Al-Imam Mohammad Ibn Saud University CS433 Modeling and Simulation Lecture 06 – Part 03 Discrete Markov Chains 12 Apr 2009 Dr.
Download ReportTranscript Al-Imam Mohammad Ibn Saud University CS433 Modeling and Simulation Lecture 06 – Part 03 Discrete Markov Chains 12 Apr 2009 Dr.
Al-Imam Mohammad Ibn Saud University
CS433 Modeling and Simulation Lecture 06 – Part 03
Discrete Markov Chains
12 Apr 2009
Dr. Anis Koubâa
Classification of States: 1
2 A
path
is a sequence of states, where each transition has a positive probability of occurring.
State
j
is
reachable (or accessible)
) هيلإ لوصولا نكمي ( from state
i
(
i
j
) if there is a path from
i
to
j
–equivalently
P ij (n)
> 0 for some
n
≥0, i.e. the probability to go from
i
to
j
in
n
steps is greater than zero.
States i and j
communicate
( i
j )
) لصتي ( if
i
is reachable from
j
and
j
is reachable from
i .
(Note: a state
i always communicates with itself)
A set of states C is a
communicating class
if every pair of states in C communicates with each other, and no state in C communicates with any state not in C.
Classification of States: 1
3 A state
i
is said to be an
absorbing state
if
p ii
= 1 .
A subset
S
of the state space
X
is a
closed set
if no state outside of
S
is reachable from any state in
S
(like an absorbing state, but with multiple states), this means
p ij =
0 for every
i
S
and
j
S
A closed set
S
of states is
irreducible
) ضيفختلل لباق ريغ
(
if any state
j
S
is reachable from every state
i
S
.
A Markov chain is said to be
irreducible
if the state space
X
is irreducible.
Example
4
Irreducible Markov Chain
p
01
p
00
0
p
10
Reducible Markov Chain
p
01
0
p
00
p
10
1
Absorbing State
1
p
12
p
21
p
12
2
4
p
14
p
22 Closed irreducible set
2
p
32
p
23
p
22
3
p
33
Classification of States: 2
5 State
i
is a
transient state
) ةرباع ةلاح ( if there exists a state
j
such that
j
is reachable from
i
but
i
is not reachable from
j .
A state that is not transient is
recurrent
) ةرركتم ةلاح ( . There are two types of recurrent states:
1. Positive recurrent
: if the expected time to return to the state is finite .
2. Null recurrent
(less common): if the expected time to return to the state is infinite (this requires an infinite number of states) .
A state
i
is
periodic
with
period
k
>1, if
k
is the smallest number such that all paths leading from state
i
back to state
i
have a multiple of
k
transitions .
A state is
aperiodic
if it has period
k =1.
A state is
ergodic
if it is
positive recurrent
and
aperiodic
.
6
Classification of States: 2
Example from Book
Introduction to Probability: Lecture Notes
D. Bertsekas and J. Tistsiklis – Fall 200
Transient and Recurrent States
7 We define the
hitting time
T ij
from state
j
to stat
i
as the random variable that represents the time to go , and is expressed as:
T ij
min
k
0 :
X k
0
i
k
is the number of transition in a path from
i
to
j
.
T ij
is the minimum number of transitions in a path from
i
to
j
.
We define the
recurrence time
state
i
.
T ii
min
k
0 :
X k
0
T ii i
as the
first time
that the Markov Chain returns to The probability that the
first recurrence to state
i
f ii
Pr
T ii
n
n
n
1
T i
Pr
T
n X i
Time for first visit to
i
0 given
X 0 i
= i.
occurs at the
n th step
i
,...,
X
1 0 is
i
The probability of
recurrence to state
i
is
f i
f ii
Pr
T ii n
1
f ii
Transient and Recurrent States
8 The
mean recurrence time
is
M i
ii
i
|
X
0
i
n
0 A state is
recurrent
f i
if
f i
=1
Pr
T ii
Pr
T i
If
M i
If
M i <
=
then it is said
Positive Recurrent
then it is said
Null Recurrent
ii
|
X
0
i
1 A state is
transient
f i
if
f i
<1
Pr
T ii
Pr
T i
|
X
0
i
1 If , then
i
1
f i
Pr
T ii
is the probability of never returning to state
i
.
Transient and Recurrent States
9 We define
N i
as the
number of visits to state
N i
i
0
n
n
i
given
X 0 =i
,
i
1 if
X
0 if
X n n
i
i
Theorem:
If
N i
is the number of visits to state
i
given
X 0 =i,
then
i
|
X
0
i
n
0
P ii
1 1
f i P ii
Transition Probability from state
i
to state
i
after n steps
Proof
Transient and Recurrent States
10 The probability of reaching state
j f ij
Pr
T ij
n
n
for first time
n
1 in n-steps
j
,...,
X
1 starting from
X 0
0
i
=
i
.
The probability of ever reaching
j f ij
Pr
T ij
starting from state
i
n
1
f ij
is
Three Theorems
11 If a Markov Chain has
finite state space
, then: at least one of the states is
recurrent
.
If state
i
is
recurrent
and state
j
is
reachable
from state
i
then: state
j
is also
recurrent
.
If
S
is a
finite closed irreducible
set of states, then: every state in
S
is recurrent.
Positive and Null Recurrent States
12 Let M
i
be the mean recurrence time of state i
M i
ii
k
1
k
Pr A state is said to be
positive recurrent
T ii
if
M i
<∞ . If
M i
=∞ then the state is said to be
null-recurrent
.
Three Theorems
k
If state
i
is
positive recurrent
and state then, state
j
is also
positive recurrent
.
j
If
S
is a
closed irreducible set
of states, then every state in
S
is
positive recurrent
or, every state in S is
null recurrent
, or, every state in
S
is
transient
.
is
reachable
If
S
is a
finite closed irreducible
set of states, then every state in
S
is
positive recurrent
.
from state
i
13
Example
p
00 0 Transient States
p
01
p
10 Recurrent State 1 4
p
14
p
12 2
p
22 Positive Recurrent States
p
32
p
23 3
p
33
14
Periodic and Aperiodic States
Suppose that the structure of the Markov Chain is such that state
i
is visited after a number of steps that is an integer multiple of an integer
d
>1 . Then the state is called
periodic
with period
d
.
If no such integer exists (i.e.,
d
aperiodic
.
=1 ) then the state is called
Example
1 0.5
0 Periodic State
d
= 2 0.5
1 1 2
P
0 0.5 0 0 1 1 0 0.5
0
Steady State Analysis
15 Recall that the state probability, which is the probability of finding the MC at state
i
after the
k
th step is given by:
i
Pr
X k
i
π
0 , 1 ...
An interesting question is what happens in the “
long run
”, i.e.,
i
k
lim This is referred to as
steady state state
probability or
equilibrium
or
stationary
Questions: Do these limits exists?
If they exist, do they converge to a legitimate probability distribution, i.e.,
i
1 How do we evaluate
π j
, for all j.
Steady State Analysis
16 Recall the recursive probability
π
k
1
π
If steady state exists, then π(
k
+1 )
P
π(
k
) , and therefore the
steady state probabilities
are given by the solution to the equations and
i
1
i
If an
Irreducible Markov Chain
, then the presence of
periodic states
prevents the existence of a steady state probability Example: periodic.m
P
0 0.5 0 0 1 1 0 0.5
0
π
1 0 0
Steady State Analysis
17 THEOREM: In an
irreducible aperiodic
positive recurrent
states a Markov chain consisting of
unique stationary state probability
vector
π
exists such that
π j
> 0 and where M
j
j
k
lim
j
1
M j
is the mean recurrence time of state j The steady state vector
π
is determined by solving and
i
i
1
Ergodic Markov chain
.
Discrete Birth-Death Example
18
p
0
1-p
p
P
p p
0 1-p
1
1
p
0
p
1 0
p p
0
i
1-p
p
Thus, to find the steady state vector
π
we need to solve and
i
i
1
Discrete Birth-Death Example
19 In other words
j
0 0
p j
1 1 1
p p
j
1 Solving these equations we get 1 1
p p
0 2 In general
j
1
p p
j
0 1, 2,...
1
p p
2 0 Summing all terms we get 0
i
0 1
p p
i
0 1
i
0 1
p p
i
Discrete Birth-Death Example
20 Therefore, for all states j we get If p<1/2, then
i
0 1
p p
i j
1
p p
j
i
j
0 1
p p
i
0, for all
j
All states are
transient
i
0 If p>1/2, then 1
p p
2
p p
1 0
j
2
p p
1
p p
j
, for all
j
All states are
positive recurrent
Discrete Birth-Death Example
21 If p=1/2, then
i
0 1
p p
i
j
0, for all
j
All states are
null recurrent
22
Reducible Markov Chains
Transient Set T Irreducible Set S 1 Irreducible Set S 2 In steady state, we know that the Markov chain will eventually end in an irreducible set and the previous analysis still holds, or an absorbing state. The only question that arises, in case there are two or more irreducible sets, is the probability it will end in each set
Reducible Markov Chains
23 Transient Set T
r i s
1
s n
Irreducible Set S Suppose we start from state i. Then, there are two ways to go to S.
In one step or Go to r T after k steps, and then to S.
Define
i
Pr
X k
0 1, 2,...
Reducible Markov Chains
24 First consider the one-step transition Pr
X
1 0 Next consider the general case for k=2,3,… Pr
X k
k
1
r k
1
T
...,
X
1 Pr
X
k
X
1 Pr
r X k
i
p ir k
1
k
1
p ij
r k
1
T
i
...,|
X
1 0
r k
1
T
...,|
r X
1
p ir
p ij
0
i
ir
0
i