 }
2
= { (0, 0, 0), (0, 0, 1), (0, 1, 0), (0, 1, 1), (1, 0, 0),
(1, 0, 1), (1, 1, 0), (1, 1, 1) }.
Let P({0, 1}3) be the power set †1 of {0, 1} 3, i.e., the set of all subsets of
{0, 1}3. A ∈ P({0, 1}3) is equivalent to A ⊂ {0, 1}3. Let #A denote the
number of elements of A. Now, define a function P3 : P({0, 1}3) → [0, 1] :=
{ x | 0 ≤ x ≤ 1 } by
#A #A
P3 (A) := = , A ∈ P({0, 1}3)
#{0, 1}3 23
(See Definition A.2), and functions ξi : {0, 1}3 → {0, 1}, i = 1, 2, 3, by
ξi (ω) := ωi , ω = (ω1 , ω2 , ω3 ) ∈ {0, 1} 3 . (1.2)
2 22 23 2 22
23 23
5 6 1
=P , = .
In the number8line
8 with binary
8 scale, it is expressed as a segment:
4
. 。
P ←→ P, {Xi}3i=1 ←→ {di}3i=1,
{(ω, pω) | ω ∈ Ω}
zz
zz X marginal distribution
1 a11 ······ a1s1
zz
X2 zz of X2
Σs 1
a21 p11 ······ ps11 pi1
i=1
..
. . . . .
Σs1
a2s2 p1s2 ······ ps1 s2 p is
i=1 2
Repeat ‘one hundred coin tosses’ 106 times, and permit S be the variety of
the arise- rences of “the coin comes up Heads 6 times in succession”
among the 106 trials. Then, by using the regulation of huge numbers,
S/106 can be an amazing approxi- mate fee of p with excessive
probability. To try this, the full number of coin tosses we need is one
hundred × 106 = 108. Of path, we do no longer toss a actual coin
9 Ma
the
ma
tics
of
coi
n
tos
Namely, if Alice chooses an ω from {0, 1}108 , and computes the valuesinof
S(ω)/106, she can get an g
approximate value of p with error less than 1/200
with probability at least 0.99.
For any ω′, the pseudorandom number g(ω′) is not a random number.
The contents of this section slightly exceeds the level of this book.
Borel’s everyday variety theorem
Rational numbers are sufficient for sensible computation, but to permit calcu- lus
be to be had, real numbers are essential. Just like this, the primary purpose why
we remember infinite coin tosses is that they’re beneficial while we examine the
limit behavior of n coin tosses as n → ∞. Indeed, the fact that the prob- potential
area for n coin tosses varies as n varies is awkward, terrible-searching, and poor
for superior take a look at of opportunity concept.
As an instance, Borel’s normal range theorem
P lim 1 Σ d = 1 =1 (1.16)
I
N→∞ n i=1 2
Asserts just one of the analytic residences of the series of functions
{di}∞i=1, however it is interpreted within the context of opportunity principle as
“whilst we toss a coin infinitely in many instances, the asymptotic restriction of
the relative frequency of Heads is half with probability 1.” It’s miles
acknowledged that Borel’s regular range theorem implies Bernoulli’s theorem
(1.thirteen). note that it isn’t smooth to comprehend the exact that means of
(1.16). Intuitively, it manner that the ‘duration’ of the set
A := ( x ∈ [0, 1) . lim 1 d (x) = 1 ⊂ [0, 1)
I
. n→∞ n i=1 2
Is equal to 1, however considering the fact that A isn’t a easy set like a semi-open
c program languageperiod, how
To define its ‘length’ and how to compute it come into query. To clear up them,
we want degree idea.
19 Ma
the
ma
tics
Construction of Brownian motion of
coi
the second one reason why we consider limitless coin tosses is that wen
tos
will assemble a random variable with arbitrary distribution from sin
g
countless coin tosses. what's more, besides very special cases†15, any
probabilistic object can be created from them. for instance, we right here
construct from them a Brownian movement—the maximum essential
stochastic technique both in theory and in practice.Define a function F :
R → (0, 1) := { x | 0 < x < 1 } ⊂ R as follows.†16
t 1 u2
F (t) := √ exp − du, t ∈ R.
−∞
2
For example, construction of uncountable (cf. p.25) independent random variables.
15
t t
16 is an abbreviation of lim , which is also called an improper integral. See
−∞ R→∞ −R
Remark 3.7.
1.5 20
.
Infi
s. Then,
nit putting
e
coi
n X(x) :=
tos −∞ (x = 0),
ses
it holds that F −1(x) (0 < x < 1),
P(X < t) := P({x ∈ [0, 1) | X(x) < t})
= P({x ∈ [0, 1) | x < F (t)})
= P( [0, F (t)) ) = F (t), t ∈ R.
21 Ma
the
ma
Here is an amazing idea: if we put tics
of
X1 := X 2−1d1 + 2−2d 3 + 2−3d6 + 2−4d10 + 2−5d15 + ·· , coi
n
X2 := X ·2−1d 2 + 2−2d 5 + 2−3d 9 + 2−4d 14 + ·· , tos
X3 := X ·2−1d 4 + 2−2d8 + 2−3d13 + ·· , sin
g
X4 := X 2·−1d7 + 2−2d12 + · · · ,
−1
X := X 2
d + · · · ,5
.
11
then each Xn obeys the standard normal distribution. We emphasize that
each dk appears only in one X n, which means the value of each Xn does
not make influence on any other Xn' (n′ /= n). Namely, {Xn}n∞=1 are
‘independent’.
Now, we are at the position to define a Brownian motion {Bt}0≤t≤π
(Fig. 1.6):
r ∞
t 2 Σ sin nt
Bt := √ X1 + π Xn+1, 0 ≤ t ≤ π. (1.17)
π n n=1
n=1
1.0
0.5
–0.5