CHAPTER 5
R E C U R R E N C E R E L AT I O N S
VARIATIONS ON SOLVING RECURRENCE
RELATIONS WITH CONSTANT COEFFICIENTS
• We will discuss some extra problems
• The first group of problems will just be variations on LHRRCC
– Never forget what LHRCC stands for
VARIATION 1: THE CHARACTERISTIC
POLYNOMIAL HAS HIGH DEGREE
• We only solved problems with a quadratic characteristic polynomial
• That doesn’t have to happen
• Here is a recurrence relation with a characteristic polynomial of higher degree
– For example, an = 2an-1 + an-2 – 2an-3
• This is solvable
– For example, an = an-1 – + an-2 – an-3
• This is s not directly solvable
• It can be “solved” numerically
VARIATION 2: THE RECURRENCE
RELATION IS NOT HOMOGENEOUS
• So now LHRRCC has change to LRCC
• For this kind of problem, we use a table
– This table is explained (in words, not as a table) on p. 244 as Problem 40
– I have included the table here
• There is an extra step for nonhomogeneous problems that we didn’t have with
homogeneous problems
– You have to find a particular solution to the nonhomogeneous problem
– You have to find a general solution to the HOMOGENEOUS version of the problem
– The solution is the sum of the two
THE TABLE FOR NONHOMOGENEOUS
PROBLEMS
• The original equation is an = c1an-1 + c2an-2 + f(n)
• The characteristic equation is then x2 – c1x – c2 = 0
Is 1 a root of the 1 is not a root 1 is a single root 1 is a double root
characteristic polynomial?
f(n) Format for solution to particular equation
(constant) D Dn Dn2
Cn D1n + D2 D1n2 + D2n D1n3 + D2n2
Cn2 D1n2 + D2n + D3 D1n3 + D2n2 + D3n D1n4 + D2n3 + D3n2
AN EXAMPLE OF A NON-HOMOGENEOUS
RECURRENCE RELATION
• Let’s try to solve an = 6an-1 - 8an-2 + 3
• The homogeneous version of the problem is an = 6an-1 + 8an-2
– The characteristic equation is?
– The roots of the characteristic equation are?
– The solution to the homogeneous equation is?
• Stop and solve this part of the problem
AN EXAMPLE OF A NON-HOMOGENEOUS
RECURRENCE RELATION
• Let’s try to solve an = 6an-1 – 8an-2 + 3
• The homogeneous version of the problem is an = 6an-1 – 8an-2
– The characteristic equation is x2 – 6x + 8 = 0
– The roots of the characteristic equation are 2 and 4
– The solution to the homogeneous equation is an = R(2n) + S(4n)
AN EXAMPLE OF A HOMOGENEOUS
RECURRENCE RELATION
• Now, we solve the nonhomogeneous problem
– The non-homogeneous part is f(n) = 3
• In order to use the table, we need to know if 1 is a root of the characteristic equation
– The answer is no, since the roots are 2, 4
• The table says we should look for a solution to have the form
an = D
• Substituting this into the non-homogeneous equation gives
D = 6D – 8D + 3
• Solving this equation gives D = 1
PUTTING IT ALL TOGETHER
• So the solution is
an = R(2n) + S(4n) + 1
• You have to go through this entire process every time
• If initial conditions are given, you can determine R and S:
• Suppose a0 = 4 and a1 = 11
• Then we have 4 = R(20) + S(40) + 1 and 11 = R(21) + S(41) + 1
• Solving this system gives R = 1 and S = 2, for the solution to the original problem being
an = 2n + 2(4n) + 1
A SECOND EXAMPLE
• Let’s try to solve an = 2an-1 + 8an-2 + 81n2
• The homogeneous version of the problem is?
• The characteristic equation is?
• The roots of the characteristic equation are?
• The solution to the homogeneous equation is?
• See if you can also determine what f(n) is
• Stop and solve this part of the problem
THE ANSWERS
• The equation is an = 2an-1 + 8an-2 + 81n2
• The homogeneous version of the problem is an = 2an-1 + 8an-2
– The characteristic equation is x2 – 2x – 8 = 0
– The roots of the characteristic equation are –2, 4
– The solution to the homogeneous equation is an = R(-2)n + S(4n)
f(n) = 81n2
CONTINUING ON
• Since 1 is not a root of the characteristic equation, the particular solution has the form
an = D1n2 + D2n + D3
• Notice that
an-1 = D1(n-1)2 + D2(n-1) + D3
an-2 = D1(n-2)2 + D2(n-2) + D3
• The recurrence relation an = 2an-1 + 8an-2 + 81n2 becomes
D1n2 + D2n + D3 = 2[D1(n-1)2 + D2(n-1) + D3] + 8[D1(n-2)2 + D2(n-2) + D3] + 81n2
FINDING THE CONSTANTS-PART 1
• We have
D1n2 + D2n + D3 = 2[D1(n-1)2 + D2(n-1) + D3] + 8[D1(n-2)2 + D2(n-2) + D3] + 81n2
• Expanding powers of n gives
D1n2 + D2n + D3 = 2D1(n2-2n+1) + 2D1D2(n-1) + 2D1D3 + 8D1(n2-4n+4) + D2(n-2) + D3 + 81n2
• Distributing and grouping based on powers of n gives
(–9D1 – 81)n2 + (36D1 – 9D2)n + (34D1 + 18D2 – 9D3) = 0
• Since this must be true for all n, we have three separate equations
(–9D1 – 81) = 0
(36D1 – 9D2) = 0
(34D1 + 18D2 – 9D3) = 0
FINISHING THE PROBLEM
• Solving the first equation gives D1 = –9
• Substituting that into the second equation and solving gives D 2 = –36
• Substituting both D1 and D2 into the third equation and solving gives D3 = –38
• So the particular solution is an = –9n2 – 36n – 38
• The solution to the original recurrence relation is
an = R(-2)n + S(4n) – 9n2 – 36n – 38
A PROBLEM TO TRY
• an = 4an-1 – 3an-2 + 20
• The homogeneous version of the problem is?
• The characteristic equation is?
• The roots of the characteristic equation are?
• The solution to the homogeneous equation is?
• See if you can also determine what f(n) is
• Stop and solve this part of the problem
FINISHING THE PROBLEM
• You can see that 1 is a root of the characteristic equation
• This means the particular solution has the form a n = Dn
• Then
an-1 = D(n-1) = Dn – D and an-2 = D(n-2) = Dn – 2D
• We need to solve an = 4an-1 – 3an-2 + 20
Dn = 4Dn – 4D – 3Dn + 6D + 20
• Continuing to solve gives the particular solution of
an = –10n
• For the full solution, find the general solution, and add –10n
THE HALTING PROBLEM
• There is one leftover idea from this chapter
• This is the Halting Problem
• Can you write a program that
– takes as its input another program and that program’s input, and
– determines if that program will run infinitely or stop?
THE HALTING DECIDER
• Suppose you did have a program that could decide
• Let’s call it the Halting Decider.
• I will create another program
• Let’s call it the Halting Undecider
– This program also takes as its input a program
– This program does the opposite of its input.
• If the input program stops, the new program runs forever, (like an infinite loop)
• If the input program runs forever, the new program stops.
HALTING OR NOT?
• Now, we take the Undecider, use it as input to itself, and give this to the Halting Decider
• Now, either a program stops or it doesn’t
– We know this because the Halting Decider can tell us this
• Let’s look at the new Undecider program
– If it stops, it runs forever
– If it runs forever, it stops
• So, there is no Halting Decider, because it can’t tell us if this program halts or not
• We call the halting problem “undecidable”
– This means that it’s impossible to decide if the problem has a solution or not
PRACTICE PROBLEMS
• p. 244, #41-47