User manual | MATHEMATICAL INSTITUTE UNIVERSITY OF OXFORD

MATHEMATICAL INSTITUTE
UNIVERSITY OF OXFORD
Computational Mathematics
Students’ Guide
Michaelmas Term 2016
by
Dr Andrew Thompson
Acknowledgments: This course guide is based largely on previous versions by Dr Colin
Macdonald and others before him. I am grateful to all previous authors.
©2016 Mathematical Institute, University of Oxford
Notes for the 2016/2017 Course
The course will be in two parts: Part I in Michaelmas Term Weeks 3–8 and Hilary Term
Weeks 1–2, and Part II in Hilary Term Weeks 3–8. For Part I, you will attend scheduled
practical sessions every fortnight starting in Week 3 Michaelmas Term. The practical sessions
are held in the Mathematical Institute, Radcliffe Observatory Quarter.
Each practical session will be run by a demonstrator and may include short lectures. You
will need to bring your laptop to these sessions. If you are unable to bring a laptop
to the sessions, please contact Nia Roderick, Academic Assistant ([email protected])
well ahead of your first session.
Please follow the instructions on the Mathematical Institute website
https://www.maths.ox.ac.uk/members/it/software-personal-machines/matlab
to install Matlab on your machine prior to the first session. Note that there are around
50 toolboxes available for download, and the only one you will require is the Symbolic Math
Toolbox. The default setting is to download all available toolboxes, which may take many
hours using a standard broadband connection. You can select just the Symbolic Math Toolbox
(i.e. deselect all other toolboxes) in the “Product Selection Window” stage of the installation
process.
Alternatively, and in particularly if you are a proficient computer user, you could try
to use Octave (http://www.octave.org) which uses a very similar command syntax and is
often regarded as a Matlab clone. See Section 1.1.2.
Time-line
A suggested rough schedule for covering the material is the following table:
Week
3–4 MT
5–6 MT
7–8 MT
1–2 HT
Chapter Covered
1
2
3
4
You should work through each chapter in order, progressing at a rate of roughly one chapter
a fortnight. Michaelmas term will place many demands on your time but you need to put in
the effort to ensure that you make steady progress and that you finish Part I of the course
by the end of Week 2, Hilary Term.
Assessment During Part I of the course, you may work collaboratively with others and—as
always—you are encouraged to discuss mathematics and your studies with your peers. Your
demonstrator may ask you to submit work related to these sessions but none of the work
in Part I will be formally assessed. Instead the material will act as a foundation enabling
you to work individually during Part II. This individual work will be assessed and will count
towards your Preliminary Examination as described in Examination Decrees & Regulations,
2016 and the current Course Handbook. (See your college tutor if you have any questions
about this aspect of the course.)
Course director The course director for this academic year is Dr Andrew Thompson.
The course demonstrators (who will be in attendance during each practical session) will try
to answer any of your questions; alternatively Dr Thompson can be contacted by email at
[email protected]
Course website This manual and any extra course material (such as lecture notes) can
be found on the course website:
https://courses.maths.ox.ac.uk/node/30
Errata list In the rather likely event that errors are found in this manual after printing,
they will be corrected and listed on the website above. It would be a good idea to check
occasionally, especially if you think you’ve found an error.
Contents
1 Introduction to Matlab
1.1 What is Matlab? . . . . . . . . . . . . . . . .
1.1.1 What is the Symbolic Math Toolbox?
1.1.2 What is Octave? . . . . . . . . . . . .
1.1.3 Learning Matlab . . . . . . . . . . . .
1.2 The command prompt, variables, getting help
1.2.1 Using Matlab’s built-in help system
1.2.2 Assigning to variables . . . . . . . . .
1.2.3 Floating-point numbers . . . . . . . .
1.2.4 Symbolic computing . . . . . . . . . .
1.2.5 Saving your work in .m file scripts . .
1.2.6 Using symbols instead of numbers . .
1.2.7 Built-in functions and help . . . . . .
1.3 2D graphics . . . . . . . . . . . . . . . . . . .
1.3.1 Simple line graphs . . . . . . . . . . .
1.3.2 Parametric and polar plots . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
1
1
2
2
2
3
4
4
5
6
9
10
10
11
2 Algebraic equations and calculus
2.1 Evaluating expressions . . . . . . . . . . . . . . . . .
2.2 The solve command . . . . . . . . . . . . . . . . . .
2.2.1 Solving equations symbolically . . . . . . . .
2.2.2 Assumptions . . . . . . . . . . . . . . . . . .
2.2.3 Solving equations numerically . . . . . . . . .
2.3 Differentiation and the diff command . . . . . . . .
2.3.1 Differentiation of unknown functions . . . . .
2.4 Evaluating limits . . . . . . . . . . . . . . . . . . . .
2.5 Integration . . . . . . . . . . . . . . . . . . . . . . .
2.5.1 The int operator . . . . . . . . . . . . . . . .
2.5.2 Quadrature: numerical evaluation of integrals
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
13
14
14
16
16
18
20
22
23
23
24
.
.
.
.
.
.
.
.
26
26
28
29
29
30
30
31
31
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 Differential equations, sums, matrices and vectors
3.1 Solving ordinary differential equations using dsolve
3.2 Vectors and lists in Matlab . . . . . . . . . . . . . .
3.2.1 Symbolic lists . . . . . . . . . . . . . . . . . .
3.2.2 Indexing in lists . . . . . . . . . . . . . . . .
3.2.3 Vector operations . . . . . . . . . . . . . . . .
3.3 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Sums and products . . . . . . . . . . . . . . . . . . .
3.4.1 Simple “numerical” sums . . . . . . . . . . .
3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
3.5
3.6
3.4.2 Products . . . . . . . . . . . . . . . . . . . .
3.4.3 Symbolic manipulation of sums and products
Arrays, matrices and vectors . . . . . . . . . . . . .
3.5.1 Matrix definition . . . . . . . . . . . . . . . .
3.5.2 Rows, columns and submatrices . . . . . . . .
3.5.3 Operators on matrices and vectors . . . . . .
3.5.4 Simultaneous equations . . . . . . . . . . . .
3.5.5 Eigenvalues and eigenvectors . . . . . . . . .
Meshgrids and 3D plots . . . . . . . . . . . . . . . .
3.6.1 Logical masks . . . . . . . . . . . . . . . . . .
3.6.2 Meshgrid . . . . . . . . . . . . . . . . . . . .
3.6.3 3D plotting . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
32
33
34
34
34
36
36
36
36
37
37
4 Loops, conditionals and functions
4.1 Loops . . . . . . . . . . . . . . . . . . . . .
4.1.1 Approximate solutions to equations
4.2 Conditionals . . . . . . . . . . . . . . . . . .
4.3 Functions . . . . . . . . . . . . . . . . . . .
4.4 Examples of functions . . . . . . . . . . . .
4.4.1 Finding the arithmetic mean of a set
4.4.2 Taylor’s theorem . . . . . . . . . . .
4.4.3 Euler’s method . . . . . . . . . . . .
4.4.4 Euclid’s algorithm . . . . . . . . . .
4.4.5 A simple matrix function . . . . . .
4.5 Debugging . . . . . . . . . . . . . . . . . . .
4.6 Further exercise . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
of numbers
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
39
39
40
41
42
44
44
44
45
46
46
47
48
A Simplification
A.1 Expand . . . . . . .
A.2 Factor . . . . . . . .
A.3 Collect . . . . . . . .
A.4 Simple and Simplify
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
49
49
50
50
50
B Other advanced topics
B.1 Matlab “handle” functions and anonymous functions
B.2 Variable precision arithmetic . . . . . . . . . . . . .
B.3 Cell Arrays . . . . . . . . . . . . . . . . . . . . . . .
B.4 MuPAD worksheets . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
52
52
52
53
53
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Chapter 1
Introduction to Matlab
In this chapter you will learn about:
• Accessing Matlab;
• using Matlab as a calculator;
• Matlab identifiers, constants and functions;
• using the Matlab help system;
• making .m file scripts;
• plotting simple line graphs.
1.1
What is Matlab?
Matlab is often described as a problem solving environment. It is a programming language
and set of tools for solving mathematical problems.
The name Matlab was originally a contraction of “Matrix Laboratory” and indeed Matlab’s core strength is numerical computing involving matrices, vectors and linear algebra. It
also has extensive plotting and graphics routines.
1.1.1
What is the Symbolic Math Toolbox?
The Symbolic Math Toolbox is an add-on for Matlab which adds symbolic computing
abilities. The strength of such a system is the ability to manipulate algebraic expressions.
The Symbolic Math Toolbox can be used to solve algebraic equations, factorize polynomials
and simplify rational expressions. It can also solve some differential equations. It can perform
many of the standard operations of analysis such as evaluating sums, limits, derivatives and
integrals. Under the hood, the Symbolic Math Toolbox is powered by a computer algebra
system called MuPAD.
The combination of Matlab and the Symbolic Math Toolbox allows the user to mix
symbolic and numerical computing within the Matlab environment.
1.1.2
What is Octave?
Some people are concerned about the ethical issue surrounding computing. Should software
be freely available? If these issues are important to you, you might want to consider Octave
(http://www.octave.org) as an alternative to Matlab. It is Free Software (also known as
1
CHAPTER 1. INTRODUCTION TO MATLAB
2
Open Source Software) which means you can use it without restriction, study its source code,
improve it, and share it with others as you wish. None of these things are true of Matlab.
OctSymPy (see http://github.com/cbm755/octsympy) is an add-on package that adds
symbolic computing features to Octave (it is analogous to the Symbolic Math Toolbox for
Matlab). OctSymPy is new software developed right here at Oxford!
The syntax of commands in Octave is almost the same as Matlab and the differences
are well-documented. But having said that, this manual will concentrate on Matlab and
this is likely the tool your demonstrators will know best. Try Octave if you wish, but do so
“at your own risk.”
1.1.3
Learning Matlab
In order to become technically adept at using Matlab it is important that you attempt
all the exercises contained in this manual; as with most tools, Matlab is best learned by
actually using it to do mathematics, and this should be practised as often as possible. Try
to incorporate Matlab into your weekly problem sheets, by using it to check some of your
hand-written answers. Examples include substituting your solution back into the equation
and plotting an answer to make sure it makes sense.
In the rest of this chapter, we will give a brief introduction to Matlab, outlining some
of its basic features and illustrating them with some short examples. The output of each
command line is not printed in this manual; after executing each line you should check that
the output is as you would expect. If it is not, then you should think about why this might
be, and if necessary consult your demonstrator.
1.2
The command prompt, variables, getting help
Note that in Matlab the prompt is the symbol “ ”. Commands are entered to the right
of the prompt, followed by the return/enter key. To illustrate this try entering the following:
12/15
which should display 0.8. Try some more commands
3 + 18
3+18
3
+ 18
40*21
factorial(3)
1.2.1
Using Matlab’s built-in help system
There are various ways of getting help on Matlab commands but one of them is built into
the command prompt and is a good “first stop” for help. To use it, type help followed by
the name of a command:
help factorial
You can also access help in various ways using the graphical user interface. (Don’t worry if
much of the information in the help text doesn’t make sense yet.) Note that this help tends
to refer to functions in CAPITALS but Matlab is case sensitive and usually the actual
function is lowercase. Presumably this is for historical reasons, left over from a bygone era
CHAPTER 1. INTRODUCTION TO MATLAB
3
before copy-and-paste1 . Throughout this manual you should use the help facility to learn
more about the commands you are introduced to.
1.2.2
Assigning to variables
Numbers can be assigned to variables
a = 4
b = factorial(4)
B = 5
Later you can recall the value of a variable
a
b
and variables can be used in further calculations
my_var = a^2 + b/10 + 2*B
It is important to think about what the symbol = means here: Matlab uses it as the
assignment operator. B = 5 means “assign the value 5 to B”. You could pronounce this as
“B gets 5”. You will sometimes see people writing the assignment operator as “B ← 5” and
some other programming languages use “B := 5” to avoid confusion. For better or for worse,
Matlab uses =. Later on, we will meet the equality operator == which tests for equality and
expresses symbolic equality: it deserves to be pronounced “equals”.
There are a few words that are reserved for Matlab operations and so cannot be used
for variable names. Examples include while, for, and function: we will encounter many of
these later the course and if you try to use one as variable name you’ll usually get an error.
But just because something is allowed doesn’t mean it’s a good idea:
pi = 3.2
sin(pi)
This seems like a good time to show you how to clear all variables.
clear
pi
a
% back to its usual value
% gives an error
Indeed, it is recommended that you start all exercises with clear. Let’s say that again in
bold:
Usually, you should start exercises with clear.
When something doesn’t work as expected, try clear.
A semicolon suppresses the output (this is useful, for example, if the output is extensive
and not specifically required). It can also make your code and output more readable: this
will be more important as you start writing longer scripts in Matlab.
a
b
c
c
1
=
=
=
=
6;
3;
a*b;
2*(c + 6)
“We didn’t have copy-and-paste in my day. I used to dream about copy-and-paste...”
CHAPTER 1. INTRODUCTION TO MATLAB
1.2.3
4
Floating-point numbers
By default Matlab produces numerical answers which are often approximations (albeit
highly accurate approximations)
5/3
pi
sqrt(2)
These numerical solutions are typically accurate to about 15 decimal places:
sin(0)
sin(pi)
Note the latter gives 1.2246e-16, that is 1.222 × 10−16 . You can increase the number of
displayed digits with
format long
sin(pi)
but note this has no effect on the actual accuracy of the expression, just its display.
Internally, Matlab by default stores numbers in “IEEE Double-Precision Floating Point”
or just double for short. Each double takes 8 bytes (that is 64 bits, “binary digits”) in the
memory of the computer. The 64 bits is used to store numbers in scientific notation with a
certain number of bits for the exponent (the “-16” in our example above) and the rest for
the coefficient (1.222 in our example).
Computers are very fast at doing arithmetic on doubles and this forms the basis for much
of computation. For example, encoding/decoding your voice, music, and video on your phone,
performing large-scale climate simulations, finding a location from GPS satellites, or finding
eigenvalues of large matrices all involve billions of operators on doubles. A surprising number
of things we do (e.g., searching the internet) are just special cases of “finding eigenvalues of
large matrices”.
However, sometimes it’s nice when 6/9 becomes 2/3 not “0.666666666666667”. Particularly when learning or studying mathematics, we are often less interested in the numerical
solution of something but rather in the how and why of it. Think about the area under a curve
(a number, rather easily approximated) versus the indefinite integral (a general expression).
We want the computer to do both these tasks.
1.2.4
Symbolic computing
Symbolic manipulation software or computer algebra systems do work with symbols rather
than numbers. They try to work out derivatives and integrals and solve equations much the
same way you do: by manipulating the symbols according to certain rules (“6/9: 6 is 2*3
and 9 is 3*3 so cancel the 3’s to get 2/3”).
The Symbolic Math Toolbox adds symbolic computing to Matlab. We can check if we
have the toolbox with the command:
ver
which might produce a lot of output but the important bit for us is
Symbolic Math Toolbox
Version 6.2
(R2015a)
CHAPTER 1. INTRODUCTION TO MATLAB
5
This manual assumes that you are using Version 6.2 or any subsequent later versions of
the Symbolic Math Toolbox. If you installed Matlab according to the instructions at the
beginning of this manual, this will certainly be the case.
With the toolbox, you enter symbolic expressions at the Matlab prompt:
sym('6/9')
sym('1')
a = sym('pi')
sin(a)
b = sym('sqrt(2)')
b^2
Those are single quotes which surround a string. Notice that you can also assign symbolic
expressions to variables.
In fact, you can leave out the quotes when entering small integers or simple fractions.
sym(1)
sym(1000)
sym(6/9)
It is recommended to use this shortcut only for integers and very simply fractions, and to
revert to the quote form if there is any doubt. The following exercise should convince you
there can indeed be doubt.
Exercise 1.1 Observe the output of sym(13/10), sym(133/100), sym(1333/1000) and follow the pattern a few more steps. What happens? Try the experiment again with quotes
(sym('13/10'), etc). (If you’re curious, help sym explains how the quoteless mechanism
works).
In general, you probably don’t want to use decimal places inside the sym command.
sym('6/9')
sym(6/9)
sym('6.0/9')
% yes
% sure, good too
% probably not what you wanted
The latter will use “variable precision arithmetic” which we won’t make much use of in this
course: “help vpa” if you want to know more, or see Appendix B.2.
1.2.5
Saving your work in .m file scripts
Instead of just typing commands into the command prompt, you can create a script. Within
the graphical user interface, one way to do this is to select “file, new, script” from the menu
bar. Add some commands to the file, for example:
% my solution to Exercise 1.1
sym(13/10)
sym(133/100)
Save the file as “ex1_1.m”. From within the editor, you can run your script by clicking on
the icon with the green triangle and the white square. You can also switch to the command
prompt and type the name of your script (without the .m)
ex1_1
CHAPTER 1. INTRODUCTION TO MATLAB
6
In either case, the contents of the script should execute in the command window. You can
alternate between editing your script and running it.
It is recommended to save each exercise as a script and you may want to use additional
scripts as you work through this manual. Find what works for you but remember you
may want to access this material later (for example, next term when you work through
the projects).
1.2.6
Using symbols instead of numbers
Numbers are nice and all, but the sym() command makes more interesting things possible
because symbolic expressions can contain symbols:
x = sym('x')
y = asin(x^3)
z = sin(y)
Another example:
x = sym('x')
y = sym('y')
f = (x^3 + x^2) * sqrt(y) * exp(x)
In fact, there is a shortcut—syms—for defined symbolic variables:
syms x y z a b c r s t
f = sin(a*x) * sqrt(tan(r*t-x) + sin(t-1) * tanh(s^2*y))
Consider
a = sym(6)
b = sym('2*a')
c = sym(2)*a
What happened? The result in b might not be what you wanted! This suggests you should
avoid putting expressions inside the quotes and instead build them out of symbolic variables or
symbolic numbers.2 For example, rather than writing f = sym('a * x + b'), it’s probably
better to do
syms a x b
f = a*x + b
In Section 2.1, we will learn about the subs command which will help push assigned variables
into expressions.
Variable types and casting
We have encountered two types (“classes”) of objects so far: doubles which store numbers
with a finite precision and syms which store symbolic expressions. It will often be important
to keep track of which is which. The whos command can help.
clear
a = sym('6/9')
b = 6/9
whos
2
In previous versions of the Symbolic Math Toolbox, it was quite common to put a lot of mathematics
inside quotes in sym() commands, so you may see references to this sort of thing.
CHAPTER 1. INTRODUCTION TO MATLAB
7
Now we can ask some useful questions about what happens when you combine objects.
Exercise 1.2 Use Matlab to evaluate the following exactly:
a = sym(2/3)
b = 5/3
c = a + b
d = b*a
e = a^b
f = (2/3) ^ (sym(5/3))
e - f
whos
From these results, what does Matlab typically do when it performs a binary operation
which combines a double and a sym? This is known as casting.
So very roughly speaking, as long as some symbols in your expression are of class sym, you
can expect the result to be symbolic too. This is good news for entering complex expressions,
for example:
x
f
%
g
= sym('x');
= 512*x^4 + 13/10*x^2 + 256*x^(2/3)
because the alternative is just horrible:
= sym(512)*x^(sym(4)) + sym(13)/sym(10)*x^sym(2)
sym(256)*x^(sym(2)/sym(3))
f - g
% but they are the same in the end
+
...
The previous example also shows how to split a very long line in two: simply interrupt the
line with “...” (three full stops) and continue on the next. You’ll also note that we have
been using the character “%” to comment our code. This is a good idea to help the reader
understand what the code does. As you’ve probably figured out, Matlab ignores these
comments when executing the code.
Sometimes it’s hard to tell if an expression like f above was entered correctly; the output
often looks as bad—or worse—than the input. Try this:
pretty(f)
That will render the expression in way that you might find easier to decipher.3 Similarly,
latex(f) will output code that can be pasted into LATEX, which is the standard tool most
mathematicians use for typesetting mathematics.4
Just as sym() converts to symbolic expressions, double() converts to double-precision
numerical values.
Exercise 1.3 Use Matlab to evaluate the following exactly:
19 × 99, 320 , 21000 , 25!,
3
26
− .
13 27
Now calculate the decimal value of following:
21
1
, 171/4 ,
, 0.216100 .
23
99!
Finally, explain what Matlab does in each of these cases:
3
Of course, beauty is subjective. . . The output of this command might have counted as “pretty” in 1992;
perhaps “retro” might be a better word.
4
Typesetting in LATEX is a useful skill to learn, especially if you end up continuing for graduate studies.
CHAPTER 1. INTRODUCTION TO MATLAB
8
a = 2/3
b = sym(2)/sym(3)
c = double(b)
Matlab has some names reserved for constants, some of which are shown in the following
table
pi
i (or “1i”)
inf
nan (Not a Number)
3.1415
...
√
−1
∞
0
0 , sin(∞)
The number e, that is 2.7182 . . ., can be entered as
double_e = exp(1)
symbolic_e = exp(sym(1))
log(symbolic_e)
Exercise 1.4 Careful it’s not sym(exp(1)): try that and see what happens. Can you explain
why?
As stated above, Matlab denotes the square root of −1 by i (actually 1i or j work too).
Suppose that we wish to express (2 + 3i)4 in the complex form a + bi, where a and b are real:
(2 + 3*i)^4
Some Matlab programmers don’t like this because they want to use i for a variable. Thus
it is very common to instead write:
(2 + 3*1i)^4
% or
(2 + 3i)^4
These examples used class double but symbolic complex numbers work too:
syms x y
z = x + 1i*y
sin(z)
You might have a sense of unease at this point: you and I might be tacitly assuming x and y
are real but of course Matlab doesn’t know that. Indeed the complex conjugate of z = x+iy
is z̄ = x − iy but Matlab says:
syms x y
z = x + 1i*y
conj(z)
We will see how to fix this later using assumptions in Section 2.2.2.
Exercise 1.5 Use Matlab to express the following in the approximate decimal form a + bi
1
1+i
4 2i 4
2
−1
2
(5 − 8i) , i , i , i 2 ,
, and
+
.
5 + 2i
3
5
CHAPTER 1. INTRODUCTION TO MATLAB
9
Exercise 1.6 Use symbolic complex numbers to verify that (a + bi)2 = a2 − b2 + 2abi. (Hint:
one approach would be to construct the left-hand side and the right-hand side. Sometimes
the simplify() command will give Matlab a nudge. . . ).
1.2.7
Built-in functions and help
As we have seen already, Matlab has many in-built functions, many of which are fairly
obvious. The table below lists some of those that you should be familiar with.
cos, cosh
sin, sinh
tan, tanh
sec, sech
csc, csch
cot, coth
exp
log
log10
abs
sqrt
cosine and hyperbolic cosine
sine and hyperbolic sine
tangent and hyperbolic tangent
secant and hyperbolic secant
cosecant and hyperbolic cosecant
cotangent and hyperbolic cotangent
exponential
natural logarithm
base-10 logarithm
absolute value (and magnitude of complex number)
square root
Matlab has extensive built-in help which you can access from the command prompt:
help cos
One important note is that help often gives you the function that operates on double
values. For example, with the command above, you’ll probably see
Overloaded methods:
sym/cos
If you are working with symbolic expressions, you can access the symbolic specific help:
help sym/cos
(An object-oriented programmer would say “cos is overloaded for sym input”.) Other commands have associated help too:
help clear
help syms
help whos
You can also search for keywords:
lookfor hyperbolic
Matlab has more detailed hyperlinked help accessed with the doc command (and also available online).
Exercise 1.7 Use Matlab to evaluate the following:
tan2 (π/3), sec(π/6), 1 + cot2 (π/4), cosec2 (π/4) e3 ln 4 .
CHAPTER 1. INTRODUCTION TO MATLAB
10
Exercise 1.8 Use the help facility to learn how to compute the binomial coefficient n Cr in
Matlab. Evaluate 10 C4 and 250 C12 . Do this both symbolically and numerically with double
precision.
Exercise 1.9 Evaluate arcsin(1/2). Evaluate sec(arctan x) in terms of x.
1.3
2D graphics
One of the very useful features of Matlab is the ease with which many different types of
graphs may be drawn. Most graphs are drawn with variants of the plot command, which
has several forms, each with many optional arguments that determine the appearance of the
resulting graph. Here we introduce a few of the simplest forms. If at any time you wish to
learn more, explore with Matlab’s help and doc pages.
1.3.1
Simple line graphs
The simplest graphs are those of functions of a single real variable; for example, the graph of
the function sin(10/(1 + x2 )) for 0 ≤ x ≤ 10 is given by the command
syms x
y = sin(10/(1+x^2));
ezplot(y)
You should see the graph on your screen, and it should be blue, which is the default colour.
You can control the domain with
ezplot(y, [-5, 10])
The range of the vertical axis can also be specified:
ezplot(y, [-5 10])
ylim([-1.2 1.2])
Exercise 1.10 Plot sin(10 cos x) on the interval −2π ≤ x ≤ 2π.
You can draw more than one curve, in different colours and line styles. For example:5
clear
syms x
ezplot(cos(x))
hold on
h = ezplot(sin(x));
set(h, 'color', 'red')
set(h, 'linestyle', '--')
set(h, 'linewidth', 3)
legend('cos', 'sin')
title('my colourful plots')
5
You will probably want to put longer examples like this in a script (see Section 1.2.5).
CHAPTER 1. INTRODUCTION TO MATLAB
11
The use of “hold on” is crucial: try it without. The American spelling of “colour” is required
(as is the American pronunciation of “EZ”). The variable h is called a “handle” and it gives
you access to Matlab “under the hood”: see get(h) for example.
You can also clear a figure with clf and control which figure window you’re drawing on:
figure(2); clf;
h = ezplot(atan(x));
set(h, 'linewidth', 4)
set(h, 'color', [0.5 0 1])
The colour was set using an “RGB triplet” a vector specifying the percentages of red, green
and blue.
Exercise 1.11 Draw the graphs of these five functions on the same plot:
y1 (x) = 1,
y2 (x) = 1 + x,
y3 (x) = 1 + x + x2 /2,
y4 (x) = 1 + x + x2 /2 + x3 /6,
and y5 (x) = exp(x), on the interval 0 ≤ x ≤ 1. Use different colours for each. Add a legend.
You can interact with figures using the graphical user interface. For example, zooming
and panning with the mouse (try this).
Exercise 1.12 Using a plot, find the approximate coordinates of all the real solutions of the
nonlinear simultaneous equations
y = sin x,
y = x3 − 5x2 + 4.
Hint: see help ginput for one interesting way to do this with a mouse. (You will see in
Section 2.2.3 how to find numerical solutions to simultaneous equations such as these.) 1.3.2
Parametric and polar plots
The command ezplot can also plot a function that is defined parametrically (with x and y
as functions of t, say). So the ellipse x = 2 cos t, y = sin t is plotted with the command
syms t
ezplot(2*cos(t), sin(t), [0, 2*pi])
axis equal
The axis equal command will ensure that a circle looks like a circle.
Exercise 1.13 The cycloid x = t − sin t, y = 1 − cos t is the curve traced out by a point on
a wheel as the wheel turns. Plot this curve for 0 ≤ t ≤ 6π.
Polar plots of r = f (θ) for a ≤ θ ≤ b can be achieved with ezpolar, for example to draw
the cardioid:
syms t
ezpolar(1 - cos(t), [0, 2*pi])
Exercise 1.14 Plot (both leaves of) the lemniscate r2 = cos(2θ).
CHAPTER 1. INTRODUCTION TO MATLAB
12
Alternatives to the ezplot
The ezplot command and the other ez commands are useful to make quick plots, particularly
of symbolic expressions. However, sometimes it might be useful to access the other Matlab
plotting commands in order to have more control over plots. We give one (non-exhaustive)
example, as you will certainly encounter this style of Matlab plotting:
syms x
y = sin(10/(1+x^2));
ym = matlabFunction(y)
xx = linspace(-5, 10, 512);
plot(xx, ym(xx))
The key idea here is to create a vector of doubles (the linspace command). We also convert
the symbolic expression into a more traditional Matlab function.
Animation
We finish this chapter with an animation:
clear
clf
syms x y
ezsurf(real(atan(x + 1i*y)), 'circ')
shading flat
camlight left
material shiny
for a = 1:90
view([-30, a])
pause(0.1)
end
(More about for loops later.)
Chapter 2
Algebraic equations and calculus
In this chapter you will learn about:
• different ways of evaluating expressions;
• solving equations symbolically and numerically;
• using Matlab to differentiate expressions;
• using Matlab to evaluate integrals (symbolically and numerically);
• dealing with limits.
2.1
Evaluating expressions
We previously saw a small example of the behaviour of the Matlab Symbolic Math Toolbox
when redefining variables in an expression. Recall:
syms x
y = 2*x
x = sym(6)
y
% still 2x
The subs command is useful here.
subs(y)
By default subs substitutes the current value of all variables into the expression. In this case,
x was 6 so it substitutes 6 in for x to give y = 12.
The subs command has a more specific form subs(expr, old, new) which replaces all
occurrences of the expressions old in expr by the expression new. For example, suppose we
want to evaluate the expression x2 + 3x − 2 at x = 1. This is achieved by using the commands
syms x
y = x^2 + 3*x - 2;
subs(y, x, 1)
A useful feature is that y itself has not changed, so that if we now wish to evaluate y at a
different value of x then this is easily done:
13
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
14
% first, note y unchanged
y
subs(y, x, 10)
subs(y, x, sym(pi))
y2 = subs(y, x, atan(x))
Exercise 2.1 In each of the following, evaluate the expression at the given value.
i) x3 − 3x2 + 2x − 1, at x = 5;
ii) sin x cos3 x, at x = π/4;
√
iii) ln u + 1 + u2 − 3 , at u = 2.
A single call of subs can bepused to perform several substitutions. As an example, suppose
that we wish to evaluate r = x2 + y 2 + z 2 at the point (x, y, z) = (1, 2, 3). This is achieved
by typing
syms x y z
r = sqrt(x^2 + y^2 + z^2)
A = subs(r, [x y z], [1 2 3])
The [] notation indicates a vector: more on this later. Recall that you can use double() to
evaluate to a decimal approximation:
double(A)
Sometimes after using subs you might need simplify to encourage Matlab to cleanup the
result.
That completes this brief introduction to evaluation. We now move on to look at solving
equations. The main aim is to introduce solve(), which can be used to solve equations,
inequalities, and systems of these. First, recall that “=” is used in Matlab to express
assignment. To express equality we use “==”.
a = sym(6)
logical(a == 6)
logical(a == 7)
logical(a < 10)
logical(a <= 6)
logical(a ~= 5)
logical(a ~= 6)
%
%
%
%
%
%
%
assign 6 to a
returns 1 for
returns 0 for
returns 1 for
returns 1 for
returns 1 for
returns 0 for
true
false
true
true
true
false
These are mostly self-explanatory except for “~=” (tilde followed by equals) which means
“not equal to”.
2.2
2.2.1
The solve command
Solving equations symbolically
The solve command can be used to rearrange simple algebraic expressions to arrive at a new
expression. For example, the solution of the equation
2x + 3 = 0
for x can be obtained using the Symbolic Math Toolbox as follows:
(2.1)
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
15
syms x
eq = 2*x + 3 == 0
solve(eq, x)
Again please note the different uses of = and == here: eq is assigned the expression of
Equation (2.1) which happens to contain an equality operator. Messing this up makes subtle
bugs that can be hard to track down.
Note that the solution, − 32 , is not assigned to x. You can do that yourself if you wish:
x
x = solve(eq, x)
eq
eq2 = subs(eq)
Suppose that we wish to solve the simultaneous equations
x + y = 2,
−x + 3y = 3.
This is done as follows:
clear
syms x y
eq1 = x+y == 2;
eq2 = -x+3*y == 3;
[xsoln, ysoln] = solve(eq1, eq2, x, y)
An alternative:
sol = solve(eq1, eq2, x, y)
xsoln = sol.x
ysoln = sol.y
where sol is something called a structure: roughly speaking it can contain “sub-variables”
known as fields.
What about problems with multiple solutions?
clear
syms x
eqn = x^2 == -25
sol = solve(eqn, x)
sol(1)
sol(2)
Here sol is a vector of two solutions, −5i and 5i.
Exercise 2.2 Find the solutions of the following equations:
(a)
x2 − x − 2025 = 0,
(b) x3 − 6x2 − 19x + 24 = 0,
(c) 2x4 − 11x3 − 20x2 + 113x + 60 = 0.
Exercise 2.3 Use the solve command to find the point of intersection, in the (x, y)-plane,
of the two lines
ax + by = A, cx + dy = B.
Having found the intersection, find an expression for the distance of the point of intersection
from the origin.
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
16
Note in the previous exercise that the Symbolic Math Toolbox made various assumptions
(for example that ad − bc 6= 0). Some other systems (for example MuPAD in its worksheet
mode) are very good at finding a wider variety of solution possibilities: this can be useful
(and/or annoying) depending on what you’re trying to accomplish.
2.2.2
Assumptions
Sometimes one makes assumptions about equations and variables. For example, you might
be interested only in the real roots of a quadratic. Or you might have a mathematical model
where you know a particular variable must stay positive (say it represents a positive physical
quantity like concentration or density). The Symbolic Math Toolbox allows you to add
assumptions to symbolic variables.
syms x
assume(x, 'real')
ineq = x^2 <= 25
S = solve(ineq, x, 'ReturnConditions', true);
S.conditions
assumeAlso(x > 0)
S = solve(ineq, x, 'ReturnConditions', true);
S.conditions
ineq = x^2 <= -25
S = solve(ineq, x, 'ReturnConditions', true);
S.conditions
To list the current assumptions, type:
assumptions
You can remove the assumptions on a particular variable using
x = sym('x', 'clear')
Recall that if you want to clear everything and start again, the usual approach is the clear
command. But it is worth mentioning that this will not remove assumptions from variables.1
The clear all command will do the right thing.
Exercise 2.4 Define z = x + iy where x and y are symbolic variables. Take the complex
conjugate of z. Now modify your code to assume that both x and y are real and run your
code again.
2.2.3
Solving equations numerically
The solve command is used for finding symbolic solutions to equations. This is not always
possible, but numerical (approximate) solutions can usually be found. If solve is unable to
find a symbolic solution, it will try to find a numerical solution.
For example, suppose we wish to solve the equation
sin x = x3 − 5x2 + 4,
and that having plotted the graphs of sin x and x3 − 5x2 + 4 we know that there are three
solutions, at approximately x = −0.90, 0.89 and 4.78 (these approximate solutions were
obtained in Exercise 1.12).
To find these solution more precisely, we try solve:
1
The variable will be deleted but not the assumption, possibly leading to very subtle and hard-to-find bugs.
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
17
clear
syms x
eqn = sin(x) == x^3 - 5*x^2 + 4;
solve(eqn, x)
This fails to find an exact symbolic solution but it finds a numerical solution
ans =
0.88543649189409090795806224578236
This solution is neither a double nor a sym class: we’ll see in a moment why it is so accurate.
But if you’re going to do something else with the result, you probably want:
y = double(solve(eqn, x))
which converts it to a double quantity.
What about the other two solutions? The most common numerical approaches involve
providing either an interval in which to search or an initial guess. Here we outline two possible
approaches.
The vpasolve command The Symbolic Math Toolbox also has the command vpasolve
which uses “variable-precision arithmetic” (see Appendix B.2 for more information.) One
way to use this command is to provide a search interval:
vpasolve(eqn, x, [-1 0])
vpasolve(eqn, x, [0 1])
vpasolve(eqn, x, [4 5])
ans =
-0.90040020886787653209849187387426
ans =
0.88543649189409090795806224578236
ans =
4.7813983927628242128187791632837
Alternatively, you can provide an initial guess:
vpasolve(eqn, x, -.9)
vpasolve(eqn, x, .8)
y = double(vpasolve(eqn, x, 4.7))
(where the last example shows conversion to double, assuming the solution is to be used in
further calculations.)
Note that the vpasolve command does not respect assumptions on variables.
Alternative approach: fzero for numerical root finding The Matlab command
fzero() does root-finding: it searches for numerical solutions of f (x) = 0 where f is a
function of one variable.
f = @(x) sin(x) - ( x^3 - 5*x^2 + 4 );
format long
fzero(f, -1)
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
18
ans =
-0.900400208867877
Some explanation: the first line creates a Matlab “anonymous function” (see Appendix B.1),
which is then passed to fzero, along with an initial guess. We can then find the third root:
fzero(f, 5)
ans =
4.781398392762824
Note that fzero does not work with symbolic expressions or symbolic functions.
Having to enter the same expression again is not optimal and potentially error-prone.
Here’s one approach to convert the symbolic equation “eqn” from above into an appropriate
function f:
% we want the left-hand-side and right-hand-side
LR = children(eqn)
f = matlabFunction(LR(1) - LR(2))
fzero(f, -1)
ans =
-0.900400208867877
Exercise 2.5 Makes a plot showing the roots of x3 + 3x2 − 2x + 1. Then use solve() to
find the roots. Use double() to find numerical approximations to the roots.
Exercise 2.6 Find real solutions of the equation
x sin x =
1
2
for x in the following ranges: (i) 0 < x < 2; (ii) 2 < x < 4; (iii) 6 < x < 7; (iv) 8 < x < 10.
Verify graphically that there are no solutions in the range 4 < x < 6.
2.3
Differentiation and the diff command
The Symbolic Math Toolbox can find the derivatives of expressions of one or several variables.
In this section we introduce the diff command, which is the basic differentiation operator
for expressions, and illustrate its use in different situations including finding the coordinates
of stationary points.
We start with an example; the derivative of sin x with respect to x is found with the
command
syms x
diff(sin(x), x)
The general syntax for differentiating an expression expr with respect to a variable var is
diff(expr, var);
and the derivative of ex
2 +a2
with respect to x is found with the commands
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
19
syms x a
z = exp(x^2 + a^2);
d1 = diff(z, x)
The second derivative of ex
2 +a2
with respect to x is
diff(d1, x)
and this same result can also be found all in one go by typing
diff(z, x, 2)
So the fifth derivative of sin(x2 ) is achieved
diff(sin(x^2), x, 5)
Partial differentiation2 of expressions of more than one variable is performed by a straightforward extension of this procedure. For example, consider the expression sin x cos y; then
its second derivatives are found as follows. First define z = sin x cos y:
syms x y
z = sin(x)*cos(y)
Then the second derivative ∂ 2 z/∂x2 is given by
diff(z, x, 2)
and ∂ 2 z/∂y 2 is found with
diff(z, y, 2)
or by nesting the diff command
diff(diff(z, y), y)
Thus one way to evaluate the mixed second derivative
∂2z
∂x∂y
is:
diff(diff(z, x), y)
This is a bit cumbersome, particularly if you want something like
can do:
∂5z
,
∂x2 ∂y 3
and instead you
diff(z, x, x, y, y, y)
For example to verify that the order of differentiation does not matter here:
diff(z, y, x)
diff(z, x, y)
Sometimes it is necessary to find the derivative of an expression at a particular point,
rather than as a function of the independent variable(s). This is achieved with a combination
of the diff, subs and sometimes double commands. Thus the value of the first derivative
of sin x at x = 6 is given by the commands
f = diff(sin(x), x);
subs(f, x, 6)
2
This topic may be new to you; it will be covered in lectures this term.
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
20
and to get a decimal approximation:
double(subs(f, x, 6))
The following exercises give practice in the diff command and also revise some earlier
commands such as solve and plot.
Exercise 2.7 Find
d
dx
(x2 + 1)4
e2x
and evaluate it at x = 1.
Exercise 2.8 Plot the second derivative of x4 /(1 + x2 ) for −5 ≤ x ≤ 5.
Exercise 2.9 Given y = 12x5 − 15x4 + 20x3 − 330x2 + 600x + 2, find the (x, y) coordinates of
any stationary points (that is, those for which y 0 (x) = 0). Use Matlab to plot y so that the
real stationary points are visible. Can you also make a plot to show the complex stationary
points?
Exercise 2.10 If z = ln(x2 + y 2 ) find the value of the constant a such that
2 2
∂z
∂z
+
= ae−z .
∂x
∂y
Exercise 2.11 Execute this code:
diff(sym(5))
diff(5)
and explain the results (hint: what does diff([1 2 5 9]) do?).
2.3.1
Differentiation of unknown functions
So far only the derivatives of expressions with an explicit form (in terms of one or more
independent variables) have been found. In this section we look at how to differentiate
arbitrary functions, say y(x) or f (x, y), without first giving them an explicit form. For
example, we might want to find dy/dx (as a function of y and x) given that x2 + y 2 = 3.
However, typing
syms x y
diff(y, x)
returns 0 because Matlab does not know that we intended y to depend on x; x and y are
free independent variables and are treated as such.
This difficulty is addressed in the Symbolic Math Toolbox by using y(x); Matlab interprets this syntax to mean that y(x) depends on x. Thus we have
clear
syms y(x)
diff(y,x)
diff(y,x,2)
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
21
Note here the syms y(x) also makes x a symbolic variable. In fact let’s take a closer look
whos
and see that y(x) is a “symfun”—a symbolic function, and x is a sym. In fact, the command
“syms y(x)” is a shortcut for:
x = sym('x')
y(x) = sym('y(x)')
Differentiation and the “symfun”
The syntax here is quite important. Note the following:
x = sym('x')
f = sym('f(x)')
% this is probably a mistake
g(x) = sym('g(x)')
Here g is a symfun but f is just the expression f (x): probably not what was intended. Note
also the difference
A = diff(g, x)
B = diff(g(x), x)
C = diff(f, x)
%D = diff(f(x), x)
whos A B C
%
%
%
%
yes, A is a symfun for the deriv
gives an expression, not a symfun
gives an expression, not a symfun
would give an error
The distinction between an expression and a symfun is usually not important but in some
cases—such as dealing with differential equations—it makes an important difference. In
summary, when it matters, the form syms f(x) is probably less error-prone.
Functions of multiple variables can also be manipulated using symfuns:
syms f(x,y)
f
diff(f, x)
diff(f, x, 2)
diff(f, x, y)
diff(f, x, x, y, y, y)
Using symfuns, the diff command knows the usual rules of differentiation, for instance
the addition, multiplication and quotient rules:
syms f(x) g(x)
diff(f + g, x)
diff(f * g, x)
diff(f / g, x)
To illustrate an application of this theory, suppose that we wish to find dy/dx (in terms of
x and y) given that x2 + y 2 = 3. To do this with Matlab, first define the given explicit
equation by typing
clear
syms y(x)
eq1 = x^2 + y(x)^2 == 3
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
22
and then differentiate this equation, assigning the output equation to a new name
eq2 = diff(eq1, x)
Let’s ask Matlab to solve the resulting expression for dy/dx:
solve(eq2, diff(y, x))
solve(eq2, diff(y(x), x))
% fail
% fail
Hmmm, we’ll have to try a bit harder:
% use subs to replace the deriv with "dydx"
syms dydx
eq3 = subs(eq2, diff(y(x), x), dydx)
% now solve for dydx
solve(eq3, dydx)
Exercise 2.12 Find the derivatives of
sin f (x), sin ef (x) ,
exp
p
1 + f (x)g(x) .
Exercise 2.13 Find dz/dp implicitly (in terms of z and p) given that
p3 + z(p)2 + 3pz(p) = 0.
Try to convince Matlab to isolate dz/dp, that is, solve for dz/dp.
2.4
Evaluating limits
This short section introduces the command limit. Its syntax is fairly self-explanatory; for
example to find
sin x
lim
x→0 x
type
syms x
limit(sin(x)/x, x, 0)
Left and right limits can also be found; for example to find
lim
x→3+
x2 − 4
x2 − 5x + 6
use the command
syms x
expr = (x^2-4)/(x^2-5*x+6);
limit(expr, x, 3, 'right')
Exercise 2.14 Use Matlab to evaluate the following limits
tan x − x
1 x
(i) lim
,
(ii) lim 1 +
.
x→∞
x→0 x − sin x
x
In each case, make a plot that clearly demonstrates that the limits are correct.
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
2.5
23
Integration
The Symbolic Math Toolbox can evaluate many integrals symbolically, and Matlab itself
can also provide numerical estimates of most definite integrals.
2.5.1
The int operator
The syntax for integrating an expression expr with respect to a variable var is int(expr,var).
2
For example, to integrate the expression xeax with respect to x type
syms x a
int(x*exp(a*x^2), x)
Note that the Symbolic Math Toolbox does not include the arbitrary constant of integration.
To evaluate the definite integral
Z
1
2
xe5x dx
0
type
int(x*exp(5*x^2), x, 0, 1);
and to evaluate
Z
0
u
√
dx
u−x
type
syms x u
int(1/sqrt(u-x), x, 0, u);
If the Symbolic Math Toolbox cannot evaluate an integral then it returns a representation of
the integral itself:
int(x^x, x)
which is a correct answer but probably not what you were hoping for!
At this point you should experiment with some integrals just to see what can be done.
Exercise 2.15 Use Matlab to integrate the following expressions with respect to x:
√
(i) ex − 1, (ii) x2 (ax + b)5/2 , (iii) sinh(6x) sinh4 (x), (iv) cosh−6 (x), (v) sin(ln(x)),
where a and b are constants.
Exercise 2.16 Use Matlab to evaluate the following expressions symbolically:
Z 1
Z ∞
Z 1
1
1
2
−1
(i)
dx, (ii)
x tan x dx, (iii)
dx.
3
(1 + x)(1 + x2 )
1/2 1 + x
0
0
From these exercises you will note that the symbolic toolbox can deal with rather nasty
integrals symbolically, and that it easily evaluates many of the integrals found in elementary
texts or standard tables.
Nevertheless one often encounters integrals where it is desirable to have Matlab evaluate
an integral numerically—for example, because the Symbolic Math Toolbox fails to integrate
or the symbolic calculation is too slow.
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
2.5.2
24
Quadrature: numerical evaluation of integrals
You may be aware that the majority of integrals cannot be evaluated symbolically. If the
symbolic toolbox fails to find a symbolic answer to a definite integral, or if you know that
the integral in question cannot be evaluated in terms of known functions, then it may be
necessary to evaluate it numerically.
R1
For example, suppose we try to integrate 0 e−t arcsin(t)dt, symbolically by typing
syms t
y = exp(-t)*asin(t);
z = int(y, t, 0, 1)
As we noted before, this returns the symbolic integral expression; the Symbolic Math Toolbox does not know the answer. Matlab has a variety of commands for doing numerical
integration. This is known as quadrature. These are independent of the symbolic toolbox so
we first convert out expression to a Matlab function and call integral(). Continuing our
example above:
ym = matlabFunction(y);
z = integral(ym, 0, 1)
The integral command has various options related to accuracy of the approximation: see
help integral.
To summarize, the int() command is part of the Symbolic Math Toolbox, and tries
to evaluate integrals symbolically; the integral() command is a Matlab command which
evaluates integrals numerically using quadrature.
Exercise 2.17
evaluate the following
numerically to six digit
Z 1Use MatlabZ to
Z 5 integrals
10
1
3
√
accuracy: (i)
ex dx, (ii)
dx, (iii)
sin ex/2 dx.
1 + x4
0
0
0
Exercise 2.18 In this exercise, we explore the function sinc x which is defined as

 sin x
x 6= 0,
x
sinc x =

1
x = 0.
(a) Assign the expression sin(x)/x to a variable and then plot it for −30 ≤ x ≤ 30.
(Note that ezplot should have no problem with plotting sin0 0 . However, you should satisfy
yourself that your graph looks correct at x = 0, given the above definition.)
(b) First, it can be shown that each maximum or minimum of the graph of sinc x corresponds to a point of intersection of the graphs of sinc x and cos x. Use Matlab to illustrate
this, by drawing the graphs of sinc x and cos x on the same plot for 0 ≤ x ≤ 10. Then use
solve to find numerical approximations for the x-coordinates of all stationary points of sinc x
for 0 < x ≤ 10, and verify that these are indeed where the two graphs that you have just
drawn meet.
(c) Use Matlab to evaluate the three integrals
Z ∞
Z ∞
Z ∞
x
x
x
(i)
sinc x dx, (ii)
sinc x sinc dx, (iii)
sinc x sinc sinc dx.
3
3
5
0
0
0
Z
What do you think the value of
0
conjecture.
5
∞Y
sinc(x/(2k + 1)) dx is? Use Matlab to verify your
k=0
CHAPTER 2. ALGEBRAIC EQUATIONS AND CALCULUS
25
Exercise 2.19 (Optional) In this exercise, we use Matlab to investigate some rational
bounds on π.
(a) Verify that
Z 1 4
x (1 − x)4
22
dx =
− π.
(2.2)
2
1+x
7
0
This is in itself an interesting result, as 22
7 is often used as a rational approximation for π.
However, this result can also be used to obtain bounds on π in the form a/b < π < c/d,
where a, b, c and d are integers.
(b) Verify by hand, no Matlab, that
1
2
Z
1
4
4
Z
x (1 − x) dx <
0
0
Z
(c) Evaluate J =
1
x4 (1 − x)4
dx <
1 + x2
Z
1
x4 (1 − x)4 dx.
0
1
x4 (1 − x)4 dx, and hence deduce that
0
1979
3959
<π<
.
630
1260
(2.3)
This gives rational bounds on π. You can now use Matlab to obtain tighter rational bounds,
in the following manner. It can be shown that the identity (2.2) can be generalized by
replacing the powers of 4 by powers of any integer multiple of 4 as follows.
Z 1 4n
x (1 − x)4n
dx = (−1)n (π − Rn ),
2(n−1)
2
(1 + x )
0 2
where n = 1, 2, 3, . . . and Rn is a rational number. (In (2.2) above, n = 1 and R1 = 22/7.)
The Symbolic Math Toolbox should be able to verify this for any given n and compute the
corresponding value of Rn . Then, following the procedure of parts (ii) and (iii) above, new
bounds on π can be found.3
(d) Use Matlab to calculate R5 .
(e) Hence find the upper and lower bounds on π for the case when n = 5. Evaluate these
both as rational numbers and in decimal form. You might need variable precision arithmetic
(Chapter B.2).
3
It can be shown that the higher the value of n, the tighter the bounds on π, but this is not considered
here.
Chapter 3
Differential equations, sums,
matrices and vectors
In this chapter you will learn about:
• solving differential equations;
• sums and products;
• the use of lists and vectors;
• 3D plotting.
3.1
Solving ordinary differential equations using dsolve
The “dsolve” command can be used to find the closed form solutions of many types of
differential equations. For example, to solve the homogeneous differential equation
d2 y
−y =0
dx2
we use the symfun class as follows:
syms y(x)
ode = diff(y,x,2) - y == 0
dsolve(ode)
Notice the way by which the symbolic toolbox gives the constants of integration. The important thing here is that y(x) was created as a symfun and that the diff(y,x) was used
to create the ODE (and not diff(y(x),x)); this is an example of the toolbox differentiating
arbitrary functions, which was discussed previously in Section 2.3.1.
Inhomogeneous equations are no different; to solve
y
dy
+ = x2
dx x
type
clear
syms y(x)
ode = diff(y,x) + y/x == x^2
dsolve(ode)
26
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 27
Sometimes Matlab does not automatically simplify the solution as much as it could—the
solution of
d3 x
d2 x
dx
−
3
+3
− x = 16e3t
3
2
dt
dt
dt
is found by typing
syms x(t)
ode = diff(x,t,3) - 3*diff(x,t,2) + 3*diff(x,t) - x == 16*exp(3*t)
soln = dsolve(ode)
simplify(soln)
To insert initial and/or boundary conditions in Matlab, proceed as in the following examples.
The solution of the boundary value problem
y 00 + 5y 0 + 6y = 0, y(0) = 0, y(1) = 3
is found as follows:
syms y(x)
DE = diff(y,x,2) + 5*diff(y,x) + 6*y == 0
soln = dsolve(DE, y(0)==0, y(1)==3)
We could then plot the solution:
ezplot(soln(x), [0 1])
ylim([0 6])
grid on
ylabel('y')
Now consider the initial value problem
y 00 = y, y(0) = 1, y 0 (0) = 0,
which could be solved as follows
syms y(t)
DE = diff(y,t,2) == y;
yp = diff(y,t);
Y = dsolve(DE, y(0)==1, yp(0)==0)
And again a plot:
Yp = diff(Y, t);
clf
ezplot(Y, [0 2])
hold on
h = ezplot(Yp, [0 2])
set(h, 'color', 'red', 'linestyle', '-.')
grid on
% note how we enter y'(x) here:
legend('y(x)', 'y''(x)')
Exercise 3.1 Use Matlab to solve the following differential equations:
(i)
d2 y
dy
+3
+ 4y = sin x,
2
dx
dx
In (ii), plot the solution.
(ii)
dy
= 3xy, y(0) = 1.
dx
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 28
Exercise 3.2 Solve the differential equation y 00 + y = 0, subject to the initial conditions
y(0) = a, y 0 (0) = b. Assign the result to a variable and then evaluate the solution at x = π.
3.2
Vectors and lists in Matlab
Matlab was originally designed to make it easy to manipulate matrices and vectors. A row
vector takes the form [expr1, expr2, expr3,..., exprn], in other words it is a sequence
of expressions enclosed in square brackets. We will often choose to think of this as a list. In
a list, order and repetition are both important.
An example of entering a list is:
L = [1, 2, 3, 4]
In fact the commas are optional:
L = [10 12 11 10 9]
Specific elements can be extracted. Here we pick out the third element:
L(3)
In a similar fashion, you can replace entries of a list:
L(3) = 42
L
There is a useful shortcut in list construction of the form “start:step:end”, for example:
L1
L2
L3
L4
=
=
=
=
1:10
3:2:13
17:-1:10
1.0:0.2:1.6
Lists can be concatenated
clear
L1 = [1 2 3];
L2 = [4 5 6 7];
L3 = [L1 L2]
Note this makes a longer list rather than making a “list of lists” (the latter can be done in
Matlab using cell arrays: see Appendix B.3). The number of elements in a list is its length:
n = length(L3)
You can make an empty list of length zero:
L = []
n = length(L)
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 29
3.2.1
Symbolic lists
We can also make lists/vectors of symbolic objects and manipulate their elements
syms x
L = [sin(x) cos(x) tan(x)]
L(1)
diff(L(1), x)
% in fact we can differentiate all elements at once:
derivs = diff(L, x)
Every element of a list must be the same type. If you mix sym and double in a list construction, everything will be cast to sym:
clear
x = sym(10); y = 11; z = 12;
L = [x y z];
a = L(1); b = L(2); c = L(3);
whos
Notice that L is a 1x3 sym and the a, b, and c are all of class sym.
Exercise 3.3 Design an experiment to check what happens if you replace one element of an
existing sym list with a double. On the other hand, what happens if you replace one element
of an existing double list with a sym?
3.2.2
Indexing in lists
Ranges of elements can be extracted using the colon notation:
L = 2:0.1:4
L(1:3)
L(2:2:12)
There is also a keyword called “end”
L(1:end-4)
L(end-3:end)
You can use one list as indices to access another list:
I = [1 5 3 2]
L(I)
% or directly
L([1 5 3 2])
This leads to a common design pattern in Matlab:
I = find(L > 3.05)
L(I)
I = find( (L >= 3.1) & (L < 3.8) )
L(I)
Here I is a list of indices satisfying the condition. You can combine the “list accessing a list”
pattern with the “end” keyword. For example, the following will “rotate” a list:
rotLeft = L([2:end-1 1])
rotRight = L([end 2:end-1])
(This is a bit of a trick—so called “syntactic sugar”—e.g., you cannot store I = [2:end-1 1].)
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 30
3.2.3
Vector operations
To do component-wise operators on a vector, we need special “dot” operators. To raise each
element to a power, use:
L = [1 2 4 8];
L3 = L .^ 3
Similarly, to multiple or divide the components of one vector by the components of a another
vector (component-wise multiplication/division), do:
M = L .* L3
Z = (L.^4) ./ L - L3
The component-wise multiplication symbol is typically pronounced “dot star”. The usual
multiplication symbol * (without the dot) means matrix multiplication in this context (see
Section 3.5).
Vectors and matrices are so fundamental in Matlab that we return to them in Sections 3.5
and 3.6 but first we look at some applications of lists, namely for set theory and performing
sums.
3.3
Sets
Matlab doesn’t have particular support for mathematical sets. (MuPAD does using its
worksheet mode, see Appendix B.4). However, using lists, we can implement some basic
ideas of set theory.
The set {1, 2, 3} could be presented by
S = [1 2 3]
Similarly the set of three polynomial expressions could be
syms x
S = [x^2 3*x 2]
Order should not be important and repetitions should be ignored in sets:
S = [1 3 4 1 2 1]
S = unique(S)
% not really a set
% much better
Given two sets A and B, Matlab can perform the usual set operations of union A ∪ B,
intersection A ∩ B and difference A \ B:
A = [1 2 3 4]
B = [2 4 6 8]
union(A, B)
intersect(A, B)
setdiff(A, B)
Exercise 3.4 Use Matlab to verify that if X = {1, 2, 3, 5}, Y = {2, 3, 4} and Z = {1, 3}
then
X − (Y ∪ Z) = (X − Y ) ∩ (X − Z) and X ∩ (Y ∪ Z) = (X ∩ Y ) ∪ (X ∩ Z).
(Of course this does not constitute a proof of the general results!)
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 31
3.4
Sums and products
In this section we look at sums and products in both numerical and symbolic forms using the
commands sum, prod, symsum and symprod.
3.4.1
Simple “numerical” sums
Suppose we want to evaluate the three sums
S1 =
10
X
r,
S2 =
r=1
100
X
2
i ,
100
X
S3 =
i=1
k=10
1
.
(k + 1)2
One approach is to construct an appropriate vector and then add up the elements:
r = 1:10;
S1 = sum(r)
For S2 :
i = 1:100;
isqr = i.^2;
sum(isqr)
% or directly:
S2 = sum(i.^2)
Note that we must use the “dot” component-wise vector operations:
k = 10:100;
S3_num = sum(1 ./ (k+1).^2)
Exercise 3.5 Use Matlab to evaluate the following
(i)
10
X
(r2 + 3r − 2),
50
X
(ii)
r=1
2i ,
(iii)
i=1
10
X
√
e−
k
,
k=1
(iv)
100
X
1
√ .
k
k=1
3.4.2
Products
Products can be
Q approached in a similar fashion using the
P prod command. If you have not
seen it before,
denotes a product in the same way as
denotes a sum. For example,
5
Y
(r + 2) = 3 × 4 × 5 × 6 × 7 = 2520.
r=1
Exercise 3.6 Use Matlab to evaluate the following:
(i)
5
Y
(r + 2),
r=1
(ii)
10
Y
r2 .
r=1
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 32
3.4.3
Symbolic manipulation of sums and products
The previous approach uses standard Matlab (without the Symbolic Math Toolbox) and
works for adding up numbers. But what about sums with variables in the limits? Or infinite
sums? For this we use the symsum command.
Let us now consider a more general case, where the upper limit is replaced by a free
variable, n. For example:
n
X
Sn =
r.
r=1
Our previous list idea won’t work here, instead:
syms r n
Sn = symsum(r, r, 1, n)
(the first r here is the summand, the second is the variable to sum over, see help symsum).
Now that we have Sn we can manipulate it in the usual ways:
subs(Sn, n, 10)
% which is the same as:
sum(1:10)
Infinite sums can be evaluated in this way too.
clear all
syms a k
S = symsum(a^k, k, 1, inf)
Why does this output look confusing? Recall that the geometric series converges only for
certain values of a. Let’s use the assumptions features of the symbolic toolbox to clear things
up a bit:
syms a b
assume(b, 'real')
assumeAlso(b >= 0)
assumeAlso(b < 1)
assumptions(b) % summarize assumptions on b
subs(S, a, b)
% replace a with b
Actually, that was “heavy handed”, how about:
% this will clear the assumptions on b
b = sym('b', 'clear')
% want the magnitude of b to be less than 1
assume(abs(b) < 1)
assumptions(b) % summarize assumptions on b
subs(S, a, b)
% replace a with b
Instead of using subs after the fact, you could also make the assumptions first:
clear all
syms a k
assume(abs(a) < 1)
S = symsum(a^k, k, 1, inf)
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 33
Exercise 3.7 Use Matlab and the Symbolic Math Toolbox to evaluate the following:
n−1
X
(i)
ak ,
(ii)
k=0
∞
X
1
.
k!
k=0
Exercise 3.8 Use Matlab and the Symbolic Math Toolbox to help verify the following:
(i)
∞
X
n=1
2
= 1,
(n + 1)(n + 2)
(ii)
∞
X
k2 + k − 1
k=0
(k + 2)!
= 0,
(iii)
∞
X
π2
1
=
k2
6
k=1
Exercise 3.9 By embedding one symsum command within another, evaluate the double sums
(i)
10 X
10
X
m
(2n + 1)e ,
(ii)
m=0 n=0
10 X
10
X
m
(2n + 1)e ,
(iii)
n=0 m=0
10 X
m
X
(2n + 1)em ,
m=0 n=0
where (i), (ii) should be approximately 4.2162e+06 and (iii) should be 3.814e+06.)
Exercise 3.10 Use symprod to evaluate the following:
10
Y
(i)
2
r ,
r=1
(ii)
7
Y
r
(1 − q ),
r=1
(iii)
∞
Y
q r , where q ∈ R.
r=1
Check that substituting q = 2 gives -78129765 in (ii). Does the Symbolic Math Toolbox give
reasonable answers in (iii)?
3.5
Arrays, matrices and vectors
We saw horizontal row vectors in Section 3.2. Recall the syntax was
h = [1 2 4 8]
The column vector a = [2, 1, 5, 9]T can be defined in Matlab by:
a = [2; 1; 5; 9]
As we saw in Section 3.2, to change the third element to the value 7, type
a(3) = 7
and then check the result by typing
a
Vectors can be defined and initialized in various ways; for example a zero column vector can
be set up with the command
c = zeros(4,1);
and then the entries may be input, for example, one at a time. Similar commands include
ones() and rand().
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 34
3.5.1
Matrix definition
Matrices—2D arrays of numbers—can be entered row by row:
A = [1 2 3; 4 5 6];
The entries of a matrix can be accessed by two indices
i = 1;
j = 1;
A(i,j)
or more directly as in
A(2,1)
Individual entries can be changed as necessary:
A(1,2) = -1;
A
Matlab does not distinguish between matrices and vectors: they are all just arrays of various
sizes. A Matlab array can contain symbolic entries but all entries of an array must be the
same class.
3.5.2
Rows, columns and submatrices
Suppose that A is a matrix. We can extract the first row of A using:
A = rand(4,5)
h = A(1, :)
and you can think of this as “first row, every column” (i.e., the first row of A). Similarly, we
get the second column of A using:
v = A(:, 2)
We can also extract a submatrix
B = A(1:2, 1:3);
More generally, the submatrix of rows i to j and columns k to l can be extracted with the
command A(i:j, k:l).
3.5.3
Operators on matrices and vectors
As you will be aware, matrix algebra can only be performed on matrices of the right shapes,
and we shall suppose that the operations defined below have a meaning; if the matrices are
the wrong shape Matlab will probably complain.
In the following table, A and B are matrices, u, v are vectors and s a scalar.
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 35
Matlab command
s*A
A*B
A*v
A+B
A^n
A.^n
A .* B
A ./ B
inv(A)
x = A \ b
A-s
det(A), trace(A)
transpose(A)
A'
u' * v
u * v'
cross(u,v)
eye(n)
zeros(n,m)
ones(n,m)
size(A)
Function
Scalar multiplication by s
Matrix multiplication of A of B
Multiplication of A by vector v
Matrix addition of A and B (component-wise)
Matrix power
Component-wise exponentiation
Component-wise multiplication (A, B same size)
Component-wise division
Matrix inverse (square matrices)
Solve systems of linear equations Ax = b
subtract scalar s from each element
determinant and trace of A
transpose of A
conjugate transpose of A
inner product (dot product) of two column vectors
outer product of two column vectors
cross product of two length-3 vectors
n × n identity matrix
n × m zero matrix
n × m matrix of ones
returns [n m] for an n × m matrix
Here are some examples:
u = rand(3,1)
v = rand(3,1)
u' * v
u * v'
The conjugate transpose and transpose() are the same thing for real matrices. Matlab
users tend to mostly use '.
In many cases, matrix operators work with symbolic matrices:
syms x
A = sym(round(10*rand(4,4)));
I = sym(eye(4,4));
B = A - x*I
inv(B)
pretty(inv(B))
The output of the above suggests that inverting matrices is rather hard. Happily, it’s not
something we have to do very often—for example, this is not how Matlab’s “\” solves
Ax = b.


2 5
1 2 3
1 3 7
Exercise 3.11 Let A =
,B=
and C =  4 6 .
4 5 6
4 −5 0
−1 0
Use Matlab to find (where possible) 3A, A − 2B, A + 2C, AC, CA, AB and AT . Verify
that (AC)T = C T AT .
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 36
3.5.4
Simultaneous equations
Exercise 3.12 Consider the set of simultaneous equations
3x + y − z = 1,
5x + y + 2z = 6,
4x − 2y − 3z = 3.
Write this as a system Ax = b, where A is a 3 × 3 matrix, x = [x, y, z]T and b is a
column vector of length 3. Use the Matlab backslash \ command to solve the system of
simultaneous equations for x.
3.5.5
Eigenvalues and eigenvectors
The eigenvalue problem for a square matrix A is to find nonzero eigenvectors x and associated
eigenvalues λ such that:
Ax = λx.
(If you haven’t seen this before, you will soon in linear algebra—skip this for now and come
back later). In Matlab:
lambda = eig(A)
[V,D] = eig(A)
The first form gives a vector of the eigenvalues and the latter gives a diagonal matrix D with
the eigenvalues on the diagonal. The columns of the matrix V contain the eigenvectors.
Exercise 3.13 (a) Use Matlab to find the eigenvalues and corresponding eigenvectors of
the matrix
2 3
.
A=
2 1
(b) Now use Matlab to find the eigenvalues and eigenvectors of
(i) A3 ,
(ii) A−1 ,
(iii) A − 6I (where I is the identity),
(iv) (A + 3I)−1 .
In each case, can you spot the relationship of the eigenvalues and eigenvectors found in (b)
to those found in (a)?
3.6
Meshgrids and 3D plots
We looked at indexing into vectors in Section 3.2.2. Understanding this enables concise code,
particularly for plotting.
3.6.1
Logical masks
Let’s reconsider an example from earlier using the find command:
L = 2:0.1:4;
% approach 1
I = find(L > 3.05);
L(I)
% approach 2
B = L > 3.05;
L(B)
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 37
The results of L(I) and L(B) are the same. But compare I and B. Note that I contains just
3
the indices of those elements which are greater than 10
. On the other hand, B contains 1’s
where the condition is true and 0’s where it is false. B is known as a “logical mask”.
Here’s an example of what can be done with this sort of masking
x = linspace(-1, 1, 512);
y = cos(2*pi*x);
y(y > 0) = 0;
y(x > 0.5) = 2;
plot(x, y, 'b.-');
Exercise 3.14 Explain what each line of the previous job accomplishes. Change “512” in
the above code to “20” and make the code explicitly display the two logical masks. If you
swap the order of the two “y(...) = ...” lines, does the picture change? Why or why not?
3.6.2
Meshgrid
We can build special 2D arrays (matrices) of x and y coordinates for use in plotting:
x = linspace(-2,2,60);
y = linspace(-1,1,40);
[xx,yy] = meshgrid(x,y);
f = sin(2*pi*xx) .* cos(pi*yy);
figure(1); clf;
surf(xx, yy, f);
axis equal
xlabel('x'); ylabel('y'); zlabel('f');
figure(2); clf;
pcolor(xx, yy, f);
axis equal; axis tight;
xlabel('x'); ylabel('y');
colorbar
Meshgrid is essentially the Cartesian product of the 1D grids in the x and y directions.
Examine the following output to understand what meshgrid does:
x = linspace(-2, 2, 5)
y = linspace(-1, 1, 4)
[xx, yy] = meshgrid(x, y)
3.6.3
3D plotting
Exercise 3.15 In the plotting code above, try adding “shading interp”, “shading flat”,
or “shading faceted” after each plot.
After the surf command, try adding these command one after another and examining
what each one does: “shading interp”, “camlight left”, “lighting phong”, “material shiny”.
Use your mouse and the rotate and zoom tools to interact with your figures.
CHAPTER 3. DIFFERENTIAL EQUATIONS, SUMS, MATRICES AND VECTORS 38
Next let’s combine meshgrids with logical masks. This gives a way to modify regions for
plotting.
x = linspace(-2,2,60);
y = linspace(-1,1,40);
[xx,yy] = meshgrid(x,y);
f = sin(2*pi*xx) .* cos(pi*yy);
I = (xx > 0) & (yy > 0);
f(I) = 0;
% or try: f(I) = f(I)+1;
figure(1); clf;
surf(xx, yy, f);
axis equal
xlabel('x'); ylabel('y'); zlabel('f');
figure(2); clf;
pcolor(xx, yy, f);
axis equal; axis tight;
xlabel('x'); ylabel('y');
colorbar
Exercise 3.16 Modify the above code so that for every point where f is negative, the value
is replaced with three times its absolute value.
Exercise 3.17 Modify the above code to plot f = 3−(x2 +y 2 ) in the interior of the following
triangle. Set f = 0 outside the triangle.
The interior of the triangle is defined by the inequalities
y < 0.25x + 0.5,
x < 1.3 + 0.15y,
and y > −0.4x − 0.5.
Also, try setting the outside values to inf and nan (not a number).
Exercise 3.18 What does this code do?
A = rand(10,10);
I = A > 0.8;
A(~I) = 0;
p
Exercise 3.19 Plot the function besselj(nu,r) where r = x2 + y 2 on a circle of radius
20 for various integer values of ν. Animate your plots over ν = 0, 1, 2, . . . using the pause
command.
Chapter 4
Loops, conditionals and functions
In this chapter you will learn about:
• using Matlab loops and conditionals;
• writing functions in Matlab;
• debugging programs.
In this chapter you will consolidate things you have learned so far and you will also be
introduced to some simple programming techniques. The material here is very important
and a lot of it will be needed in the Hilary Term projects; it is therefore crucial that you start
Hilary Term Week 3 having completed Chapter 4.
In addition to being an environment for problem solving and exploring mathematics,
Matlab is also a programming language. By this we mean that a series of command lines
can be written which, when executed, perform a particular task. You have seen several
simple programs already (and hopefully written a few too). In this chapter you will look at
programming in more detail, to develop longer and more structured command sequences. In
particular you will look at some of the main building blocks of programming, namely loops,
conditionals and functions.
4.1
Loops
Sometimes a Matlab command (or a sequence of commands) may need to be executed a
number of times. The for. . . end control structure is provided exactly for this purpose. For
example, say we wanted to output the squares of the integers from 1 to 5. This could be
achieved with the following command:
for j = 1:5
j^2
end
Essentially, any list can be provided in the “for j = ...” part, although you’ll find most
frequently it is some set of integers.
39
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
for
i
end
for
x
s
end
40
i = 1:2:7
c = [4 7 1 4]
= sym(pi)/4*j
= sin(x)
As another example, a for loop could be used (instead of sum) to find the sum of the cubes
of the even integers between 50 and 100 inclusive.
sumcubes = 0;
% set sumcubes to zero initially
for i = 50:2:100
sumcubes = sumcubes + i^3;
end
sumcubes
Note also that the command line within the loop has been slightly indented; this makes the
whole structure easier for human beings to read (Matlab doesn’t care).
Exercise 4.1 Use a for loop structure to plot cos(nπx/2) for n = 0, 1, 2, . . . 10.
Exercise 4.2 Use one or more for loops to find
N
X
(2n − 1)2 for each N = 5, 6, . . . , 30.
n=1
Another form of loop is the “while” loop. For example, the sum-of-cubes-of-even-integers
problem from above could be solved using a while loop as follows:
sumcubes = 0;
j = 50;
while (j <= 100)
sumcubes = sumcubes + j^3;
j = j+2;
end
sumcubes
Exercise 4.3 In a while loop, is the condition—in the case above “(j<=100)”—checked
before or after each iteration?
4.1.1
Approximate solutions to equations
In this section the for...end structure is illustrated by looking at Newton’s method for
finding approximate solutions to equations of the form f (x) = 0. The process depends on
the function being “approximately linear” when viewed on a small enough interval (the First
Year course should make this more precise).
Starting with a guess at the solution, a, the equation of the tangent through (a, f (a)) is
y − f (a)
= f 0 (a),
x−a
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
and this cuts the x-axis at â = a −
41
f (a)
. This is (we hope) a better approximation to the
f 0 (a)
root than a.
Say we want to find an approximate solution to
f (x) = x5 − x1.33 − 1 = 0
and we know that an initial approximation to the solution we are interested in is a = 1.5
(e.g., from a plot). Newton’s method can be used to find a better approximation with the
following commands:
clear
syms x
f(x) = x^5 - x^(sym(133)/100) - 1
fp = diff(f)
% now use the correction in double
a = double(1.5);
a = a - double( f(a) / fp(a) )
If we repeat the last command over-and-over (try pressing the up-arrow), we ought to get a
better-and-better approximation to the root. However, this is a little clumsy. Instead we use
a loop, repeating until the approximation is accurate to some required number of significant
figures:
% f and fp as above, then convert to matlab functions
f_m = matlabFunction(f)
fp_m = matlabFunction(fp)
a = 1.5;
b = inf;
tol = 10^(-8);
while ( abs(b-a) > tol )
b = a;
a = a - f_m(a) / fp_m(a);
end
a
(Recall that double variables have roughly 15 decimal digits of accuracy, so convergence
cannot be expected if tol is taken smaller than, say, 1e-14.)
Exercise 4.4 Adapt the Newton’s method example above to find a root of
f (x) = 5 −
1
cos(3x) = x,
4
with a starting initial guess of a = 5. Show that the root is 5.24979 to six significant figures.
How many iterations does it take? (Modify the code to tell you this.) Try to use solve to
check your answer.
4.2
Conditionals
The “if. . . else. . . end” construction allows us to choose alternative courses of action during
the execution of a sequence of commands. The general structure is:
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
42
if <conditional expression>
<statement sequence>
elseif <conditional expression>
<statement sequence>
else
<statement sequence>
end;
where there can be as many elseif’s as needed (or none at all) and else is also optional.
Here is a simple example:
x = -3
if (x > 2)
y = x + 1;
elif (x > 0) && (x <= 2)
y = x;
else
y = x - 1;
end
y
The <conditional expression> mentioned above is a Matlab statement which is either 0
(false) or non-zero (true). This will usually be a comparison using the equality and inequality
operators on page 14 combined with logic operators: and (“&&”, two ampersands), or (“||”,
two vertical bars), not (“~”, tilde).
4.3
Functions
Matlab functions are sets of commands which are needed to be used repeatedly, often with
different parameter values each time. Some programming languages call these “procedures”
to distinguish them from mathematical functions. At any rate, a Matlab function could be
used to define a (mathematical) function, to return a matrix, to plot a graph, or to perform a
particular calculation. For example, create an file called “f2c.m” with the following contents:
function c = f2c(x)
%F2C Convert Fahrenheit to Celsius
%
F2C(x) converts a temperature in degrees Fahrenheit to
%
degrees Celsius.
c = 5/9*(x-32);
% Typically, your function would be longer and would
% include more steps here.
end
% optional
Having written this function, you can now use it on the Matlab command prompt:
f2c(100)
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
43
to see that the answer is approximately 37.8o C. Once you have a function, it can be re-used
easily:
f2c(0)
f2c(32)
f2c(80)
help f2c
lookfor celsius
The main points to notice are:
• Each function lives in its own .m file. The file name should match the function name.
• You can call functions which are either in the current working directory or in your path.
• Functions may have one or many arguments, or none at all.
• A function can return one or more values, or none at all.
• All communication between the function and whoever called it is through the arguments
(inputs) and return values (outputs). That is, variables are “local” to the function: you
cannot access variables from wherever the function was called from.1
One of the most important reasons to write functions is to encapsulate a smaller part of a
problem: debug it, polish it, document it, and then use it as part of a larger program without
worrying about how it works. Thus with a few well-designed functions, a programmer can
avoid trying to think about her entire program at once and instead break it down in to
manageable chunks.
To help a function be re-usable, there is a specifically formatted comment which forms
the “help text” at the start of the function:
• The first line consists of the name of your function and a short summary:
%function y = myfunc(x)
%MYFUNC What is does
• The remaining lines of the comment give detailed usage notes for your function:
%function y = myfunc(x)
%MYFUNC What is does
%
A more detailed explanation, by convention intended
%
three spaces...
For shorter functions, there are also “inline functions” (also known as anonymous functions) which do not need a separate .m file. These can be created using the “@” syntax or they
can be converted from symbolic expressions. More details can be found in Appendix B.1.
Exercise 4.5 Write a Matlab function called c2f which converts temperatures in degrees
Celsius to degrees Fahrenheit. Check your function by evaluating c2f(0) and c2f(37.8). Exercise 4.6 Write a function which defines the function f given by
(
x
if x ≥ 0,
f (x) =
−x otherwise.
Check your function by finding f (2) and f (−3). Hint: Use the if...end structure within
your function.
1
Global variables are an exception. Many programmers avoid global variables whenever possible. Anyway,
type “help global” to find out more.
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
4.4
44
Examples of functions
This section presents some non-trivial examples of functions. In some cases, there may be
existing implementations in Matlab or we may not implement the “optimal” or most general
approach. In particular, experienced Matlab programmers will often try to solve problems
by manipulating lists rather than using loops. We make no such attempt in this section. At
any rate, if you think you know a better approach, implement it and try: the encapsulation
principle of functions means it is simple to “drop in” a replacement implementation.
4.4.1
Finding the arithmetic mean of a set of numbers
This section revises the material on lists covered in Section 3.2. Suppose we wanted to find
the arithmetic mean of a dataset given as a list of elements.
function s = arithmetic_mean(data)
%ARITHMETIC_MEAN Find the arithmetic mean of a list of data
n =
s =
for
s
end
s =
end
length(data);
0;
i = 1:n
= s + data(i);
s / n;
An example call of this function would be:
arithmetic_mean([1 2 3 4 5 6])
4.4.2
Taylor’s theorem
Taylor’s theorem tells us how well a function which is sufficiently
proximated by polynomials in the neighbourhood of a point. The
Symbolic Math Toolbox calculates the Taylor series of a function
of the size of the remainder after the maximum degree polynomial
Thus
differentiable can be aptaylor command in the
together with an ‘order’
terms which you specify.
syms x
taylor(cos(x), x, 0, 'order', 5)
gives the Taylor expansion of cos(x) about zero with the remainder of degree five. The
following function plots the function f and its Taylor polynomials about x = a up to and
including a maximal order for x in a specified interval.
function plot_taylor_poly(f, a, maxdeg, xlim)
%PLOT_TAYLOR_POLY Plot partial Taylor series of a function
figure(1); clf;
h = ezplot(f, xlim)
set(h, 'linewidth', 2, 'linestyle', '--')
leg{1} = char(f);
hold on
for i = 1:maxdeg
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
45
tay = taylor(f, 'ExpansionPoint', a, 'Order', i)
h = ezplot(tay, xlim);
set(h, 'color', [i/maxdeg 0 0])
leg{i+1} = ['taylor' num2str(i-1)];
end
title('partial taylor series')
legend(leg)
For example, try
syms u
f = cos(u)
plot_taylor_poly(f, 1, 4, [0 pi])
The command that builds the legend uses a cell-array (see Appendix B.3 for details).
4.4.3
Euler’s method
You have seen how dsolve can find analytic solutions to some differential equations. However,
not all differential equations have analytic solutions and in such cases approximate solutions
may be computed using numerical methods. Matlab includes the commands ode45 and
ode15s for solving initial value problems. But here we write our own basic solver using
Euler’s method.
The Mean Value Theorem, which holds for a wide class of functions, says that y(x +
h) − y(x) = hy 0 (x + θh) for some θ depending on x and h. For suitable functions we can
therefore hope that hy 0 (x) is a reasonable approximation to y(x + h) − y(x). This is the basis
of “Euler’s method”.
For an equation in the form y 0 (x) = f (x, y(x)) together with an initial condition y(a) = α,
say, we first choose a “step length” h which gives discrete values of xi = a + (i − 1)h, i =
1, 2, . . .. Then set y1 = α and compute approximations yi to y(xi ) for i = 2, 3, . . . using
yi = yi−1 + hf (xi−1 , yi−1 ).
Here is one possible implementation, eulers_method.m, of Euler’s method for y 0 = f (x, y),
with y(a) = α, for a ≤ x ≤ b = nh (n steps of step size h). The x coordinates are output in
xx and the solution sequence in yy.
function [xx,yy] = eulers_method(f, a, b, alpha, n)
%EULERS_METHOD Solve initial value problem ODEs
%
[xx,yy] = eulers_method(f, a, b, alpha, n) calculates an
%
approximate solution to solves y'(x) = f(x,y(x)) with initial
%
condition y(a) = alpha up until x = b using n steps.
%
%
Increasing n generally increases accuracy.
%
%
This version works for scalar problems but can be generalized.
h = (b - a) / n;
% stepsize
% n steps means n+1 values, allocate storage for soln
xx = linspace(a, b, n+1);
yy = zeros(size(xx));
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
46
% initial condition is first entry of solution
yy(1) = alpha;
% now loop, performing Euler updates
for i = 2:(n+1)
yy(i) = yy(i-1) + h*f(xx(i-1), yy(i-1));
end
Hence the approximate solution to
dy
= f (x, y) = −xy,
dx
y(0) = 1,
for 0 ≤ x ≤ 10 using 50 steps can be plotted as follows:
f = @(x,y) (-x*y);
x0 = 0; y0 = 1;
[x,y] = eulers_method(f, x0, 10, y0, 50);
clf;
plot(x0, y0, 'ro');
hold on;
plot(x, y, 'k.-');
For reference, here’s how to use ode45 for this problem:
[xx2,yy2] = ode45(f, [0 10], y0);
plot(xx2,yy2, 'g.-')
legend('IC', 'Euler', 'ode45')
4.4.4
Euclid’s algorithm
Euclid’s algorithm for finding the greatest common divisor of two natural numbers is essentially based on the observation that if a > b then gcd(a, b) = gcd(a − b, b). Things can be
improved by taking off as much b as possible:
function gcd = euclid(a,b)
%GCD
Find the greatest common denominator
if b == 0
gcd = a;
else
gcd = euclid(b, a - b * floor(a/b));
end
Note this function is recursive—it calls itself.
4.4.5
A simple matrix function
Recall we looked at matrices in Section 3.5. The next function creates an n × n matrix
with 0’s on the diagonal, 1’s everywhere below the diagonal and −1’s everywhere above the
diagonal.
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
47
function A = mymat(n)
%MYMAT Make a matrix
A = zeros(n,n);
for i = 1:n
for j = 1:n
if (i > j)
A(i,j) = 1;
elseif (i < j)
A(i,j) = -1;
else
A(i,j) = 0;
end
end
end
This is perhaps not the most “stylish” way to write this function. A Matlab pro might
write a different function:
function A = mymat2(n)
%MYMAT Make a matrix
A = triu(-1*ones(n),1)
+
tril(ones(n),-1);
But they both work just fine:
mymat(4)
mymat2(4)
4.5
Debugging
When you have written a program, you should not be too surprised if it does not work the
first time you try to run it. The program may either crash completely or may give the wrong
answer. If you’re really on a roll, you might take down Matlab or your operating system
too. Don’t worry though, the rest of the internet is (probably) immune to the bugs in your
Matlab code (this should not be interpreted as a challenge).
• If the program crashes: This is the easy case, because you often know where to
start. Matlab’s error message might point you to a line number. At any rate, you
should try to locate the line where it dies.
It may be helpful to remove semi-colons; the program will then print out intermediate
calculations which should make it easier to find the offending line. You should then try
to work out what is wrong. Remember that the error is very often above the line with
the apparent error.
Mixing up “=” and “==” is a very common syntax error. If you are copying code from a
hardcopy, be careful about typing the letter O instead of the number zero (don’t laugh:
it happens).
• If the program runs but gives the wrong answer: This is the harder case: it
may prove quite difficult to locate and correct the error. The first thing is to convince
yourself that your program is trying to solve the right problem. You should try to
CHAPTER 4. LOOPS, CONDITIONALS AND FUNCTIONS
48
run the program for very simple cases, and find the simplest case in which it gives the
wrong answer. Remove all code that is not used in that particular calculation; this is
relatively easy to do by inserting % to make them into comments.
If you have written functions, test them separately to make sure they behave as you
expect.
Add some “debugging output” by removing semi-colons or otherwise printing out intermediate results. Trace this output manually to find a point where the results don’t
match what you expect.
• Watch out for floating point when you were expected symbolic results (and vice
versa). “whos” is your friend.
• The Matlab graphical user interface provides various features to help you debug code
such as breakpoints. The command “keyboard” is useful for debugging functions: add
this command to your function and when Matlab gets there, it gives you access on
the command line to the variables inside you function.
4.6
Further exercise
The following is a final optional exercise on functions.
Exercise 4.7 Write a function that displays the rows of Pascal’s triangle up to and including
the nth row for a given positive integer n.
Appendix A
Simplification
A key strength of computer algebra systems like the Symbolic Math Toolbox is the ability
to manipulate mathematical expressions symbolically. The symbolic manipulation of expressions occasionally (on a bad day you might say “often”) creates results in a different form
than what the user needs. It is thus useful to know how to persuade the toolbox to change
the form of a particular expression into what is needed. This is more of an art than a science
but experience and a working knowledge of the commands and techniques introduced here
can be helpful.
A.1
Expand
The expand command does exactly what its name implies. In this section we will look at
how it works when applied to polynomials and to trigonometric functions. For example the
expansion of (x + 1)(x + z)2 is found by typing
syms x y z
expand((x+1)*(x+z)^2)
pretty(ans)
When applied to trigonometric functions expand uses the sum rules1 to remove multiple
angles, replacing them with powers:
expand(cos(2*x))
(remember that in the output cos(x)^2 means cos2 x, etc.) Other examples of the expand
command are
expand(sin(x+y))
expand(cos(2*x+y))
expand(cos(5*x))
pretty(ans)
Perhaps in the last example above we wished instead to obtain the partial expansion cos 5x =
cos 4x cos x − sin 4x sin x; we could use a little trickery:
c = expand(cos(x+y))
c2 = subs(c, y, 4*x)
Hyperbolic expressions can be expanded in much the same manner; for example try
expand(sinh(3*x))
1
cos(x ± y) = cos x cos y ∓ sin x sin y is an example of a sum rule.
49
APPENDIX A. SIMPLIFICATION
A.2
50
Factor
For polynomials the factor command is in some ways the opposite of expand, and often
does as one would expect:
syms x y z
factor(x^5 - x^4 - 7*x^3 + x^2 + 6*x)
f = x^4 + 4*x^3*y - 7*x^2*y^2 - 22*x*y^3 + 24*y^4
factor(f)
In the second example above f has been defined first; this is a useful trick with long expressions
to check that they are typed correctly before proceeding. The factor command can also be
used on rational expressions:
g = factor( (x^3 - y^3) / (x^4 - y^4) )
pretty(g)
(where the common factor (x − y) in the numerator and denominator has been cancelled).
Exercise 5.1 Expand the function f = (x + y + 1)3 and then factorise f − 1.
Exercise 5.2 Use Matlab to show that
n
X
r=1
n
X
r=1
n
X
r=1
1
r = n(n + 1),
2
1
r2 = n(n + 1)(2n + 1),
6
1
r3 = n2 (n + 1)2 .
4
(You might want to review Section 3.4.3.)
A.3
Collect
The command collect is useful when you want to collect terms in powers of a known
expression.
syms x
g = (cos(x) + 1)*(cos(x) + 2)*(cos(x) + 3)
collect(g, cos(x))
A.4
Simple and Simplify
The commands simplify() are simple() are the most general of the Symbolic Math Toolbox’s simplification commands and are usually the ones to be tried first. However, they are
not always the most appropriate commands and results can be unpredictable. simple() is
particularly interesting because it outputs the various attempts it makes to the screen.
Positive integer powers of trigonometric and hyperbolic functions are simplified by these
rules as far as possible, as illustrated by the following attempt to simplify sinh4 x − cosh4 x:
APPENDIX A. SIMPLIFICATION
51
syms x y z
f = (sinh(x))^4 - (cosh(x))^4
simplify(f)
The simple command will display its various attempts:
simple(f)
One of the most useful applications of simplify or simple is when trying to verify that
expr == expr2, where expr1 and expr2 are both expressions. Here is example, an attempt
to verify that
1 + tanh2 x
= cosh 2x :
1 − tanh2 x
expr1 = (1+tanh(x)^2) / (1-tanh(x)^2);
expr2 = cosh(2*x);
simplify(expr1 - expr2)
The following exercises will give you some practice using the various simplify commands
and also some earlier commands.
Exercise 5.3 Use simplify or Simplify to show that
if
if
1 + sin x
1 + cos x
r
1+x
y = ln
1−x
y=
if y = ln
(1 + x)1/2
(1 − x)1/3
then
then
dy
cos x + sin x + 1
=
;
dx
1 + 2 cos x + cos2 x
dy
1
=−
;
dx
−1 + x2
!
then
−5 + x
dy
=
.
dx
6(−1 + x2 )
Exercise 5.4 Consider the general cubic polynomial
1
f (x) = ax3 + bx2 + cx + d,
3
where a, b, c and d are real constants. If the stationary points of f are at x1 and x2 , use
Matlab to show that
a
f (x1 ) − f (x2 ) = − (x1 − x2 )3 .
6
Exercise 5.5 Show that the function f = 1/r, where r2 = (x − a)2 + (y − b)2 + (z − c)2 and
a, b and c are constants, is a solution of Laplace’s equation
∂2f
∂2f
∂2f
+
+
= 0.
∂x2
∂y 2
∂z 2
Appendix B
Other advanced topics
B.1
Matlab “handle” functions and anonymous functions
We have seen that functions can be created using .m files (Section 4.3). You can create
function handles using “anonymous functions”. This is most helpful for making a short
function, for example:
alpha = 0.5
y = @(x) sin(alpha*x)
y(pi)
You can also create these sorts of functions from symbolic expressions:
syms x
y = sin(x)
ym = matlabFunction(y)
See help function_handle, help anondemo, and help matlabFunction for more information.
B.2
Variable precision arithmetic
“Out-of-the-box” Matlab focuses on supporting double precision floating point (the double
class). Recall this class represents numbers using a precision of roughly 15 decimal digits.
The Symbolic Math Toolbox (based on MuPAD) adds support for symbolic expressions (using
the sym and symfun classes). The Symbolic Math Toolbox also adds support for Variable
Precision Arithmetic via the vpa class. Having floating point numbers with more than 15
digits can be a useful feature and so this chapter is essentially for your own interest or
reference.
To use variable precision arithmetic, first set the desired precision. Then use vpa() or a
quoted symbolic constructor with a decimal place.
digits(64)
a = vpa(2/3)
b = sym('2.0')/3
Note that b = sym(2.0)/3 won’t work (why not?). Probably the vpa() command is less
prone to accidental unwanted use.
It is worth noting that the clear command does not reset digits to its default value. But
clear all will.
52
APPENDIX B. OTHER ADVANCED TOPICS
53
Exercise 6.1 All rational numbers p/q, where p and q are integers (q 6= 0), have a terminating or (eventually) repeating decimal form. By increasing the precision using digits()
as appropriate, find the exact decimal form of 21/23.
Exercise 6.2 The following two expressions are both approximations to π that were discovered by the Indian mathematician Ramanujan (1887–1920):
√
√
√
12
π1 = √
ln((2 2 + 10)(3 + 10))
190
and
sr
π2 =
92 +
192
.
22
Use VPA to find the absolute errors |π1 − π| and |π2 − π|. Hence determine how good these
approximations actually are.
B.3
Cell Arrays
Recall that a vector in Matlab is homogeneous: everything inside it must be the same type
(double, sym, etc). How dull! Cell arrays on the other hand can store anything, even other
cell arrays. Make one like this:
A = {'Freedom!', sym(2/3), rand(3,3)}
A{1}
A{2}
A{3}
A common use of cell arrays is to store a “list of strings”. This is because an array of strings
just concatenates them. For example:
L
%
S
%
= ['red' 'lorry' 'yellow' 'lorry']
output is 'redlorryyellowlorry'
= {'red' 'lorry' 'yellow' 'lorry'}
output is a 4-element cell array
This is useful for constructing legends for plots when using loops.
B.4
MuPAD worksheets
In addition to the interfaces to Matlab using the Symbolic Math Toolbox, MuPAD has its
own graphical user interface and its own language. The user interface is based on a worksheet
model which is popular with many computer algebra systems. To explore MuPAD, begin by
typing
mupad()
at the Matlab prompt.
Academic Honesty
In Michaelmas Term you are encouraged to co-operate with fellow students in learning to
use the system and to use Matlab. The Hilary Term projects, however, must be your own
unaided work.
The description of each project gives references to books covering the relevant mathematics;
if you cannot understand some of this then you are free to consult your tutors or others about
it, but not about the project itself. You may discuss with the demonstrators and others the
techniques described in the Michaelmas Term Students’ Guide, and the commands listed in
the Hilary Term Students’ Guide or found in the Matlab Help pages. You may ask the
course director to clarify any obscurity in the projects.
The projects must be your own unaided work. You will be asked to make a
declaration to that effect when you submit them.
University ICT Regulations
All use of the computing and network facilities in the Oxford University Computing Services,
as well as all other computing and network facilities throughout the University of Oxford and
associated Colleges, is subject to certain rules. These rules concern what is considered to be
unacceptable behaviour and misuse, as well as what may infringe licence terms or may be
otherwise illegal. Note that all use is permitted for bona fide purposes only, and is subject
to proper authorization (which may be provided either explicitly or implicitly).
The University regards computer misuse as a serious matter which may warrant disciplinary
(or even criminal) proceedings.
The rules by which you have agreed to be bound can be found at the following webpages:
http://www.admin.ox.ac.uk/statutes/regulations/196-052.shtml
https://www.maths.ox.ac.uk/members/it/it-notices-policies
54

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project