You are on page 1of 6

Iteration of Linear Functions

Mckenzie H. Fox
February 2, 2016

1 Introduction
1.1 Abstract
Iteration in mathematics, is the repetition of a mathematical or computational procedure applied to the
result of a previous application, typically as a means of obtaining successively closer approximations to
the solution of a problem. In this article we will be studying the effects of iteration on linear functions
by by taking successive iterations of a linear function of the form f (x) = ax + b, starting with an initial
seed value designated x0 . The iteration process can be summarized with the recurrence relation:

xn+1 = axn + b (1)


Using the data obtained from this experiment, we will form conjectures pertaining to the effects of
iteration on linear functions, and develop mathematics that will predict the behavior of iteration on
linear functions based on values of; a, b, and x0 .

1.2 Definitions
1.2.1 Linear Function
A linear function is a first degree polynomial of the form,
f (x) = mx + b
where m and b are constants and x is a real variable. The constant m is called slope and b is called
y-intercept. The y coordinate is obtained by setting y = f(x).

1.2.2 Recurrence Relations


A recurrence relation is an equation that defines a sequence based on a rule that gives the next term of
said sequence, as a function of the previous term(s). If we denote the nth term in the sequence by xn , a
recurrence relation is of the form
xn+1 = f (xn )
To generate sequence based on a recurrence relation, one must start with some initial seed value(s). For
a first order recursion, a single seed value of x0 can generate all remaining terms using the recurrence
relation.

1.2.3 Convergence of a Sequence


A sequence Sn converges when it’s terms tend towards a constant value L. This constant value L is
the limit of the sequence. In other words, a sequence Sn converges to L if there exists a function f :
R>0 → R such that;
∀x > 0 : (L − x, L + x) ⊃ {sn : n > f (x)}

1.2.4 Divergence
If Sn does not converge as described above, then the sequence diverges, or does not tend towards any
constant value L.

1
1.2.5 Geometric Series
For the purpose of algebraic simplification
P later in this article, a definition of the geometric series is
included as follows. A geometric series k ak is a series for which the ratio of each two consecutive
terms ak+1 /ak is a constant function of the summation index k.

For the simplest case of the ratio a( k + 1)/ak = r equal to a constant r, the terms ak are of the form
ak = a0 rk . Letting a0 = 1, the geometric sequence ak nk=0 with constant |r| < 1 is given by

n
X n
X
Sn = ak = rk = 1 + r + r2 + r3 . . . + rn
k=0 k=0

Simplification of this formula yields


n
X 1 − rn+1
Sn = rk =
1−r
k=0

2 Technical Results
2.1 Iterative Analysis in Spreadsheet
A spread sheet was created to track the values of xn+1 in Equation (1) and to study the effects that
variation of parameter values a, b, and x0 had on xn+1 as n → ∞. The value of a was varied between -2
and 2 in increments of 0.2, the value of b was varied between -3 and 3 in increments of 0.2. A portion
of this spreadsheet has been included below for reference.
a b x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 x10
-2.00 -3.00 3.00 -9.00 15.00 -33.00 63.00 -129.00 255.00 -513.00 1023.00 -2049.00 4095.00
-1.80 -3.00 3.00 -8.40 12.12 -24.82 41.67 -78.00 137.41 -250.33 447.60 -808.68 1452.62
-1.60 -3.00 3.00 -7.80 9.48 -18.17 26.07 -44.71 68.54 -112.66 177.25 -286.60 455.57
-1.40 -3.00 3.00 -7.20 7.08 -12.91 15.08 -24.11 30.75 -46.05 61.47 -89.06 121.68
-1.20 -3.00 3.00 -6.60 4.92 -8.90 7.68 -12.22 11.67 -17.00 17.40 -23.88 25.65
-1.00 -3.00 3.00 -6.00 3.00 -6.00 3.00 -6.00 3.00 -6.00 3.00 -6.00 3.00
-0.80 -3.00 3.00 -5.40 1.32 -4.06 0.24 -3.20 -0.44 -2.65 -0.88 -2.29 -1.17
-0.60 -3.00 3.00 -4.80 -0.12 -2.93 -1.24 -2.25 -1.65 -2.01 -1.79 -1.92 -1.85
-0.40 -3.00 3.00 -4.20 -1.32 -2.47 -2.01 -2.20 -2.12 -2.15 -2.14 -2.14 -2.14
-0.20 -3.00 3.00 -3.60 -2.28 -2.54 -2.49 -2.50 -2.50 -2.50 -2.50 -2.50 -2.50
0.00 -3.00 3.00 -3.00 -3.00 -3.00 -3.00 -3.00 -3.00 -3.00 -3.00 -3.00 -3.00
0.20 -3.00 3.00 -2.40 -3.48 -3.70 -3.74 -3.75 -3.75 -3.75 -3.75 -3.75 -3.75
0.40 -3.00 3.00 -1.80 -3.72 -4.49 -4.80 -4.92 -4.97 -4.99 -4.99 -5.00 -5.00
0.60 -3.00 3.00 -1.20 -3.72 -5.23 -6.14 -6.68 -7.01 -7.21 -7.32 -7.39 -7.44
0.80 -3.00 3.00 -0.60 -3.48 -5.78 -7.63 -9.10 -10.28 -11.23 -11.98 -12.58 -13.07
1.00 -3.00 3.00 0.00 -3.00 -6.00 -9.00 -12.00 -15.00 -18.00 -21.00 -24.00 -27.00
1.20 -3.00 3.00 0.60 -2.28 -5.74 -9.88 -14.86 -20.83 -28.00 -36.60 -46.92 -59.30
1.40 -3.00 3.00 1.20 -1.32 -4.85 -9.79 -16.70 -26.38 -39.94 -58.91 -85.47 -122.66
1.60 -3.00 3.00 1.80 -0.12 -3.19 -8.11 -15.97 -28.55 -48.69 -80.90 -132.44 -214.90
1.80 -3.00 3.00 2.40 1.32 -0.62 -4.12 -10.42 -21.76 -42.17 -78.90 -145.02 -264.04
2.00 -3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00

Made apparent from the data is the following;


1. When −0.8 < a < 0.8, the iterative sequences converge to a limit regardless of other parameter
values. What the limit is for any given iterative sequence has yet to be determined however and
varies greatly depending on parameter values. It is assumed that the values in the iterative sequence
will converge when |a| < 1, and diverge otherwise.
2. Whenever xn = xn+1 the iterative sequence also converges. The limit in these cases appears to be
x0 .
To better visualize convergence and divergence of Equation (1) when varying parameters, a chart was
created to visually track these phenomena.

2
2.2 Visual Analysis of Convergence

Figure 1: Convergence Chart

Upon inspection of the chart, we can see that conjecture 1 appears true with the exception of when
xn = xn+1 but we have not used values between 0.8 and 1.0 or -0.8 and -1.0 so the interval of convergence
for ”a” may or may not be [-0.8 0.8]. The outliers that occur when xn = xn+1 appear to occur linearly
as seen on the chart.

2.3 Mathematical Analysis of xn+1 = axn + b


So far we have only performed computational analysis of the iterative sequence to determine and identify
trends. In this section we will;

1. Analyze the iterative sequence to determine a general formula for Xn .


2. Use mathematical induction to prove the general formula is correct.
3. Analyze the specific cases promulgated in conjectures 1 and 2 found under section 2.1, and if
possible, provide proof of convergence.

3
2.3.1 General Formula for Xn
To find the general formula for Xn we will write out the first few terms of the sequence abstractly so
that terms may be combined and simplified.

x1 = ax0 + b
x2 = a2 x0 + b(1 + a)
x3 = a3 x0 + b(1 + a + a2 )
x4 = a4 x0 + b(1 + a + a2 + a3 )
x5 = a5 x0 + b(1 + a + a2 + a3 + a4 )

We can see by inspection that the second term in the iteration is the same as the geometric series
expansion shown in definition 1.2.5 above. When simplified by substituting the geometric series for the
second term, it can be seen that any term in the iteration sequence can be directly calculated according
to the following formula for Xn .
n−1
X
xn = a n xo + b ak
k=0

1 − an
xn = an xo + b
1−a

2.3.2 Proof by Induction of Xn


Claim 1. For any n > 0 ∈ N,
1 − an
xn = an xo + b = axn−1 + b
1−a
Base Case:

P (1) : x1 = ax0 + b = ax0 + b True for n=1

P (2) : x2 = a2 x0 + b(1 + a) = a2 x0 + b(1 + a) True for n=2

Induction Hypothesis:

Suppose P (k) is true for some n = k, k ∈ N


k
P (K) : xk = ak xo + b 1−a
1−a = axk−1 + b. Then,

Inductive Step:

Let n = k + 1. Then,
1 − ak+1
P (k + 1) : xk+1 = ak+1 xo + b = axk + b.
1−a
1 − ak
→ = a(ak xo + b )+b
1−a
1 − ak+1
→ = ak+1 xo + b
1−a
k+1
1 − a 1 − ak+1
ak+1 xo + b = ak+1 xo + b
1−a 1−a
P (k + 1) : True for n = k + 1

4
Conclusion: By the principle of mathematical induction, the theorem holds for all n > 0 ∈ N.

2.3.3 Convergence and Divergence of xn+1 = axn + b.


As promulgated in conjecture 1 under section 2.1, we will check for convergence of the terms of the
iterative sequence when |a| < 1. If |a| < 1, then a can be represented by the rational number p1 .
p > 1 ∈ R. In other words,
1
a=
p
1
→ an = (± )n
p

Substituting (± p1 )n into the general equation for Xn and taking the limit as n → ∞ shows us that the
the terms of the iterative sequence converge to a single value when |a| < 1.
1 b b
lim ± n
x0 + =
n→∞ p 1−a 1−a
It can also be seen from the equation above that when |a| > 1 the terms in the iterative sequence will
grow to infinity.There are two other cases of Xn that are dependent on a however, when a = ±1.

When a = 1:

Xn+1 = (1)xn + b
X1 = X0 + b
X2 = (X0 + b) + b
X3 = (X0 + 2b) + b

We can assume by inspection that when a = 1, the nth term in the iterative sequence an be deter-
mined by Xn+1 = X0 + nb and if that is true, as n → ∞, the terms will grow ever larger. The data in
the spread sheet coincides with this conjecture.

When a = −1:

Xn+1 = (−1)xn + b
X1 = −X0 + b
X2 = −X0 + 2b
X3 = X0 − b
X4 = X0
X5 = −X0 + b

In this case, we see that the iterative sequence has a cycle of values. As n → ∞ the value of Xn will
oscillate between a set few numbers. This shows that when a = −1, the sequence will diverge.

5
The final case we will be looking at, derived from the experimental data is conjecture 2. We have
seen that when Xn = Xn+1 , the iterative sequence appears to converge immediately to its limit.

When Xn = Xn+1 :

a(X0 ) + b = X0
X0 − a(X0 ) = b
X0 (1 − a) = b
b
X0 =
1−a
b b
When X0 is equal to 1−a , each following term in the sequence will also be equal to 1−a . So the
iterative sequence converges for these initial seed values. It is show here that the outliers that appear
b
linear on the convergence chart are in fact linear, with a slope of 1−a .

3 Conclusion
We have looked at iteration of linear functions and recognized trends in the behavior of Xn , when varying
the sequences other parameters. Using this information we developed mathematics and provided proof
of their use for analyzing the behavior of the iterative sequence. We determined under what conditions
the sequence will converge and diverge when dependent on a, as well as determined the limit of the
sequence when it does converge.

Further analysis of of the cases when a = ±1 could be performed to provide proof of both the general
formula for when a = 1, Xn+1 = X0 + nb, and for the oscillating nature of Xn+1 = axn + b when a = −1.
Also, making the steps in variation of applicable parameters smaller may help to identify further trends.
Both are worth looking into.

You might also like