Scribd
Upload
98 views

Module 3 Markov Chains

Uploaded by

Deepika
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
98 views

Module 3 Markov Chains

Uploaded by

Deepika
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 25
acy 420 ENGINEERING MATHEMATICS -IV 5.2] Stochastic Process 1 We have already stated that in a random experiment, if a real variable X is associated with every out come then it is called a random vai riable or stochastic variable, This is equivalent to, having a function on the sample space $ and this function is called a . random function or a stochastic function. In this article we discuss a stochastic process called the Markov process which is such that the generation of the probability distributions depend only on the present state. Before we take up the actual discussion of this Markov process we present some basic. definitions and concepts relating to stochastic process. . © Classification of Stochastic Processes Let Sbe the sample space of a random experiment and R be the set of all real mimbers. A random variable X is a function f from $ to Rie, X = f(s),s € S. We define an index set T C Rindexed by the parameter t such as time. Let us suppose that the value of a random variable defined on $ depends ons € Sandt T. In this context a Stochastic process is a set of random variables { X(t), t € T} defined on $ with a parameter f. Here Xj = X (0) is called as the initial state of the system. ‘The values assumed by the random variable X (+) are called states and the set of all. possible values forms the state space of the process. If the state space of a stochastic process is discrete then itis called a discrete state process also called a chain. On the other hand if the state space is continuous then the stochastic process is called a continuous state process. Similarly if the index set T is discrete then we have a discrete parameter process. - Otherwise (ie., when T is a continuous set) we have a continuous parameter process, A discrete parameter process is also called a stochastic sequence denoted by” { x} neT. . ‘The classification of the four different type of stochastic processes are presented in the form of a table. — Discrete Index Set- F Continuous Index Set -T Discrete Discrete parameter Continuous parameter State Space stochastic process (chain) stochastic process (chain) Continuous Discrete parameter Continuous parameter State Space continuous state continuous state stochastic process stochastic processSTOCHASTIC PROCESS 5.21 Probability Vector : tc “(,, Up, ..:0,) where the quantities 7), ¥,, ... vector, : : A vector 0 = (01, 0) 0 Examples‘: a =.(1,0):; probability vectors. Definitions 421 By a vector we simply mean n tuple of numbers v, are Called components of the v,) is called. a probability vector if each one of its components are non negative and their sum is equal to unity. (12, w), w= (4, 4,12) are all Note: Ifv is not a probability vector but each.one of the v, (i = 1 to m) are non negative " then Xv isa probability vector where: & = 1/ Ev, i=t For example if v = (1, 2, 3) then 2. = 1/6 and (1/6, 2/6, 3/6) is a probability vector. Stochastic Matrix : A square matrix’ P'= (p,j) having every row in the form of a Probability vector is called a stochastic matrix. Examples *: (i), Identity matrix (J of any order, 100 10 tay |; 01.0 2 3 _ @Alo1 we[sis @) [01 > By dataa, +a, = 1,b, +b) = 1,242) a, 4 vA=[4, %| b, =[aate hy, % +0, | We have to prove that (0, a, +0, b; )+( 0; 4) +0, by ) = LHS= 0, (4 +4) +0, (bj +b) = 0) «140,» 12, +0, 1 Thus v Ais also a probability vector. 48. Prove with reference to two second order stochastic matrices that their: product is also a” stochastic matrix. >> Let A have, ay, 4, by 4, HP | and B =|," ’;?? | be two stochastic matrices, Hence we "21 2 bay bap i, My tO. = 1p by thy iho . @ gy + lgg = 1; bay ty = 1 : ™STOCHASTIC PROCESS MH %2 || on Oe AB = api M29 || Bay bog wart aba yyy tay bp oy bay Fan boys My Byy + yg byy We have to show that, My yy Fy bai +, bay + yy bay and yy by, +p boy + dy bay + byy = 1 LHS of (ii) can be written as, My (Dy +b i2) + yy (boy bay) Fi a+ tay -1= ay +a = 1, byusing (i. LHS of (iii) can be written as, ag (yy +Byp) + Mog (Bay +b pp) = yy» Lp +1 = 1 Thus AB is a stochastic matrix. 423 Gi) Remark : In particular we can say that A" (n = 1,2,3,... )areall stochastic matrices. 49. If A is a square matrix of order n whose rows are each the same vector @ = (ay, a, ++°0,) andif v= (01, 0), ---0,) isaprobability vector, prove that vA=a 4% {My M2 My >> By datawehave, A =424 ENGINEERING MATHEMATICS - IV.» = [evetenet tans mts Rete tt tHe] =[n(ty £40, ) 5 dy (OB 4D, FFD, )/ dy (ay Hey to 8, )] =[a. ty, va, oa since 0, +0) 4+++4+0, = 1 Thus vA = a as required. 50. Find the’ unique fixed probability vector of the regular stochastic matrix nal 344 “| V2 1/2 >> Wehave to find v = (x,y) where x+y = 1 such that 0A =v | o[e alte tall] | ie, [Seady trata le L [Seedy daetylepe, 9] = 3y4t attoy 11 . | getgyry ++ ii) We can solve either of the two equations by using y = 1-. 3) Oe Using y= 1-xin (i) wehave, 2,4G%., or kt 2-2e ade ET Hence y = 1-x = 1/3 and v = (xy) = (2, 18) Thus (2/3, 1/3) is the unique fixed probability vector. BL. Find the unique fixed probability vector for the regular stochastic matrix 0 o 1 | A=|16 12 173 ia 0 23 143 i | >> Wehave to find v = (x,y,z) wherex+y+z = 1suchthatvA =v : | ' o 1:0 | vy 2}| 6 12 18/=[x, y j - ada [eaie, y= 6x, 6x+3y 442 = by ,ytz = 3z , ; . £ ie,” 'y = 6x, 6x—3y+4z = 0,y-22 = 0 iid Usingy = 6xandz = 1-x—y = 1-x-6x = 1-7xin 6x—3y+4z = 0 wehave, 6x—18x+4-28x = 0 - x= 1/10 “Hence y = 6/10, 2 = 3/10 © Thus the required unique fixed probability vector v is given by é v= (1/0, 6/10, 3/10) Z 52.: With reference to the stochastic matrix A in Example-24, verify the property that the ¢€ sequence A*, A?, A approaches the matrix whose rows are each the 2 ficxed probability vector. : é >> We have A= eS vas we have obtained in Problem-50 the fixed : © piobability vector v= (2/3,° 1/3). « | Let B be the matrix whose each row is v.: & ne 243 3 . wo a-[38 4) d tan oh 27] € Consider A = 5 [? al Nowa2-= {9 2][3 1]_ 1/11 5] _[o06875 o3125 a w= 612 2] 12 2)" Te] 10 6 || 0.625 0375 | 22[31]f 1 5]_ 1 [43 21]_Posrs7s ‘osesi2s © 64/22 10 6| 64 42 22 |” | 0.65625 0.34375 ek 3.1][43 21]_ 1 [171 85] [oer oss 256 | 2 2] 42 22} 256 | 170 86 |* | 066 034 Each row of A‘ is approaching v = (2/3, V3) = (0.67, 0.33)426 ENGINEERING °MATHEMATICS - 1 53. Find the unique fixed probability vector of the regular stochastic matrix - 0 V2 v4 14 . v2 0 V4 V4 v2 V2 0 0 vw2v20 0 >> Wehavetofind v = (a,b, c,d) where at+b+c+d = 1 such that vP =v 0 V2 4 14] : v2 0 14 V4 . v2 12 0 0 [ala ber dl 212.0 0 > [a,b, c,d] 1 =F (btord) =a or bectd = 1 p (atetd) = b or atetd = 2 7 1 4 (ath) =e or ath = 4 Fash) =d ot atb=4d Byusing b+c-+d = 1-a and atc+d = 1-b (j) and (ii) respectively becomes 1-a = 2n and 1-b = 2b a= and b=13 Hence we have from (iii) and (iv), Ac = 2/3 and 4d = 2/3 c= 1% and d= 1/6 ‘Thus v = (1/3, 1/3, 1/6, 1/6) isthe required unique xed probabity ve vector. 1-b b 54, Show that (a, b) isa fixed point of the stochastic matrix p-[ ae the associated fixed probability vector ? Hence write down the fixed probbility vector of enc of the following matrices, V3 23 P= 2 12 7/10 3/10, % 1 0 v3 3 _ (8210: 2/10} » sjSTOCHASTIC PROCESS 427 >>, Let x =.(a, b) and consider the matrix product to fink be rP=ta.bi| A ia = [e(1-b)¢bn, ab+b(1-a)] = [a,b] Thus xp = x = (a, b) isa fixed point of P. Also v= (a/a+b, b/a+b) is the required fixed probability vector of P. Comparing P,, P,, P, with P we have respectively @=1,b=28 5 0223,b=12; 0=8/10,b=3/10 atb=5/8 jatb=7/6 — ;a+b= 11/10 ‘The corresponding fixed probability vectors of P,., P,, P3, be respectively denoted by v1, ¥), v5 where we have in general .-0 = (a/at+b, b/atb) ‘Thus 0 = (3/5, 2/5) 5 0, = (4/7, 3/7) | v3 = (8/11, 3/11) are the required fixed probability vectors of P,, P,, Py in the respective order. 1-a a 1-b 6 55. If P, -| j sy] Be aes show that Py, Py and P, P, are stochastic matrices. >> In P; wehave (1-a)+a=1 and b+(1-b) =1 In P, wehave b+(1-b) = 1 and a+(1-a) = Thus P, and P, -are stochastic matrices. _[i-a a@ ]fi-> » Now MyPy=|"y deal] e aa (1-a) (1-b) +a, (1-0) b+a(1—-a)]_ [4% ( = i: =] ap, | ou b(1-b)+a(1-b), P+(1-b) (1-a) ab, We shalll show that a, +b; = 1 and a,+b, = 1 Now a, +b, = (15a) (1-b) + (1-a)b+a?+a(1-a) = (1=a) (1-b+b] +2 [a+1-a} sleata=l 6 a+b =1428 ENGINEERING MATHEMATICS - I Also a+b, = b(1— b)+W40(1-b) + (1-b) (1-2) = 6 {(1-b)+b}+C1- ~b) {a+(1-a)} Sb+I-b= 1, Thus a+b, = 1 Thus P,P, is a stochastic matrix. | o 100 Z 56. ShowthatP=| 0 O 1 is a regular stochastic matrix, Also find the associated . ~ 12120 unique fixed probability vector. o 10 o 10 oo >> Consider P=| 0 0 1f// 0 0 1J=|1212 0 12120\/12 120] | 0 17212 10 21/2 0 P.P= p= } o1 wind =| 0 v2V2 121/20/| 0 17212) ..] 174144 172 0° 1 offa242 0 01212 P-P-pi-| 0 01]| 0 w212/=/141412 121720|/141412| [141214 | 0 10 0 12 2 41/4 1/72 | ppaP i O 1s) 2|=| 1741/2 174 1/2 1/20 }| 1/4 1/2 1/4 V8 3/8 1/2 We observe that in P® all the entries are positive. } Thus P isa regular stochastic matrix Next we have to find v = (a, b, c) where a+b+c =1 suchthat vP =v 010 => [a,b,c]| 0 0 1/=[2, b,c} V21220 ce ie, [ser] [a, b,c] b bse @ ate 2 wey Using ¢ = 2n and b= ¢ = 22 in atb+c = 1 weget Sa=lora=1/5 Hence b=c= = 25 Thus (1/5, 2/5, 2/5) is the required unique fixed probability vector of P. > | |- Probability p,; eye 429 ‘STOCHASTIC: PROCESS . 5.22| Markov Chains A stochastic process which is such that the generation of the probability distribution sdepend only on the present state is called a Markov proces. ; If this state space is discrete (finite or countably infinite) we say that the process is a diserete state process of chain. Then the Markov process is known as aMarkov chain. ~, Further if the state space is continuous, the process is called a continuous state process, We explicitly def fine a Markov chain as follows. Let the outcomes X, Xz, «+, of asequence of trials satisfy the following properties. (i)’ Each outcome belong to the finite set (state space) of the outcomes Mr ty yh Gi) The outcome.of aity trial depend at_most upon the outcome of the immediate preceeding trial! is associated with every pair of states (4,4) that a; occurs immediately after a; occurs. Such a, stochastic process is called a fete Matos Tie chin P;;) which arenon zero real numbers are called transition probabilities and they form a square matrix of order m called the transition probability matrix (p.m) denoted by P, Pru Paz ++ Pam , ©] This implies that after 5 throws the probability that the ball is with A is 1/8, theball with B is 3/8, the ball with C is 1/2. © Stationary distribution of regular Markov chains ‘A Markov chain is said to be regular if the associated transition probability matrix P is regular. If P isa regular stochastic matrix of the Markov chain, then the sequence of » step transition matrices P?, P?, ---P™ approaches the matrix V whose rows are each . the unique fixed probability vector v of P. Wehave pt") =p) p™ where, pl") -[A”. n.. w] Furtheras > ©, p{") =v; where i= 1,2, 3, --+m.. 433 STOCHASTIC PROCESS * ‘Whig '$8-"callled” “the stationary distribution of the, markov-chain and 2, U, +1-2,,). is called the stationary (fixed) probability vector of the Markov chain. Referring to the Illustrative Example - 2, the t.p.m of the Markov chain is : 0.10 P=/.0 O11 “| 2-172 0 and by Worked Problem - 56 the’unique fixed probability vector of P is (14,275, 245), * 2 “Hence we conclude that in the long run (11 —> se) A_will have thrown the ball 20% of the time; while B and C will have thrown the ball 40% of the time. Note: A Markov chain is said tobe irreducible ifevery state can be reached from every other state in a finite number of steps. That is to say that p{") > 0 for some n > 1, This is equivalent: to saying that a Markov chain is irreducible if the associated transition probability matrix is regular. { * Absorbing state of a Markov chain In a Markov chain the process reaches to a certain state after which it continues to remain in the same state. Such a state is called anabsorbing state of the Markov chain. ( In an absorbing state the transition probabilities p,; are such that Pj = for i= j and p,, = 0 otherwise. Thus a state 4;.of the Markov chain is absorbing if the i" row of the tip.m has 1 on the principal diagonal and zeroes elsewhere. Examples : The state a, is absorbing, 0 2 pe? m . a The states a, and a, are absorbing, % aj % 0 O1~ 1434 eae = ENGINEERING’ MATHEMATICS -1V WORKED EXAMPLES 57. The transition, matrix P of a Markov chain is given ala val with the initial 3 1/4 probability distribution p) = (1/4, 3/4). Define and find the, flowing @ AD Gd PD Gi pf. Gey PO (o) the vector p\) y* approches. (vi) the matrix P” approaches. 3 > >> (i) p{?) is the probability of moving from state a, tostate a, in’3steps. This can be‘obtained from the 2 - step transition matrix P? 2 2 pr | V2 V2}[v2 12] [58 38 )_ PD pip 3/4 1/4|| 3/4 1/4)” | 9716 7/16] | (2) 4f2) Pa’ Po} «AP = 9716 Gi) _p\2 is the probability of moving from state a, toa, in two se {2) = 3/8 * (ii) __p©) is the probability distribution of the system after 2 steps. (2) = (0) p2 — 5/8 -3/8 a 37» 27 Be AF = [UA ¥4]( Se one |" | ea 64 Thatis pl? = (37/64, 27/68] = 1), PP] (iv) _p(?) is the probability that the process isin the state a, after 2steps. Hence > = 37/64. (wv) The vector pi) P" approaches the unique fixed probabilty vector of P and we shall find the same. Let v = (x,y) where x+y =1 and we musthave vP = 2-1 “ That is [x, n[e valet me . a XDA By/h = x and x/L4Y/A ey Using y = 1-x the first equation becomes x, 3(1-2) 2 4 since x = 3/5, ¥ = 2/5 =x or 243(1-x) = 40 : 23/5STOCHASTIC PROCESS 435 The vector p°) P" approaches the vector (3/5, 2/5) (i) P™ approaches the matrix V whose rows are each the fixed probability vector of P. " = | 3/5 2/5 P" approaches the matrix [ a Z| 58. The t-p- mt ofa Markov chain is given by 20 12 P=!1 0 0 412 144 and the initial probability distribution is p) = (1/2, 1/2, 0) Find pD, “PB, pl and pl? >> First let us find the two step transition matrix P* V2.0. 72][172 0 172 3/8 1/4 3/8 Peel dio o}li1 0 ol=|12 0 12 V4i1/2 V4)| 17412 4] | 11/16 18 3/16 ae 2)_ ’ Ag) = 3/8 and phf) = V2 3/8 1/4 3/8 pl?) = pl) p? [12, 12, 0] v2 0 12 oy [Mas vs sae} =[7/16, 18, 716 | ~ — % p) = (7/16, V8, 7/16) and pl? = 7/16 59. Prove that the Markov chain whose t-p-m is 0 243 13 P=|1/2 0 1/2} is irreducible. 172 172 0 Find the corresponding stationary probability vector. >> Weshallshow that P isa regularstochasticmatrix, For convenience weshall write the given matrix in the form1 j | 436 ENGINEERING MATHEMATICS -IV 1 042);//042 1 18 6 12 Consider P? = 3 303//303/= 7/9 m6 330//3.30 9 12 15 Since all the entries in P* are positive we concliide that the t.p:m-P-is regular. Hence the Markov chain having tp.m P is irreducible. Next we shall find the fixed probability vector of P. If v= (x, y, z) weshall find v such that vP xtytzel v where 1042 Thatis [x, y, z] “GE |303/=Lx yz] 330 1 =F (8y +e, 4et32, 2e43y] = [2, y, 2] Solvingthesebyusing x+y+z = 1 weobtain x= 1/3, y = 10/27, 2= 8/7 Thus 0 = (18, 10/27, 8/27) is the required stationary probability vector. 60. Ahabitual gambler is a member of two clubs A and B. He visits either of the clubs everyday ‘for playing cards. He never visits club A on two consecutive days. But, ifhe visits club B on a particular day, then the next day he is as likely to visit ciub B or club A. Find the transition matrix of this Markov chain, Also, (a) show that the matrix isa regular stochastic matrix and find the unique fixed probability vector. | 1 { | | | > By tBz = 6x 5 Ax 432 = by ; 2e4dy = & | | i (b) if the person had visited club B on Monday, find the probability that he visits club A ont Thursday. >> The transition matrix P_ of the Markov chain is formulated as follows, AB ! be 4 0 17} Mt 8 ~ pl V2 V2} | ay ay t | | The first row corresponds to the fact that he never goes to clubA on two conseciative |, | days which implies that he is sure to visit club B. The second row corresponds to the” fact that if he goes to B on a particular day he visits B or A on the following day. Probability of going to A is 1/2 and probability of going to B is.also 1/2. : . : !STOCHASTIC PROCESS sb eomertel 22 o 4 O.t |_[iw2 2 (@) Now consider P* = [a lla zai = a Al Since ail the entries of P? are positive P is a-regular stochastic matrix, ‘We shall find the unique fixed probability vector. That is to find v'= (x,y) such that oP = y where x+y =1 [+ Ifa aa)-L* 9] ~ [feed] Thus 0 = (1/3,2/3) _ 437 (b) Let us suppose Monday as day 1, then Thursday will be 3 days after Monday. Given that the person had visited club B on Monday the probability thathe visits club A after 3 days is equivalent to finding a{*) from P°. -_ 12 V2)f 0 1 ]_[waava eH Pe Pola 12 172|"| 3/8 5/8 a3) = 3/8. Thus the required probability is 3/8, 61. A student's study habits are as follows. fhe studies one night, he is 70% sure not to study the next night. On the other hand if he does not study one night, he is 60% sure not io study the next night, In the long run how often does he study ? >> The state space of the system is{ A, B} where A Studying B: Not studying, The associated transition matrix P is as follows. . AB 4f03 07 Be oles oe «In order to find the happening in the long run we have to find the * probability vector v of P. That is to find x,y) suchthatvP =» where x+y =1 unique fixed ‘ ‘438 ENGINEERING MATHEMATICS «IV ie (m9) te ae L 2] ie, [03x+04y, 0: ‘+ 06y ]=[x, i = 03¢+0.4y O7e+06 y = y Using y = 1—x in the first of the equations we have 03x+04(1-x) =x or Lix= 04 . x= 4/11 Since x= 4/11, y= 7/11, v= (4/11 and 7/11) = (P41 Pp) Thus we conclude that in the Jong run the student will study 4/11 of the time or 36.36% of the time. 62. A man’s smoking habits are as follows. If he smokes filter cigarettes one weck, he switches to non filter cigarettes the next week with probability 0.2. On the other hand, if he smokes non filter cigarettes one week there is a probability of 0.7 that he will smoke non filter cigarettes the next week as well. In the long run how often does hte smoke filter cigarettes ? >> Thestate spacé of the system is{ A, B'} where A: Smoking filter cigarettes, B: Smoking non filter cigarettes ‘The associated transition matrix is as follows. AB p— *[08 02]_ [810 210]_ 1/8 2 ~ 3/03 07 |~ | 3/10 7/710|~ 1013 7]. We have to find the unique fixed probability vector, v= (ayy) such that » where x+y = 1 ie [wo] aa[8 a] 2] ie, [sx+9y, 2e47y] = [ 10%, 10y] > Bet By = 10x, 2x+7y = 10y Using y = 1x in the first equation, we get, 8r+3(1-x) = 10x o | £535 y= = (3/5, 2/5) = (Pas Bg) In the long run, he will smoke filter cigarettes 3/5 or 60% .of the time, Se EE, Hence v = (x.ySTOCHASTIC PROCESS 439 63. Each year a man trades his car for a new cnr in 3 brands of the popular company Maruti Udyog limited. If he hes a ‘Standard he trades it for ‘Zen’, If he has a "Zen he lrades it fora Esteem’. Ifhe has a ‘Estee’ ke is just as likely to trade it for a neta ‘Esteen’ or for a Zen’ or a ‘Standard’ one, In 1996 he bought lis first car which twas Esteem, (i) Find the probability that li ins (a) 1998 Esteem — (b) 1998 Standard (c) 1999 Zen (@ 1999 Esteem. Gi) In the Long rie, how often till he have a Esteem? >> The state space of the system is{ A, B, C} where A:Standard B:Zen C: Esteem. ‘The associated transition matrix is as follows. AB Cc Aro 1 0 1 42 Sa P=Bl 0 0 1 |=) 4% ay my cL YS 13 V3] Jas, ay O55 (i) With 1996 as the first year, 1998 is to be régatded as 2 years after and 1999 as 3 years after, We need to compute P* and P® o 1 ojfe 1 oj}'f'o 0 1 P=/o 0 1{/0 0 1 |=|18 13 3 13 13 13||13 14 13] |19 49 49 0-0 1)f0 1.0 1B 13 V3 P=/13-1313|}0 0 1/=|19 49 4/9 yo 4/9°4/9 || 139 173 V3) | 4/27 7727 16/27 (a) 1998 Esteem, = 0(2) = 4/9 a with reference to P* (b) 1998 Standard = af?) = 1/9 (e) 1999 Zen = al) = 7/27 a with reference to P® (4) 1999 Bsteem , = a3) = 16/27 * Gi) We have to find the unique fixed probability vector » = (x,y,2) such that vP.= v where x+y+z=1 070 a is : i] is A is ale é ‘l“a ENGINEERING MATHEMATICS,- IV : ie, [= y=] ; oor =[s ¥ 2] ie, [= 3x42, sy+2z] =[3s, 3y, 3: = z= 3x, 3rt¢z = 3y, By +z = 32 Consider 3x+2 = 3y ; Using z = 3x and y = 1—x~-z we get 6x = 3(1-x-z) or 6x = 3-31-32 or 8x =3 0. x= 16 Hence we obtain y = 1/3, z = 1/2 o=[xy z]=[1%, 14, v2] = [A, ?, ] : In the long run, probability of he having Esteem is f°) = 1/2 Thus in the long run in 50% of the time he will have Esteem. 64. Three boys A, B, C are throwing ball to each other. A always throws the ball to B and B always throws the ball fo C. C is just as likely to throw the ball to B as to A If C was the frst persom to throw the bal find the probabilities that after three throws @ A has the ball Gi) B has the ball (iii) C has the ball >> State space ={A, B, C} and the associated t- p-m isas follows. ABC 470 10 P=Bl 0 01 c [12 12 0] Intially if C has the ball, the associated initial probability vector is given by py) =(0,0,1) ' 7 Since the probabilities are desired after three throws we have to find pl) = p() Pe 1212 0 Refering to the Problem -56, P? =| 0 1/2 1/2 14 1/4 172 3). (ps2 2 12) _y sy cay ay Pp) = pI P(E Ea] [a Pa Thus after three throws the probability that the ball is with Ais 1/4, with B is 1/4 and with C is 1/2,STOCHASTIC PROCESS * 441 65. Two biys'B, ,B, and hoo girls G,, Gare trang bl from one tothe otter. Each ~ bay throws the ball othe other boy with probability 1/2. and toench gil wit probability 1/4. On the otherhand each girl throws the ball to each boy with probability 1/2. and never to the other girl. In the long run how often does each receive the ball. >> State space: = {3,, By, Gy, G} and the associated t-p+im P isas follows. B, BG, G B. 1 *2 “r Yo To 12 v4 Blin o w414 2 G)1212 0. 0 G [2120 0 P= Weneed to find the fixed probability vector 0 = (a,b, ¢, d) such that uP = 0 Refering to Problem -53, Wehave 9 = (1/3, 1/3, 1/6, V6) "Thus we can say that inn the long run each boy receives the ball 1/3 of the time and each girl 1/6 of the time. 66. A gimbler’s luck follows a patiern. fhe wins a game, the probability of winning the next $amé is 0.6, However if he lses a game, the probabil lity of losing the next game is 0.7. ‘There isan even chance of gambler winning the frst game. Iso (a) What isthe probability of he wintiing the second game? (): What is the probability of he ‘winning the third game? (©) Inthe long run, how aften he will win? >> State space { Win (W), Lose (L)} Wo pu” [06 04]_1f64 “iL £03.07}" 10/37 Probability of winning the first gameis “1/2, iniitial probability vector p\°) = (1/2, 1/2) and the associated t- p-m is.as follows. (19'S (0p 2 Af 64] a (a) Now pI = AO P= FULL a5] Sa =e IS, M1] Henee p!) = [9/00, 120] =[p", fh] Thus the probability of he winning the second game is 9/20.44; . ‘. 2 ENGINEERING MATHEMATICS -IV » PoBnege, 1] 33 |= ate 70 Hence p'?) = > =[ 87/200, 113/200 ] =[o (WY pt ] Thus the probability of he winning the third game is 87/200. (&) We hall find the fixed probability vector = (x,y) suchthat vP =v where x+y = 1 Thatis [x,y] “spl s7]7 a => 6r+3y = 10x, Ax+7y = 10y or —3y = 4x and by using y = 1-x weget B(l-x)=4% 40 x= 3/7 and y= 47 Hence v= [37,47 ]}=[p, pf] ‘Thus in the longrun he wins 3/7 of the fime. EXERCISES 1. Identify the probability vectors from the following. (a) (2/5, 3/5) (b) (0, -1/3,4/3) (©) (1, 0, 1%, 1/2, 1/3) 5 @ (13,0, 14%, 1/2) (ey (0.1; 02, 03, 04) 2, Find the associated probability vector to each of the following tuples. @) (1,3, 5) (b) (4, 0,1, 2) © (12, 2%, 0, 2, 5/6) : 3A = [ay ] isa stochastic matrix of order mx n and pe (oy, ¢ -v,) isa probability vector, show that vA. is ia qotatitly vector. 4, If A and B aretwo stochastic matrices of order 3 x 3, prove that AB isalso a stochastic matrix.STOCHASTIC PROCESS 443 5. 2 x 10. Show that (cftcerde, aftbf+ac, ad + bd +be) isa fixed point of the stochastic matrix l-a-b a b e l-e-d d e f l-e-f » Show that the following matrix P isa regular stochastic matrix and also find its unique fixed probability vector. : 05 0.25 0.25 P=/05 0 05 o 1 Oo " 1 0 an ee i ae . Given the t-p-m P= [is iA with initial probability distribution py) = (1/3, 2/3), find the following © PE? & PO © A? A software engineer goes to his office everyday by motorbike or by car, He never goes by bike on two consecutive days, but if he goes by car on a day then he is equally likely to goby car orbybikethenext day. Find the t- p-m ofthe Markov chain: If car is used on the first day of the week find the probability that after 4 days (a) bike is used (b) car is used. 1. Asalesman’s territory consists of 3cities A, B, C. Henever sells in the same city for 2 consecutive days. Ifhe sells in city 4, then the next day he sells in city B. However if he sells in either B ot C, then the next day he is twice as likely to sell in city A ‘as in the other city, In'the long run how often does he sell in each of the cities ? Show that the Markov chain with tp’ m given by 622 70 | 1 8 1| is ieteducible. Find the corresponding stationary probability 604 vector.te ENGINEERING MATHEMATICS -1V 1 (a), (4), (e) are probability vectors. 2 (a) (19, 3/9, 59) () (4/7, 0, 1/7, 2) (9 (18, 1%, 0, 1/2, 5/24) 6 (4/11, 4/11, 3/11) 7% (a) 78 (b) (1112, 1712) () 1/12 BoC i | Bio 1 8 ale | (@) 5/6 =~) 1116 9. v = (04, 0.45, 0.15). po long sie al ofthe tins incl A, 45% of the time in B, 15% of the time in ‘C 10. v = (4/10, 4/10, 2/10)

You might also like