You are on page 1of 24

Alexander White

1

MEI C3 Coursework
Introduction

During this coursework I will be using 3 different numerical methods to solve a range of polynomials,
for each method I will be giving a brief explanation, an example of a root being found successfully
and then an example of the method failing. Once I have demonstrated the methods I will then be
comparing each method against each other.

Change of Sign

When using a change of sign method you would first find the intervals that the roots lie in, either
graphically or by substituting values of into the equation and looking for a change in sign of the
value. Once you have your intervals you would then split this interval down again looking for a
change of sign, and then repeat this process until the root is found to the required accuracy.

I will be using decimal search; in this method once I have found my first interval for a root I would
then go up in the interval in increments finding another change of sign, which would then give me a
new smaller interval, which I will then take smaller increments of finding another change of sign, I
will be repeating this method until I find the root to a required accuracy.

I will be finding one root of the equation

to 5 decimal places using the decimal


search method



This is the graph of the function



The graph of the function

shows that the roots lie in the intervals [-1, 0] [0, 1]


and [4, 5]







2 2 4
15
10
5
5
x
y
y=x5x+2
Alexander White

2

I will find the root that lies in the interval [0,1] to 5 decimal places

First I will show that there is a change of sign at the interval by calculating ()


when and

()

()


= 2

()

()


= -2

There is a change in sign of () and () which means that the root does lie in the interval [0,1], I
will now take increments of 0.1 to look for another change of sign

() ()

()


= 1.951

() ()

()


= 1.808

I will continue these calculations in Microsoft Excel

x f(x)
0 2
0.1 1.951
0.2 1.808
0.3 1.577
0.4 1.264
0.5 0.875
0.6 0.416
0.7 -0.107
0.8 -0.688
0.9 -1.321
1 -2


From these calculations I can see there is a change of sign between () and () this gives me
a new interval of [0.6, 0.7]

I will now take increments of 0.01 to look for another change of sign

() ()

()


= 0.366481

() ()

()


= 0.316328




Alexander White

3

I will continue these calculations in Microsoft Excel













From the calculations I can see there is a change of sign between () and () this gives me
a new interval of [0.68, 0.69]

I will now take increments of 0.001 to look for another change of sign

() ()

()


= -0.002432

() ()

()


= -0.00841

I will now continue these calculations using Microsoft Excel

X f(x)
0.68 0.002432
0.681 -0.00298
0.682 -0.00841
0.683 -0.01383
0.684 -0.01927
0.685 -0.02471
0.686 -0.03015
0.687 -0.0356
0.688 -0.04106
0.689 -0.04652
0.69 -0.05199

From the calculations I can see there is a change of sign between () and () this gives
me a new interval of [0.680, 0.681]




X f(x)
0.6 0.416
0.61 0.366481
0.62 0.316328
0.63 0.265547
0.64 0.214144
0.65 0.162125
0.66 0.109496
0.67 0.056263
0.68 0.002432
0.69 -0.05199
0.7 -0.107
Alexander White

4

I will now take increments of 0.0001 to look for another change of sign

() ()

()


=0.001891

() ()

()


=0.001349

I will now continue these calculations using Microsoft Excel

X f(x)
0.68 0.002432
0.6801 0.001891
0.6802 0.001349
0.6803 0.000808
0.6804 0.000266
0.6805 -0.00028
0.6806 -0.00082
0.6807 -0.00136
0.6808 -0.0019
0.6809 -0.00244
0.681 -0.00298

From the calculations I can see there is a change of sign between () and () this
gives me a new interval of [0.6804, 0.6805]

I will now take increments of 0.00001 to look for another change of sign

() ()

()

+2
=0.000266

() ()

()


= 0.000158

I will continue these calculations in Microsoft Excel

X f(x)
0.6804 0.000266
0.68041 0.000212
0.68042 0.000158
0.68043 0.000104
0.68044 4.98E-05
0.68045 -4.4E-06
0.68046 -5.9E-05
0.68047 -0.00011
0.68048 -0.00017
0.68049 -0.00022
0.6805 -0.00028
Alexander White

5

From the calculations I can see there is a change of sign between () and () this
gives me a new interval of [0.68044, 0.68045]

I will now take increments of 0.000001 to look for another change of sign

() ()

()


= 4.44
x10
-5

() ()

()


=3.9
x10
-5


X f(x)
0.68044 4.98E-05
0.680441 4.44E-05
0.680442 3.9E-05
0.680443 3.35E-05
0.680444 2.81E-05
0.680445 2.27E-05
0.680446 1.73E-05
0.680447 1.19E-05
0.680448 6.47E-06
0.680449 1.06E-06
0.68045 -4.4E-06

From these calculations I can see there is a change of sign between () and ()
giving me a new interval of [0.680449, 0.68045]

As both of these numbers round up to 0.68045 to 5 decimal places

To 5 decimal places




0.2 0.4 0.6 0.8 1
2
1
1
2
x
y
Root
()
()
Alexander White

6

This is a zoomed in picture of my graph showing that the root does lie in the first interval [0,1], the
decimal search method then works by finding a smaller interval that the root lies in, for this root the
next interval was [0.6,0.7]


This shows that the next interval for this root is [0.6, 0.7]

Failures of Change of sign methods

Under certain situations the change of sign methods will not work, for example the equation





The graph of ()

shows that it has a repeated root in the


interval [-1, 0], this will cause problems with the change of sign methods as at a repeated root there
is no change of sign, which means that a new interval will not be found.






0.6 0.62 0.64 0.66 0.68 0.7
1
0.5
0
0.5
1
x
y
Root
2 1 1 2 3 4
6
5
4
3
2
1
1
2
x
y
y=x-2.25x-1.3125x-0.171875
()
()
()
()
Root
Alexander White

7

x f(x)
-1 -2.10938
-0.9 -1.54213
-0.8 -1.07388
-0.7 -0.69863
-0.6 -0.41038
-0.5 -0.20313
-0.4 -0.07088
-0.3 -0.00762
-0.2 -0.00738
-0.1 -0.06413
0 -0.17188

Taking increments of 0.1 in the interval [-1, 0] shows that there is no change of sign



A zoomed in graph of the repeated root shows that there is no change of sign; this means that a
change of sign method will fail as I will not generate a second interval to continue the process with.
















0.3 0.28 0.26 0.24 0.22 0.2
0.01
0.008
0.006
0.004
0.002
0
x
y
Root
() ()
Alexander White

8




Newton-Raphson method

The Newton-Raphson method is an example of fixed point iteration, unlike change of sign methods I
will use an iterative process to converge to the roots. After finding the interval that a root lies in I
would take first estimate of the root

I would then plot the point (

)) and draw a tangent to


the curve at this point, giving me a second estimate at the point where the tangent crosses the
axis (

), this process is repeated until my estimates round to a degree of accuracy giving me the
root. However since drawing in the tangents would be inaccurate it can be done using an iterative
process and achieve the same results.
I will solve the equation

to 5 decimal places using the Newton-Raphson method


Graph of ()



From the graph I can see that the roots lie in the intervals [-2,-1] [-1, 0] and [1, 2], I will now use
Newton-Raphson to find these roots
The general formula Newton-Raphson is

)
(

)

Before I can start finding the roots I need to find () for my function
()

()


Now that I have this I can generate my iterative formula for this function


2 1 1 2
6
4
2
2
4
x
y
y=x-5x-2
Alexander White

9

Now that I have my iterative formula I can now begin to find the root in the interval [-2,-1], I will use

as my first estimate

()
()

()
()


= -1.6800002

()
()

()
()


= -1.4795169
I will now continue these calculations using Autograph

From these calculations I can see that Autograph has found the root
The root is to 5 decimal places
To confirm this I will now use error bounds to see if there was change of sign either side of the root
( ) () -4.089022992
x10
-5
() 8.621627124
x10
-5
Since there is a change of sign in my error bounds is a root of





Alexander White

10


This shows how Autograph found the root using Newton-Raphson, after

was imputed it found


by drawing a tangent to ()

at the point ( ()) and found

as the point where


this tangent crossed the -axis, it then used

to find

and so on until it found the root.


I will now find the other 2 roots of

to 5 decimal places using Autograph


In the interval [-1, 0] using





2.4 2.2 2 1.8 1.6 1.4 1.2 1
25
20
15
10
5
0
x
y
X1
Tangent to y=x5x2 at X1
X2
Tangent to y=x5x2 at X2
Root
Alexander White

11

The root in the interval [-1, 0] is to 5 decimal places
( ) () 1.207918099
x10
-5
() -3.598372658
x10
-5
Since there is a change of sign is a root of


In the interval [1,2] using



The root in the interval [1, 2] is to 5 decimal places
( ) () 2.429750662
x10
-4

() -2.023798041
x10
-5

Since there is a change of sign is a root of

















Alexander White

12



Failures of Newton-Raphson

This method will sometimes fail to find a root, for example the graph of

, this
shows a root in the interval [0,1]




If I attempt Newton-Raphson on this root starting with

I see that the result is divergent,


meaning that Newton-Raphson will not find the root



This is because is a turning point of the function

and therefore the


tangent to the curve at has a gradient of 0 and will never intersect the axis and this means it
will never give a value for



The iterative formula for this equation is



2 1 1 2
4
3
2
1
1
2
3
4
x
y
y=4x2x1
Alexander White

13

Showing this for

gives


()

()

()

()


= Error

This shows that it will fail as it means dividing by 0, and because this fails for

it means

will not
be found as there is no value to put into the iterative formula.

This shows that the tangent will never intersect the axis meaning that the Newton-Raphson
method will fail as a value for

will never be found.























1 0.5 0.5 1
3
2
1
1
x
y
Tangent at


Alexander White

14

Rearrangement method

The rearrangement method is another example of a fixed point iteration method, this works by
taking an equation () and then rearranging it into the form (), and then any value of
for which () will be a root of the original equation () .

The equation I will be solving using the rearrangement method is

to 5 decimal
Places.

Graph of ()




From the graph I can see that the roots lie in the intervals [-3,-2], [0, 1] and [1, 2]

Now I will rearrange () into the form () making the subject of my formula.

()



This means that the iterative formula for my equation is.



Now that I have this I will now plot and

on the same axis to find the coordinates


of the roots of

.
3 2 1 1 2 3
3
2
1
1
2
3
4
5
x
y
y=x4x+1
Alexander White

15


This graph shows that the points of intersection of and

have the same coordinate


of the roots of

, these arent the roots as the coordinate isnt 0, however we can


use this to find the roots.

I will now look for the root in the interval [0, 1] to 5 decimal places.

I will take my first estimate to be


()


= 0.5


()


=0.28125

I will now continue these calculations using Autograph.

3 2 1 1 2 3
3
2
1
1
2
3
x
y
y=x
y=(x+1)/4
Alexander White

16



From these calculations I can see that the root in the interval [0, 1] is to 5 decimal
places.

( ) () -1.2605
x10
-5

() 2.9558
x10
-5

Since there is a change of sign is a root of




This is a graph showing how the root was found using the rearrangement method; the staircase
diagram shows that at

a line is drawn from the coordinate (

)) to the point on
where (

) then the coordinate of this point is

and this process is then repeated using


to find

until the root is found.



The rearrangement method was able to find this root as the gradient of

satisfies the
condition
()

I can see this from looking at the original graph of

against as at the root

is
less steep than which has a gradient of 1.













0.2 0.4 0.6 0.8 1 1.2 1.4
0.2
0.4
0.6
0.8
1
1.2
1.4
x
y
y=x
y=(x+1)/4
X1
X2
X3
Alexander White

17

Failures of the Rearrangement method

Like all numerical methods there are some situations where this method will fail to find the root,
usually one rearrangement of () will not find all the roots


The graph shows me trying to find the root in the interval [1,2] using the rearrangement

starting with a value

when this value is put through the iterative formula a value of

is found, this is further away from the root.





This happens because the gradient of

at the point

is greater than 1, which is the


gradient of at that point leading to the failure. For a rearrangement () to find a root of
() it must satisfy the condition below.

()

From the original graph I can see this as at the root

is steeper than meaning its


gradient is as the gradient of is 1.




1 2 3 4
1
2
3
4
x
y
y=(x+1)/4
y=x
Root
X1
X2
Alexander White

18

To find the other roots I then must rearrange into a different form of ()

()


This then gives me a new iterative formula of




This show and

plotted on the same graph and that the value for () at the
point

satisfies the condition meaning that the root will be found.




()


= 0.3644

Now that I know that this satisfies the condition I can find the root, I will do the calculations using
Autograph.
3 2 1 1 2 3
3
2
1
1
2
3
x
y
y=x
y=(4x1)^(1/3)
Root
Alexander White

19



This shows that the root is to 5 decimal places.

( ) () 5.8429
x10
-5

() -5.4495
x10
-6

Since there is a change of sign is a root.






















Alexander White

20

Comparison of methods.

After successfully finding roots of equations using the 3 methods I will now compare them, to do this
I need to find the same root of an equation using all of the methods to the same accuracy to give a
fair comparison.

I will be using the equation

to compare the methods.




I will find the root in the interval [0,1] using all three methods.

I have already found the root using the change of sign method to be 0.68045 to 5 decimal places

Newton-Raphson

The iterative formula for this equation is



I will take





This shows that the root is to 5 decimal places.

2 1 1 2 3 4 5 6
20
15
10
5
x
y
y=x5x+2
Alexander White

21

( ) () -3.1467
x10
-5

() 2.2718
x10
-5

Since there is a change of sign is a root.

Rearrangement

I will rearrange () into the form




Graph showing and

on the same axis and the root as the point where they
intersect






This shows that the root is to 5 decimal places.
2 1 1 2
1
0.5
0.5
1
x
y
y=x
y=((x+2)/5)
Root
Alexander White

22


( ) () -3.1467
x10
-5

() 2.2718
x10
-5

Since there is a change of sign is a root.

Now that I have found the root using all 3 methods I can compare them.

Speed of convergence

To find the root using each method required a number of calculations.

Decimal Search required 62 calculations.

Newton-Raphson required 4 calculations.

Rearrangement required 9 calculations.

From this it is clear that decimal search took the most calculations to find the root, 58 more than
Newton-Raphson and 53 more than rearrangement, making this the method with the slowest speed
of convergence. Rearrangement was the 2
nd
fastest method requiring 5 more calculations than
Newton-Raphson but 53 less than decimal search. Newton-Raphson was the fastest method for
acquiring the roots only requiring 4 calculations to find the root to the same degree of accuracy as
the other methods looked at.

Method Difficulty.

All of the methods used could be done with a calculator and paper or using computer programs to
achieve the same results.

Decimal Search

With decimal search the first step with either equipment would be to find the interval that the root
lies in, using a calculator this would require setting up a table of integer values and looking for a
change or multiple changes of sign depending on how many roots the equation has. Whereas with a
program such as Autograph the graph will be drawn for you providing you with the intervals that the
root or roots lie in.

Once the intervals have been found increments will then be taken to find a smaller interval once
again looking for a change of sign. Using a calculator this would once again require a table of values
going up the interval in increments looking for another change of sign, this would require the input
of the equation and the increments into the calculator. With a computer a program could be used
such as Microsoft Excel or auto graph, this will also require the input of the equation and the
increments, using a program like excel could be difficult if not familiar and due to the way equations
are inputted could cause problems with the table of values.

Once a new interval has been found the process will be repeated once again using a table function
on a calculator or on a computer program, for each the only thing needed to change is the
increments as it requires the same equation. Now this will be repeated a number of times until the
Alexander White

23

root is found, this will require the same number of steps regardless of which equipment is being
used.

Newton Raphson

This will require the finding of the intervals to be done in the same way a decimal search first, either
by a table of values or a program like Autograph.

Once this is done an iterative formula is needed, a program like Autograph will do this automatically.
However if done using a calculator the function requires differentiation which must be done by
hand. If done by hand then each new estimate must be calculated individually; however the answer
button can be used to speed this up if

programed to be the answer on the calculator, and then


set up the formula so instead of the next value the ANS symbol visible just pressing equals key will
give the next estimate without needing to re-input the equation. Whereas using Autograph all once
the Newton-Raphson iteration button is pressed all thats needed is a first estimate and all the
calculations including finding () will be done by the program, all thats required is the pressing of
the equals key on Autograph. Both pieces of equipment will use the same number of estimates to
find the root.

Rearrangement

Once the intervals that the roots lie in have been found using the same process as the other
methods () must be rearranged into the form () neither a calculator or computer
programs will do this so it must be done by hand.

This also gives you your iterative formula if being done using a calculator, and will also tell Autograph
what the iterative formula is. When using a calculator if its set up using the ANS key like with
Newton-Raphson it will then just require pressing the equals key until the estimates round to the
same value at a given accuracy. When using auto graph the graphs of and ()must be
plotted on the same axis, and when theyre both selected the () iteration button must be
pressed, and a first estimate imputed and the equals key pressed until the estimates round to the
same value. Both pieces of equipment will require the same number of steps to find the root.

Conclusion

After using all the methods and comparing them, I think that when using a computer program the
simplest method is Newton-Raphson due to the lack of calculations needed by hand and the speed
the root is found. Whereas using a calculator I would use the rearrangement method as the first step
would be the same irrespective of the equipment and a calculator is quicker at finding the roots as
no additional graphs are needed. The decimal search method is the simplest to understand out of
the 3 I used but it also takes the most calculations and longest time to complete, no matter which
equipment is used.









Alexander White

24
































-Find one of the roots already found with one method using the other 2
-Compare the speed of convergence for each method
-Compare ease/difficulty of each method if it was being done using a calculator or a computer

PLAN OUT

Using

for Newton-Raphson

Using and

for rearrangement with



SHOULD TAKE UP TO

with both

You might also like