Eigenvalues and eigenvectors Contents: A B & D Eigenvalues and eigenvectors ‘Matrix diagonalisation Matrix powers ‘Markov chains328 _FIGENVALUIFS AND FIGENVECTORS (Chapter 19) OPENING PROBLEM Tn springtime in a particular seaside town, it has heen observed that: © Each day, the weather is one of two states: sunny or rainy. Tomorrow's weather only depends on today’s weather, © If itis sunny today, the probabi © If it is rainy today, the probability that it will be rainy tomorrow is 0.4. that it will be sunny tomorrow is 0.8. Things to think about: called a transition matrix. a The matrix T= (a a 0.2 04 1 = 0.8. is the probability that it will be sunny tomorrow given that it is sunny today, What is the meaning of 1 ty =04 Wt =02 Wt, =0.6? & The column matrix s) = (a) is a state matrix. which indicates that it is sunny today. The state matrix for tomorrow is 8; = Tso = (ie) 1 What does s; mean’? i What do you think sy =T2s_will represent? € Can you use the matrices 'Y and s9 to tind the probability that tt wall be sunny: Tin 4 day’s time ii ina fortnight’s time? d= ii Is there a column matrix s such that Ts — s? Hi Accnming that such a matrix § exists, what might it mean for the weather in the town in the long term? ‘The weather mode! described in the Opening Problem using state and transition matrices is an example of a Markov process. In this Chapter we study eigenvalues and eigenvectors, which are properties of square matrices. We will explore how these properties provide understanding to the behaviour of Markov processes over time. We will sce further applications of eigenvalues and eigenvectors later in the course, including geometric, transformations, and in understanding the behaviour of physical systems. Suppose A is a square matrix, If x is a non zero vector and \ is a constant such that Ax — Ax, then \ is an eigenvalue of A and x is its corresponding eigenvector. FINDING EIGENVALUES If Ax=Ax then \x~Ax=0 . (I= A)x=0 {since Ix =x}Now if det(Al — A) #0, there For x to be a non-zero vector, we therefore require det(AI — A) =0. We conclude that: unique solution to the matrix equation, but itis the solution x = 0. The eigenvalues of A are the solutions to det(AI — A) =0. The equation det(AL — A) =0 is a polynomial equation to be solved for A. We therefore detine: The characteristic polynomial of ann x n_matrix A is p(A) = det(AI — A). The eigenvalues of A are the solutions to p(A) = 0. In this course, we only consider the eigenvalues of 2 x 2 matrices. The characteristic polynomials for 2x2 matrices are always quadratic. so there may be two real and distinct eigenvalues. a single repeated eigenvalue, or else eigenvalues which are complex conjugates, Eo Dt For the matrix A = ( y i) find: aoe the characteristic polynomial the eigenvalues. (A) = det(Al — A) (A) 4 3 w MA +A-12=0 4 At (A=a)(A+4) =0 =AA+1)-12 / A= Bor —4 =NRiA 122 the cigenvalues are 3 and —4, ELL Der Find the eigenvalues of: G3) If det(Ml—A) = If det(At— A) =0 A+B 1 A-2 -5| then -1 A+1 then 1 {7° 2. AFIAFDY-(-)=0 MA — 2) — ~ M4404341 -2445=0 . P+4Q+4=0 40)(5) a. (+2) the cigenvalue is —2. _2k4i 2 S121 the cigenvalues are 1+2i and 1—2i330 FIGENVALDIES AND FIGENVECTORS (Chapter 19) EXERCISE 13A.1 1. For the matrix A = G i) find: a the characteristic polynomial the eigenvalues. . a=(2 4) t a=(2, *) 1) must only have the single 2 Find the eigenvalues of: sae(4 0) ¢a-(¥ 7) 3 a Use the definition of eigenvalues to explain why I ( eigenvalue | 1 0 Perform calculations to verify this result. I 4 84 4 (2) on (Hf) eneeaneonnran(® 4} ‘a Use the definition of eigenvalues and eigenvectors to find the eigenvalues of A. Hence write down the characteristic polynomial of A. 5 Consider A = ( iq 3) 3 a Find A? and A“? b Find the eigenvalues of A, A?, and A“? 6 Use the definition of eigenvalues and eigenvectors to explain the connection between the Pigenvalies of a Aand—A b A and a? © Annd AW? FINDING EIGENVECTORS Having found the eigenvalue(s) for the matrix A, we return to the matrix equation (AI — A)x = 0 to find dhe eonespouding eigenvectors a, If we find an eigenvector x which satisfies the equation, then any scalar multiple of x will also satisfy the equation. Therefore, there are infinitely many eigenvectors corresponding to a particular eigenvalue. Groce DEST Ie deyiar—a) =o ten | fits AA +2) — (-1)(-3) = 0 +2\-3=0 (Q-DA+3) * A=Lor-3 The eigenvalues are 1 and —3.For X= 1, consider (AT— A)y = x= {4 For any given eigenvalue, the ¢ system (MI — A)x = 0 gives 1 -3\/a)_ (0 ‘yuatious which ae seal <1) % b}~ \o multiples of one another. , a—3b=0 tO, then a= 2t 5 x= (3)u e408 Any vector of the form ( a A Ja +40 is an eigenvector corresponding to the eigenvalue 1. For \=~3, consider (AI—A)x= 0. with x-(3) as ia 0) 1 1 bb} \o -a-b= Letting b=t, t £0, then a=-t : (a) e¥0 Any vector of the form ( Fy )e 1 #0 is an eigenvector corresponding to the eigenvalue —3. EXERCISE 134.2 1 Find the eigenvalues and corresponding eigenvectors of each matrix: Eee Ea) 53 3-1 @ Show that A = ( ) has a single eigenvalue 2. & Find the conesponding eigenvector a. € Verify that Ax = Ax. MO 0» Show that the cigenvalues of A are Ax and A». and find the corresponding eigenvectors. _f{a0 via aa(§ 2), oxo Show that the only eigenvalue of A is a. 3 Let a-( ) where Mi, 2 £0 and Ar # dy. Bay Shviy Seater eneera yet (:) a an eigenvector: oF ¢ Explain the result in 6 using the definition of eigenvalues and eigenvectors. -1 -1 56 Pinu de eigenvalues and eoresponding eigenvecars of a= (15) b Hence find the eigenvalues and corresponding eigenvectors of A?i (4 _(10 eta a=(t $) ma n=(! 2) Show that A and i have the same eigenvalues, Find the corresponding eigenvectors for matrix A. Are the eigenvectors of A also eigenvectors of B? Show that the eigenvectors of G : 3) are mutually perpendicular Gane Pear Consider A = é A) which has eigenvalues 1+ 2i and 1—2i. 0 Find the corresponding eigenvectors of A. a For 1=149%, eonsider QT AWw=0 with v= (% (aac nate (3 iit a} (42 )()-(0) . at (1+ 2i)b = Letting b=, £0, then a= me (Gari): +40 Any vector of the form (142i ¢ = (-1-2i)t ii . £40 is an eigenvector corresponding to the eigenvalue 1+2i. For X —2i, consider (AL—A)x=0 with x= (aaceny 0-5 (3) (0) 0-(-1) 1-2-0) \s 0 (77% 2) ()-@) 1 1-2) 6 0 * at(1—2i)b=0 Letting b=t, t40, then a=—(1—2i)t = (-1+2i)t ae (Gea) Pant ‘Any vector of the form ( a a 470 is an eigenvector comesponding to b the eigenvalue 1 — 2i.Fil FNVALUIFS AND FIGENVECTORS (Chanter 19) 333 Oo -1 8 Let A= ( 0 ) a Show that ( 1) +£0 are eigenvectors of A. State the corresponding eigenvalue. b State the other eigenvalue of A. ‘€ Hence find the remaining eigenvectors of A. (pals il _ (3 -5 eek) | ae HISTORICAL NOTE One of the first known applications of eigenvectors came from Leonhard Euler’s study of the rotational motion of rigid bodies. The Italian mathematician Joseph-Louis Lagrange recognised at the principal axes for dhe tation conespomd w the eigenvectors of the inertia matrix. In the 19th century, numerous mathematicians and physicists considered the properties of eigenvalues and eigenvectors in their studies. However, it was not until 1904 that the Prussian nrathematician David Hilbert gave theur the Gein description eigen, meaning “own”. ‘A non-zero square matrix is said to be diagonal if the elements not on its leading diagonal are zero, So,a 2x2 diagonal matrix has the form ( ; ) where a, b © Rand are not both zero. ‘A square matrix A is diagonalisable if there exists a matrix P such that D = P-!AP isa diagonal matrix. We say that P diagonalises A. In this course we only consider the diagonalisation of 2x 2 matrices. ICA isa 2x2 matrix with distinct eigenvalues 1, do P= (xi| x2) is the matrix and corresponding eigenvectors x1, xz, then P = (x; |x2) ‘with first column ; and iy second column x2. diggonatises A, and P ‘AP = (yy, ). = If A only has a single eigenvalue, then either: @ A Sal a £0, in which case A already dingonalised, or © the matrix P= (x|x) formed from its eigenvector is singular. and so A cannot be diagonalisedExample 5 CRTs ‘The matrix 3 1 eigenvectors x: = (}) and x= (7!) respectively Show that P; = (x;|xg) and P2=(xz|x1) both diagonalise a. Pe=(ealn)= (4! ) ) has eigenvalues X; =1 and 9 = 4, with corresponding el tel) mee Show that Py = (Xi|x2) and Pa eigenvectors xy = (X2|x1)_ both diagonalise A. The matrix A = G 4) has eigenvalues 4, = —3 and Ay = 6, with corresponding eigenvectors x1 -( 2) and x9 = (1) respectively. Show that Py = (x:|x2) and P2 = (xz[x1) both diagonalise A. Example 6 CORT Let a-(i Find the eigenvalues \1, 42 (Ar > 2) and corresponding eigenvectors x1, x2. Verify that P = (x1 |x2) diagonalises A.a If det(AL A) —0 then da1+ Vi For 4; =1+V6, (AI—Ajx=0 with == (()) v6-1 5 a 0 wm (1 a) (5) (0) . -a+ (1+ v6)b=0 Letting b=t, #0, then a=(1+VO)t (3 )-( 146 ; Ji teo 1+V6 1 ‘We could choose any choosing #3, 1 — (1499) i a siamostor corresponding to 4; = 1+ V6. For 4) =1-V6, (MI—A)x=0 with x= () ee -1-V6 -5 a\_ (0 gi -1 1-v6)\b)~ \o alo. v6)=-0 edie vee OM eaten a 1-v6 (3)= ("7") choosing t ie ( ria) is an eigenvector corresponding to A» 1+Vv6 1-v6 Let P= (xs]m2) = ( Fi ) Mat PxNat AxMat P THB 0 0 16 1 14V6 0 v tar= ( 0 ne a) {using technology} P diagonalises A. 3. For each matrix A: Find the eigenvalues 1, Az (A1 > Az) and the corresponding eigenvectors x1, x2 Verify that P= (x1]x2) diagonalises A. (4)336 _FIGENVALUIFS AND FIGENVECTORS (Chapter 19) 4 Consider the matrix A = ( ?) @ Pind the ci Jculwis ay, a2. walues Ay, Ay (Ay > Ag) aud their euntespuudia 2 ythat: = | POLAP= (3 A ii rues ( ; sige b& If P=(x |x), v We have already seen that we can calculate powers of a matrix A, provided that A is square. For larger powers, this is tedious to caleulate by hand However, if matrix A with eigenvalues 1, Az_ is diagonalisable by P, then prar= (4) QP Or» 0» Pe-! = PDP” {pre-multiplying by P and postmultiplying hy P~!} A= Ppp"! {pp-! a )(PDP~!)(PDP~* times = PDP~'PDP~'PDP™*....PDP~ =PD D D — DP- = PD"P DORSTarrey Emo 0 fl ‘G ) Use the diagonalisation of A to find A® Bee . 2 10 aa ) diagonolises A with P~1AP — (G em | The mati (FIGENVALIIES AND FIGENVECTORS (Chanter 13) 337 a ne oye (EG oo) Sut) 1 ele mc cnn evil dl Cheek =i 38 gt _1f3+3% 3-37 “4\ 1-38 1437 _ fin Rae ~\ 182 547 EXERCISE 13€ Use the diagonalisation of A to find A°. 34 2 Consider the matrix A = G 2) @ Find the eigenvalues of A and their corresponding eigenvectors. 'b State a matrix P which will diagonalise A. € Write A” in the form PD"P~!, di Ilence find the matrix P~1AMP, 2 A= GC tp find the matrix A® exactly. 02 4 we= (2 i ). find the matrix C20" exactly. We have just seen that if a matrix A is diagonalisable, with eigenvalues 1, Az, then its powers can Eo be written in the form anal 5 we for ne Zt. We can readily extend this to fractional powers. For example, we can define the “square root” of Aas VA = A? = P 1 Would all matrices necessarily have a “square root™? y 2 B= A? isa solution to the matrix equation B? = A. Is it the only solution? 2 Use the above definition to verify that (a338_FIGENVALUIFS AND FIGENVECTORS (Chapter 19) In the real-world models we have considered in the course so far, we have based our model on whatever data was available and used it to generate a function or other exact representation of the system. We know that our model is not exact, but we have made no attempt to include the random variation in the system within the model. A stochastic model includes the random variation in a system by using probability to predict the ‘outcomes of future events. is a stochastic model which describes how a sequence of random events evolves over ‘© Time is measured in diserete steps. ‘© Each time step, the possible outcomes of each event are the states that the system could be in. ‘© The probability distribution for the possible states of the system at the next time step depends only on the state at the current time step. ‘This is known as the memoryless or Markov property. For example, the Opening Problem describes a Markov chain. ‘© Time is measured in days, ‘© Each day, the weather can take two possible states: “sunny” or “r ‘© Tomorrow's weather depends only on today’s weather. TRANSITION DIAGRAMS We can illustrate how the weather changes or transitions from one day to the next using a transition Giages probability it will Be rainy tomorrow sven it was sunny today | 02 ‘ S (a) " probability it will be sunny tomorrow ie probability it will be rainy tomorraw sven it was sunny today { sven it was rainy today probebilty it will be sunny tomorrow given it as rainy today Notice that: ‘© Fach state is represented by a circle. ‘* Arrows show movement between states. ‘¢ Each arrow is labelled with the corresponding transition probabilEIGENVALUES AND FIGENVECTORS (Chanter 19) 339. HISTORICAL NOTE ‘The word “stochastic” comes from the Greek word sfokhastikéis which ‘means “acting based on guesswork”. Stochastic processes were the main research focus of the Russian mathematician Andrey Markoy (1856 - 1922), after whom Markov chains were named, Today, Markov chains have many applications in genetics, economies, and telecommunications. r Andrey Markov MATRIX REPRESENTATION ‘Une method for studying Markov chains is to use matrices. STATE MATRICES The state matrix sy shows the state of the system at time n. The suite matrix may represent either: ‘+ the probability or proportion of the population that are in cach state, or ‘* the actual number from the population that are in the given state, The al state matrix is sp. For example, in the Opening Problem we suppose today represents time step nm =0. If it is sunny + Its sunny today with probability 1. 1 today then the intial state metrix: soi — ‘o= It is rainy today with probahility 0. TRANSITION MATRICES The transition matrix T organises the transition probabilities Ina transition matrix T = (tu): The columns represent the current state. ‘The rows represent the next state. ty is the probability of moving to state i from state j. t; = P(next state is i| current state is j) The sum of each column is 1. ‘The transition matrix for the Markov chain in the Opening Towtay Problem is shown alongside sunny rainy _ (08 0.6) sunny a (Gs ot) rainy morrow340 FIGENVALDIFS AND FIGENVECTORS (Chapter 19) USING MATRICES TO CALCULATE FUTURE PROBABILITIES When we studied conditional probability, we used tree diagrams Today Tomorrow help us per foun calculations. eee i 08, sumy @ For example: sa P(sunny lomorrow) = 1x 0.8+0x 0.6 {paths and@} Lei rainy —os sumy and Pliny tomariow) <1 x 124i std {pacha Dand@} nN win chance that it will be sunny and a 24% chance that it will be rainy = Check these calculations with the In the formula 3, — T'"sy, what assumption do we make about T? Discuss whether thio assumption is necessarily realistic, EXERCISE 13D.1 1 Bioblatt is a new paper supplier. The transition diagram below chows the percentages of customers who choose each week between Bioblatt and its competitors. 20% 80% 6S: 90% 10% @ Construct a transition matrix T with elements in decimal form, b In the matrix T, state the meaning of the element: 102 09 ¢ Initially Bioblatt has uo share of the market, so the initial market share column matix is = ( ) Calculate Ts and interpret your answer Find Bioblatt’s market share after: 1 two weeks il four weeks.FIGENVALIFS AND FIGENVECTORS (Chanter 13) 341 2 A country town of 1000 residents has two main grocery stores named Arrows and Bells. ‘Of those who shop at Arrows one week, 90% shop at Arrows again the next week, and 10% shop at Rells the next weak Of those who shop at Bells one week, 70% shop at Bells again the next week, and 30% shop at Arrows the next week. a Display this information using: 1 a wansition diagram Ml a transition matrix T. © In the current week, 600 people shopped at Arrows, and 400 pcople shopped at Dells. State the initial state matrix so. ¢ Find the expected number of people who will shop at Arrows: i 1 week later fi 3 weeks later. 3 Ata local high school, 30% of students who bring their lunch from home on one day will buy their lunch at the canteen the next day. 50% of students who buy their lunch at the canteen on one day will bring their lunch from home the next, 4. Display dhs information using | a transition diagram ta transition matrix‘. On Monday, 400 students from the school were surveyed. 243 brought their hunch from home, while the remainder bought their lunch at the canteen, Construct an initial state matrix sp for this situation, ¢ How many students do you expect to buy their lunch at the canteen on: 1 Tuesday fl Priday? 4% Two brands of sheep's mifk cheese are available. Brand bought this week The current transition patterns for sales of the [Baan cheeses are shown alongside. Brant bough [l : ‘@ State the meaning of the value 0.4 in the yor nok 08 06 table. © Construct a transition matrix T. ‘€ Suppose the market shares for the first week are given os > 0.2 ) 1 Find Ts; and explain what it represents. fi Find Baaah’s market share in the third week. iil, Find Sheez’s market share for the sixth week. by the state matrix. s; = ( 5 100 smokers are attempting to quit. They meet for a barbecue on the first Saturday of each month to discuss their progress and encourage each other. In January, 62 members of the group were still smoking, while the rest of the group had gone all of the previous month without smoking. St some of those who had gone without smoking will return to the habit. The proportions are described below. eg indicate that in such citations, some of the smokers will give up in the next month, while This montha Write a matrix to deseribe the state of the membership at the end of January. 08 0.1 02 09 Evaluate 19 li Estimate how many members will not have smoked between the January and February meetings. ii Evaluate T?so, and interpret the result. b Suppose T ¢ How many members do you expect to be smoking in January the following year? EL oo | 4) Self Tutor | Scientists arc attempting to reverse the effect of salt damage on land adjacent to the Hudson River. Data was collected and examined from a property near the river over several years. This allowed the scientists to construct the following table which compares the proportions of very good, usable, and very poor land between one year and the next. Situation this year Situation next @ Draw a transition diagram for this situation. b Write down the transition matrix T. € In the matrix , state the meaning of the value: i 0.91 0.03 i 0.28 d_ Find the matrix T?, and explain the meaning of the value in: i row 1, column 2 row 3, column 3. 0 © The initial land quality matrix is { 0.1 ]. Explain what this means. 09 f Find the tand quatity after: 1 one year two years Wii four years. 0.91 0.30 0.02 a b T=| 0.06 065 0.28 oi 0.03 0.05 0.70FIGENVALIIES AND FIGENVECTORS (Chanter 13) 343 ¢ I The value 0.91 means that 91% of the land which is very good this year will be very good next year I ‘The value 0.03 means that 3% of the land which 1s very good this year will be very poor next year, ill The value 0.28 means that 28% of the land which is very poor this year will be usable next year. 0.91 030 0.02\* /0.8467 0.469 0.1162 d T= | 0.06 065 0.28 0.102 0.4545 0.3792 0.03 0.05 0.70 0.0513 0.0765 0.5046 1 In row 1, column 2 we have 0.469. This means that 46.9% of land which is usable now will be very good in two years’ time, fi In row 3, column 3 we have 0.5046, This moans that 50.46% of land which is very poor now will be very poor in two years’ time. 0.9 J-—very poor and 90% is very poor this year. 091 0.30 0.02\ / 0 0.048 0.06 0.65 0.28] | 0.1 UsI7 0.635 0.03 0.05 0.70) \o.0 After one year, 4.8% is very good, 81.7% is usable, and 63.5% ie very poor 0 \ —very goad e@ = (¢i) she ) means that there is no very good land, 10% is usable, fis Tso 0.151 ii sy = Tso ~ | 0.387 After two years, 15.1% is very good, 38.7% is usable, and 0.462 46.2% is very poor. 0.363, iii sy =T's) ~ { 0.366 } After four years, 36.3% is very good, 36.6% is usable, and 0.270 27.0% is very poor. 6 A farmer is starting a soil restoration programme. For the soil type and conditions of his land, the transition matrix from one year to the next is: VG U VP VG (08 0.2 0.0 VG = Very good T= Uf[o1 o7 03 U= Usable 01 01 0.7 VP = Very poor a Draw a transition diagram for this situation, State the meaning of the number in: 1 row 3, column 1 fi row 1, column 3, ¢ Find T®, and state the meaning of the number in: 1 row 3, column 2 Ii row 2, column 3. The present land is all very poor. Write down the initial land quality matrix. © Describe the land quality after 3 years.344__FIGENVALUIFS AND FIGENVECTORS (Chapter 19) 7 A hockey team’s fitness coach has observed the following trends over a long period: Condition this week “Fully fit | Geuing weaiment | Cannot play Condition Fally fit 0.88 0.75 0.08 next — | Getting treatment [ 0.06 O17 0.42 week Cannot play 0.06 0.08 0.50 2 Draw a transition diagram for this situation, © Write down the transition matrix TT for the condition of the team’s players. © State the meaning of the number in: i row 1, column 1 li row 2, column 3. Find T?, and state the meaning of the number in row 1, column 2. The club has 30 fully fit players, 6 getting treatment, and 2 who cannot play due to injury. How many players would the coach expect in each category: 1 next week i in two weeks’ time? ACTIVITY 1 Markov chains are not always directly observable. A Markov chain will sometimes. t9Pe¥ generate a sequence of events that we can see, but we know nothing about the underlying chain itcelf, In this cace we call it a hidden Markoy chain, In this online Activity we study a hidden Markov chain based on the Opening Problem. STEADY STATES Uh aslcZ valoda! Click on the icon to access a simulation of the weather in the town in the Opening _sreaby stares Problem, What to do: 1 Run the simulation for: a month b 100 days © 1000 days. % Describe how the proportion of sunny days changes during the simulation. 3° Run the simulation again with different initial state matrices So. Comment on your results. 4 Use the formulas, =T"s) to fi as & S100 © Si000 Comment on your results 5 Repeat 4 using different initial state matrices $. What do you notice?Fil FNVALLIFS AND FIGENVECTORS (Chanter 19) 345 From the Investigation, you should have found that the proportions of sunny and rainy days converge to 0.75 =4 and 0.25 = + respectively, regardless of the initial state matrix so. When there is very little or no change in the values of the state matrices from. ‘one time step to the next, we say the system has reached a steady state, FINDING THE STEADY STATE OF A MARKOY CHAIN Ina steady state, S41 = Sp Typ — Sn _ Ts=s {letting s ay We cau derefure define, The steady state matrix s of a Markov chain is an eigenvector of the transition matrix T with eigenvalue 1. From Seetion A, we know there are infinitely many eigenvectors corresponding to a particular eigenvalue. Ihe eigenvector we need 1s the one whose elements have the correct sum. If the state matrix is a matrix of probabilities, its elements sum to 1. Otherwise, the elements sum to the total population being considered. For example, for the weather scenario in the Opening Problem, T = ( A Re) If s= 5 is the steady state matrix then Ts =s 08 0.6\(a\_ (ua 0.2 04 b)\b oga+06b\ (a 0.2a+046) ~ \b b 0. Using either row, 0.24 + 0.4l In this case, s is a matrix of probabilities, so a+b=1 3b+b=1 Hows += ( . b=4 and a= SUMMARY a ) which agrees with our observations in the Investigation, ‘We can find the steady state of a Markov chain with transition matrix T using either: Method I: Calculate s, = Tsp for large values of n. Method 2: Find s such that Ts = s. Note that in general, the steady state does not depend on so. So, if using Method 1 and sy is not given, choose any s» # 0.346, FIGENVALUIFS AND RIGENVECTORS (Chapter 19) EXERCISE 13D.2 1. Jada has noticed that some people at her work keep leaving the fridge door partly open. She estimates 3 Suppose T= ( n'a 9) is a tansition mattix for a discrete dynamical system. The state mate 4A netball squad has 40 players. 30 are currently fully fit, 5 are getting treatment, and 5 are unavailable due to injury. The weekly transition matrix for the fitness status of the players is: that the door is currently left open 10% of the time, and she suspects the problem is getting worse. Jada has noticed that the state in which a person leaves the fridge depends on the state in which they found it, as described in the table alongside: Si OL 0.3 0.05 suman (38) ann (88 B88 State when they arrive State when 30% 5% Dulles as a Explain the significance of sq and T. b Find algebraically, the steady state of the system, s. € Hence prediet the proportion of time the fridge will be left open in the long term. Ts Jada’s concen that the problem is getting worse justified? Use Tsp to check your answer to b. Iwo bus services, Clydes and Roos, operate between Now ‘two cities, The transition matrix alongside chows the cB preferences of passengers who use these servi monthly basis. This month Clydes carried 425 passengers and Roos soma 4 _ (084 0.21) C Next 0.16 0.79) R- month ried 716 passengers, Write down the initial state unatsia 90 © Predict the monthly number of passengers for each bus service T's for large n. € List two assumptions you made in 6, Do you think these assumptions are reasonable? d_ Use b to estimate the proportion of passengers that use each bus service in the long term. Find the steady state proportions for the system algebraically. the long term by considering 01 04 of the system gives the proportions of the population with a particular characteristic, a Use the charactenstie polynomial ot 1 to tind its exgenvalues, & Find the corresponding eigenvectors of T. ¢ Hence find the steady state matrix s of the system. 0.86 0.68 0.00 \\ — fully fit 0.12 0.24 0.32 }— getting treatment 0.02 0.08 0.08 / — injured a How many players would be expected in each group: i next week fi after two weeks Iii in the tong term’ b IP the squad fields three teams of 9 cach week. discuss the implications of your answer to a iil. © Under these transition conditions, how many members would the squad require to ensure they would have a fit team for each of the three games? Explain your answer.FIGENVALIFS AND FIGENVECTORS (Chanter 13) 347 5 Each year there is considerable bird migration between the islands Paua, Manu, and Chalk. Over time, @ researcher has observed the trends shown in the table. a Find the transition matrix T which shows the Next migration pattern of the birds from one year to Year the next & Find 2, and explain what the numbers in its third column represent. ¢ In 2019, 26% of the birds lived on Paua, 39% lived on Manu, and 35% lived on Chalk. Estimate the populations on each island in: 1 2020 Nl 2022, 1d Tind the steady state proportions of birds living on cach island, ‘@ Once the steady state proportions have been reached, do the birds stop migrating between the islands? If not, what percentage of birds move between islands each year? 6 A legal firm has four city offices A, B, C, and D. A courier delivers original documents between the offices. After each delivery he waits for a new delivery from that location. The probabilities for the 0 0.4 0.2 0.1\-—A : ae a _ {05 0 05 03)—B courier’s movement are summarised in the transition matix T= {93 gi 9 ge —c: a a5 aa 9/1 a If the conrier is now at R, what is the chance that his: i first delivery goes to C fi second delivery goes to A Ill. third delivery goes to D? b If the courier is now at B, where is he most likely to be: i after 2 deliveries it after 3 deliveries? © In the long run, which office does the courier visit the most? l-p and a state matrix which p i-@ 7 Consider a Markov chain with transition matrix‘ ( represents proportions of the population. a Show that the eigenvalues of T are 1 and 1—p—q. a p+a Pp p+a jcance Of the steady state proportion Show that the steady state matrix for this system is © Interpret the physical signi348 FIGENVALUIFS AND FIGENVECTORS (Chapter 19) ACTIVITY 2 In the transition matrices we have studied so far, the elements have all been probabilities. In the ease of populations, they describe how the members of the population change from one state to another. Each member always takes a state, so the sum of the elements in each column is 1. By conttast, if we wish fo study the age distribution of a population, we need to include births and deaths. This means that new members are added to the population, while existing members are removed. The elements of the transition matrix are not necessarily probabilities, but are rates. The transition matrix for a model of this form is called a Leslie matrix, named after its inventor, the Scottish and later Australian mathematician, Patrick Leslie (1815 - 1881). A Leslie matrix has the structure ft fa fe Sa ‘We can leave ort 3 0 0 0 the last column if 0 82 0 0 itis all zeros. a 6.) ae where fi is the average number of female offspring born to each female in the ith age group 4; is the proportion of female offspring from the ith age 1up which survive w dhe next The difference between a Leslie population model and a Markov chain is that fj +s; is not necessarily 1 What to do: 1 Each yeu, a female adult penguin lays an average of 4 eggs, each of which lms x 50% chance af containing a female chick Fach year, 80% of the female chicks die, and each fomale adult has a 60% chance of surviving to the next year. M 2 a The Leslie matrix for this system is L — ( nonGG represents, Construct a transition diagram for the system, € Suppose that initially there are 250 female chicks and 300 adult female penguins in the cna = (22) ) Explain what each element of L 1 Calculate Ls and Ls, and interpret their meaning, fi By calculating L*sp and L'°so, predict what will happen to the population over time. dA particularly harsh winter means that in one year only a small number of chicks survive. ‘The population is reduced to sy = (a) Is the population likely to recover? Explain your answer e Tet s= (3) he a matrix of the pmportions of chicks and adults in the steady state female populatiFIGENVALIIES AND FIGENVECTORS (Chanter 13) 349 Research into a rare butterfly has revealed that after 24 hours, caterpillars will tum mto buttertlies, and after 18 hours, the butterflies are ready to mate. Any female butterflies that survive this long will give birth id die after 72 hours. $ of the female caterpillars die before pupating into die a buuerfly, and 3 of dhe adult female buwer before mating. ‘On average, the female butterflies that survive to mate produce 12 female caterpillar offspring. At a particular time there are 120 female caterpillars, 15 adolescent female butterflies, and $ adult female butterflics in a colony. 00 12 a The Leslie matrix for this system is L= {> 0 0 030 Explain what cach clement of L represents. b& Construct a transition € According to the model, how many females will there be in each category after: i 24 hours i 48 hours iii 72 hours? Predict what will happen to the butterfly population in the future using s, = Ls for large values ot 7, \gram for the system, REVIEW SET 13A 1 » = ee a the characteristic polynomial b the eigenvalues. —7 e 3-7 eee) a Find Ax, and hence state the eigenvalue corresponding to x. Find the remaining eigenvalue. Consider the matrix A = ‘6 el a Find the eigenvalues A1, A2 (Ar > Aa) and corresponding eigenvectors x1, % of A. 1b Verify that Ax; — Ajay and Axy = ox. Find the eigenvalue(s) and corresponding eigenvector(s) of: ‘co aaq(4, 5 baA= 24 Let A= é H a Find the eigenvalues ,, Lz (A; > As) and the corresponding eigenvectors x, X2 b Verify that P = (x: |¥2) will diagonalise A.350) FIGENVALIIFS AND RIGENVECTORS (Chapter 19) 10 -9 ( a } ( ry } Use the diagonalisation of A to find A®. Ir c e vi find the matrix C! exactly. Each night, the 500 passengers on a cruise ship can eat at restaurant A or restaurant B. 70% of passengers who eat at A one night will eat at A again the next night. 80% of passengers who eat at B one night will eat at B again the next night. a Display this infarmation using: Ta transition diagram Ha transition matrix. © On Sunday, 250 passengers ate at each restaurant. How many passengers will eat at: I restaurant A on Monday Hi restaurant B on Tuesday? ‘A university student is researching the connection hetween adults who smoke and the smoking habits of their children. Her results are presented in the table below. This generation ‘Smoker | Non-smoker | 55% 5% 45% 95% Next generation Construct a transition matrix from the table. & In the matrix T, state the meaning of the value in row 2, column 1. € For the generation in which data was first collected, 60% of adults smoked, so re (Gea Calculate Tsp and interpret your answer, d__ Find algebraically the steady state of the system s, Hence predict the proportion of adult smokers in the long term. Billi a taxi diver. He collects customers from taxi rank A, B, This time or C, After driving a customer to their destination, he returns i to a taxi rank to collect his next customer. mao Ihe table alongside shows the probabilities with which he Next = returns to cach taxi rank Pisa) ao a Display this information nsing: ea 03 1 a transition diagram Wa transition matrix T. Interpret the element in row 1, column 3 of T. € Bill started his shift at taxi rank B. He is now returning to a taxi rank to collect his third customer. Which taxi rank is he most likely to return to? In the Jong run, which taxi rank does Bill use most often?FIGENVALIIES AND FIGENVECTORS (Chanter 13) __ 351 tS ET 1. Find the eigenvalue(s) off Sea) Oe (ae) | ( ie a) has eigenvalues 3+ V8. Find the value of k. 2 The matrix B=(, 5 3. Consider the matrix A = (G a Sed renee eecarrrnatemeeeeteen 6 Without performing any calculations. find the eigenvalues and corresponding eigenvectors of 3A. een aus a= (3 5) a Show that ( a ) 4 t£0 are cigenveetors of A. State the corresponding cigenvaluc. b State the other eigenvalue of A. € Hence find the remaining eigenvectors of A. 5 Find the eigenvalue(s) and corresponding eigenvector(s) of: 2-9 5 -2 ane aa-(i ra va-( 7) «a-( 1 | 6 Consider the matrix B= «@ ‘) a Find the eigenvalues and corresponding eigenvectors of B. '& Write down a matrix P that diagonalises B, and find P~'BP. © Hence calculate. iB ii BY. ~, 76 7 Consider the matrix A = 89 a Find matrices P and D such that A = PDP! & Hence fiud dhe exact value uf A" 8 Chris and Suzie are candidates seeking to be elected mayor at the next council elections. Telephone polls of people living in their electorate have revealed the trends in voting patterns shown in the table. @ Display this information using: 1 a transition diagram Ha transition matrix T b Calculate T? and T°, © Explain what column 2 uf T? sucan, Ne [Cinis | 307% [15% week [Suzie | 20% | 85%352_FIGENVALDIFS AND FIGENVECTORS (Chapter 13) d__ The polls currently suggest that Chris is winning with 60% of the vote. Who do you expect will win if the election 1s held in: i 2 weeks Hi 4 weeks? Y Doctors classify people as underweight, healthy weight, or overweight. The following table shows the connection between the weight classifications of people and their children. This generation Underweight] Healthy] Underweight] 0.43 | 0.18 Next = generation |_Heatthy. Oe eM Tees | Overweight [0.10 0.26 @ State the uansition matrix T, and calcula T?. 1 What percentage of underweight people of the current generation are expected to have: 1 healthy weight children Wi overweight grandchildren? © Currently the weight status of people is 15% underweight, 56% healthy, and 29% overweight. Find how these proportions are expected to change: 1 in the next generation Win two generations, Tiind the stcady state for this model, and interpret your answer @ Do you think the underlying assumptions of a Markov chain are appropriate in this case? Explain your answer, 10 Ina weekly art class, 120 students can choose between painting and pottery. Each week, 40% of the painting students move to pottery, and 20% of the pottery students move to painting. Initially, 60 students chose painting and 60 students chose pottery. a Write down the initial state matrix sp and the transition matrix T. Find Ts, and interpret your answer. Find the eigenvalues and corresponding eigenvectors of T. Write on expression for the number of painting students after m weeks, m CN. Hence find the number of painting students each week in the long term.