Markov Processes-III
Presented by:
Outline
• Review of steady-state behavior
• Probability of blocked phone calls
• Calculating absorption probabilities
• Calculating expected time to absorption
Preview
• Assume a single class of recurrent states, a-periodic:
• Plus transient states, then
• Where Does not depend on the initial conditions
 jij
nn r  )()lim(
 j
 jn
ijn xx  )|()lim( 0
Preview
• Can be found as the unique solution to
the balance equations
• Together with
 m
1
mj
k
kjkj P 1,  
 
j
j
1
Example
1 2
7/5,7/2 21
 
5.0 5.0
8.0
2.0
Example
• Assume that process starts as state 1
)99()1,1( 11111001 rPxx andP 
prxx andP
1211101100
)100()21( 
The Phone Company Problem
• Calls originate as a poison process, rate
 Each call duration is exponentially distributed(parameter
 B lines available
• Discrete time intervals of( small) length



The Phone Company Problem
  ii
i
EquationsBalance
1
:


B
i
iiii
i
ii
0
00
!//1!/ 

i
Steady State Probability:
Example#1
• Consider a Markov chain with given transition
probabilities and with a single recurrent class
which is a-periodic
• Assume that for the n-step transition
probabilities are very close to steady state
probabilities
• Find
500n
 lJKJjkij PPr
XXXX ilkJP


)1000(
)|,,( 0200010011000
Steady State Probability:
Example#1
• B)
• Solution:
• By using Bay’s Rule:
?/( 10011000
 jiP XX
 jiji P
XXXXX jPjiPjiP
/
)(/),()/( 10011001100010011000


Steady State Probability:
Example#2
• An absent minded professor has two
umbrellas that she uses when coming from
home to office and back. If it rains and
umbrella is available in her location, she takes
it. If it is not raining, she always forget to take
an umbrella. Suppose it rains with probability
‘P’ each time comes, independently of other
times, what is the steady state probability that
she wet during a rain?
Steady State Probability:
Example#2
• Markov mode with following states:
• State ‘i’ where i=0,1,2
• ‘i’ umbrellas are available in current location
• Transition probability matrix:
01
10
100
pp
pp


Steady State Probability:
Example#2
• The chain has single recurrent class which is a-
periodic
• So, steady state convergence theorem applies.
0 2 1
p1
p1 p
p
1
Steady State Probability:
Example#2
• Balance Equations are as below:
1
)1(
)1(
021
102
211
20








p
pp
p
Steady State Probability
Example#2
• After Solving:
• So, the steady state probability that she gets
wet is times the probability of rain=
)3(/1
)3(/1
)3(/)1(
1
1
0
p
p
pp






0
p0
Calculating Absorption Probabilities
• What is the probability that: process eventually settles in
state 4, given that the initial state is i?
ai
3
2
5
4
1
1
1
2.0
2.0
3.0
4.0
5.0
6.0
8.0
Calculating Absorption Probabilities
SolutionuniqueiotherallFor
iFor
iFor
apa
a
a
j
j
iji
i
i



0,5
1,4
Expected time to absorption
• Find expected number of transitions until reaching the
Absorbing state, given that
the initial state is i?
3
2
4
1
1
2.0
5.0
4.0
5.0
6.0
8.0
i
Expected Time to Absorption
SolutionuniqueiotherallFor
ifor
j
jiji
i
p



1:
40
Absorption Probabilities
Example#1
• Consider the Markov Chain
Mean First Passage and Recurrence
Times
• Chain with one recurrent class;
• Fix s recurrent
• Mean first passage time from s to i
 


j
jiji
s
m
ni
siallfor
tosolutionuniquetheare
isthatsuchnE
tpt
t
ttt
XXt
1
0
,.
]|}0[min{
21
0

Mean First Passage and Recurrence
Times
• Mean recurrence time of s:




p
ssthatsuchnE
aj
j js
ns
tt
XXt
1
]|}1[min{ 0

Markov process

  • 1.
  • 2.
    Outline • Review ofsteady-state behavior • Probability of blocked phone calls • Calculating absorption probabilities • Calculating expected time to absorption
  • 3.
    Preview • Assume asingle class of recurrent states, a-periodic: • Plus transient states, then • Where Does not depend on the initial conditions  jij nn r  )()lim(  j  jn ijn xx  )|()lim( 0
  • 4.
    Preview • Can befound as the unique solution to the balance equations • Together with  m 1 mj k kjkj P 1,     j j 1
  • 5.
    Example 1 2 7/5,7/2 21  5.0 5.0 8.0 2.0
  • 6.
    Example • Assume thatprocess starts as state 1 )99()1,1( 11111001 rPxx andP  prxx andP 1211101100 )100()21( 
  • 7.
    The Phone CompanyProblem • Calls originate as a poison process, rate  Each call duration is exponentially distributed(parameter  B lines available • Discrete time intervals of( small) length   
  • 8.
    The Phone CompanyProblem   ii i EquationsBalance 1 :   B i iiii i ii 0 00 !//1!/   i
  • 9.
    Steady State Probability: Example#1 •Consider a Markov chain with given transition probabilities and with a single recurrent class which is a-periodic • Assume that for the n-step transition probabilities are very close to steady state probabilities • Find 500n  lJKJjkij PPr XXXX ilkJP   )1000( )|,,( 0200010011000
  • 10.
    Steady State Probability: Example#1 •B) • Solution: • By using Bay’s Rule: ?/( 10011000  jiP XX  jiji P XXXXX jPjiPjiP / )(/),()/( 10011001100010011000  
  • 11.
    Steady State Probability: Example#2 •An absent minded professor has two umbrellas that she uses when coming from home to office and back. If it rains and umbrella is available in her location, she takes it. If it is not raining, she always forget to take an umbrella. Suppose it rains with probability ‘P’ each time comes, independently of other times, what is the steady state probability that she wet during a rain?
  • 12.
    Steady State Probability: Example#2 •Markov mode with following states: • State ‘i’ where i=0,1,2 • ‘i’ umbrellas are available in current location • Transition probability matrix: 01 10 100 pp pp  
  • 13.
    Steady State Probability: Example#2 •The chain has single recurrent class which is a- periodic • So, steady state convergence theorem applies. 0 2 1 p1 p1 p p 1
  • 14.
    Steady State Probability: Example#2 •Balance Equations are as below: 1 )1( )1( 021 102 211 20         p pp p
  • 15.
    Steady State Probability Example#2 •After Solving: • So, the steady state probability that she gets wet is times the probability of rain= )3(/1 )3(/1 )3(/)1( 1 1 0 p p pp       0 p0
  • 16.
    Calculating Absorption Probabilities •What is the probability that: process eventually settles in state 4, given that the initial state is i? ai 3 2 5 4 1 1 1 2.0 2.0 3.0 4.0 5.0 6.0 8.0
  • 17.
  • 18.
    Expected time toabsorption • Find expected number of transitions until reaching the Absorbing state, given that the initial state is i? 3 2 4 1 1 2.0 5.0 4.0 5.0 6.0 8.0 i
  • 19.
    Expected Time toAbsorption SolutionuniqueiotherallFor ifor j jiji i p    1: 40
  • 20.
  • 21.
    Mean First Passageand Recurrence Times • Chain with one recurrent class; • Fix s recurrent • Mean first passage time from s to i     j jiji s m ni siallfor tosolutionuniquetheare isthatsuchnE tpt t ttt XXt 1 0 ,. ]|}0[min{ 21 0 
  • 22.
    Mean First Passageand Recurrence Times • Mean recurrence time of s:     p ssthatsuchnE aj j js ns tt XXt 1 ]|}1[min{ 0