Stochastic process consisting both coding and proof, and some computation for constructing a transition matrix

Oscar Albert
Oct15_ClassNotes.pdf

SIMULATION ( CONT 'D ) .

Recall :

Enemies simulate X. - Unit 142, . . . ,n ? Note PfX=i } - ft .

i=1 . . - in .

O ow.

If - in's U, I then X - j ⇒ If j = Lund -11 , then X. =j

⇒ X : - Idt i . is discrete Uniformly distributed on ↳ 4k , 2, . . - in} .

Enample : Xr Geometric Cp) . Simulate from knifco, '! -

Pdx .si?=fPCtBt "

sista . . . . . . } O O - W . .

show that I := fhaana-jy.se, f. n-Geometmp.GS.ji. j- 2

-

N mm

tf . osuep.tl -PI

- '

⇒ x =p

If PLU Eptpftps . ⇐ X -

- 2. ftp.h-ps.

I pk ,

Z K P -- gtf p-EU-DLUEp.EU -D. ⇐ 11=3 .

: -

:

u El.PL?o.ltpk,pE.oli-p*f ⇐ x=j

lllfxcnl. -

a

s -

Iain .

"

p.IE?!Ci-pMs.UEp.Ei.oCi-p3k-z=sx--j ⇒ til Icu . #t÷÷¥: ) ⇒ I - Li-P) 't

't L U

.

E l - f -DJ

⇐ - Li - p) 't 't L U - l E - f - p)

't

⇐ ( i -Pt 1. I - u e. LI-Dtt.

⇐ of '

bn h-P) .

I .

ln.CI -al L ly :c ? In - Li -p?

⇐ j E . that. t j - i . In Cl - p).

i. x.si#x:--k.Y.$

Both U & - I - U . Nllnifco, i ).

SIMULATING A MARKOV CHAIN WITH FINITELY

MANY STATES :

suppose { Xn} n >o.

forms a MC . , OI ni an - A

initial distribution '

, P. as a transition

'

mat M .

→ Nz D: - - - - N N ,

D= #PCninlei, JEN .:-# Rn . 2-plni.mg:p Flaring

-t

- I

4:{"÷.fi E- ng .

Let L be a large number . Assume Xo -- n , . We know that 3¥

, pea , nj ) =L .

if Us . Elif ,

planar .

) , it

, pin , ,nr!)

T

' Einar. .⇐÷:÷÷, ÷ ,

⇒ Xz = nm .

U L .

- Ps : I - Generate L independent =

uniforms. fan L "

barge '"

t

[ 14 , . . . .sk} . Let IEREL. Z - Let ' R -- I

xi-nj.5tpEgpcm.ar.sc@EIE.pcniinr.s og - op . -

4€. E 3. - let - R - Rtl .

" I:÷ :: "

: Generating a path for the process Xm . based on P

-

F-

*In >o

{ Xo . , Xi , Xu - - - - - }. = { Xo -- Nj *

l X

' -

Mqm. ' Xz? Nj , - -

- .

CHAPTER. 2- COUNTABLE MARKOV CHAIN -

#

t.es : at Random walk

.

on Zt- ltimple ). -

.

I-p. p.

c-→.

I l S "

Pli , it) =p.

Pli ,

i- is = ,- p. } for i EET (2) RW - on It . -

.

→ I

n y .

Define plays. = Pdx, - yl Xo -

- n} .

i . I

.

yes. Puig ). =L .

11

¥ .

Then the n- step transition is given by : Pncn , y) = PfXn=y/Xo=n}.

CHAPMAN - KOLMOGOROV EQUATIONS

let OL man co

Pmtn . Caryl. = PfXm+n . =ylXo=n}

= [ .

PlXm+n .

-

ay , Xm=Z/Xo=n} 2- E. It . - phXm=zIXo=n}

= E PLXmtn-y.fi/m=z } ZEE

.

pyxm.az/Xo=nf.--zEfzPfxn--YlXo--Z3PfXm--zlXo--a3. Pmentacy ! Z Zz # Pricey

) Pm last?

Enough : RW - with partially reflecting boundaries :

let 0C pal . i $ -- fo , I , 2 , . . . . . } P l-p. P

'

c-→

I - P' se - I n sett

p (n, att ) =p.

p la , a - l ) = I - p. , n

> O-

p 10,0 ) = I -p

-

Emanate : Queue wig Model . (Discrete Time) .

Xn = # of customers waiting aniline for some service . -

During . each time internal. ; by - a person arrives

. we

prob p. - a person gets serviced & leaves the queue M prof q .

S '

=L 0 , I , 2 , - - - - - } .

p la , n - i) = PLXN -- a - i lxm

,

= a } .

= P L tonne one left} P{ New person entered '}

= gyu - pl.

p Cn , nti ). = p.li-q).

p Cn

, m)

.

=

qp + Cl-g) Ci -p). for

scoopco , D .

= p .

SECTION 2-2 : RECURRENCE & TRANSIENCE : -

Recall : Irreducible Markov chain .

Suppose Xn ni an irreducible MC . on a countably

infinite state space S '

& transition prob . play !

Def :- - The chain Xn is nonevent if for each state

set & i

p { Xn -n. for infinitely many n } = I . -

- The chain Xh mi transient if for each

state n E S '

i

Pf Xn = n . for infinitely many n } = O .

Let Xo = se wLott .

Define - is Rn . = -2. I ( Xn - n).

A-0

If A chain is recurrent R, = b- t ' NES .

÷

If the chain is transient Ras b -

Y prob I -

E