Numerical methods

Annie Fournier-Gagnoud

Phelma 2016-2017
OUTLINE
● Introduction
● Programming with Matlab
Introduction
Algorithm
Recommandations
Matlab software
● Ordinary Differential Equations ODE
● Partial Differential Equations PDE

A. Fournier-Gagnoud Numerical Methods 2016-2017 2

 Introduction
To develop a program few steps are useful:
 Choice of language,
 Design the algorithm,
 Construct the program,
 Test the program step by step (sequence by
sequence).

A. Fournier-Gagnoud Numerical Methods 2016-2017 3

OUTLINE
● Introduction
● Matlab software
Introduction
Algorithm
Recommendations
Matlab
● Ordinary Differential Equations ODE
● Partial Differential Equations PDE

A. Fournier-Gagnoud Numerical Methods 2016-2017 4

 ALGORITHM
 Algorithm = sequence of finite instructions, used
for calculation and data processing
 Algorithms are essential in computational
process
 Algorithm must be rigorously defined
 Algorithm is a precise list of precise steps
 Order of instructions must to be listed explicitly

A. Fournier-Gagnoud Numerical Methods 2016-2017 5

 ALGORITHM DESIGN
Algorithm design is a specific method to create a
mathematical process in solving problems.

In a first step we must analyze the program to

solve a given problem.

We have to identify the input data, the output

data, to choice the variables.

A. Fournier-Gagnoud Numerical Methods 2016-2017 6

 ….
Initialization of input data …
ALGORITHM DESIGN

Calculations
…..
….. Calculation 1
.….
Calculation 2

…
Saving results ….
…

Graphic representation

A. Fournier-Gagnoud Numerical Methods 2016-2017 7

ALGORITHM DESIGN

Main algorithm Function / subroutine

Main steps
of the calculation. Input parameters
Main program Output parameters
calls functions Successive Steps
or subroutines

A. Fournier-Gagnoud Numerical Methods 2016-2017 8

ALGORITHM DESIGN
Iterative algorithms: repetitive constructs like loops to solve
the given problems
An algorithm may be viewed as controlled logical deduction:
logical instruction

Serial Algorithms: computers execute one instruction of an

algorithm at a time
Parallel algorithms: take advantage of computer architectures
where several processors can work on a problem at the same
time

A. Fournier-Gagnoud Numerical Methods 2016-2017 9

OUTLINE
● Introduction
● Programming with Matlab
Introduction
Algorithm
Recommendations
Matlab software
● Ordinary Differential Equations ODE
● Partial Differential Equations PDE

A. Fournier-Gagnoud Numerical Methods 2016-2017 10

RECOMMENDATIONS
 Planning the program
a series of smaller, independent tasks
each task as a separate function
 General Coding Practices
variable names: easier to understand
< 80th column
 Importance of comments %
 Coding in Steps
 Making Modifications in Steps
 Testing the Final Program

A. Fournier-Gagnoud Numerical Methods 2016-2017 11

OUTLINE
● Introduction
● Programming with Matlab
Introduction
Algorithm
Recommendations
Matlab software
● Ordinary Differential Equations ODE
● Partial Differential Equations PDE

A. Fournier-Gagnoud Numerical Methods 2016-2017 12

MATLAB

MATrix LABoratory

Commercial software developed by the

MathWorks Inc.
Numerical calculations on vectors, matrices,
real or complex
Freeware : freemat, octave, scilab
Unix, Linux, Windows implementation
A. Fournier-Gagnoud Numerical Methods 2016-2017 13
MATLAB

Interactive software system for numerical

computations and graphics
Matlab is designed for Matrix computations:
Solving systems of linear equations
Computing eigenvalues - eigenvectors
Factoring matrices

A. Fournier-Gagnoud Numerical Methods 2016-2017 14

Beginning with Matlab

 Start : double clicking the MATLAB shortcut

 Desktop MATLAB appears, Including
 Some menus
 Help is very useful Getting Started
 Three windows
Command window : Enter Matlab statements at the prompt
Showing you what you type and the result produced
At this level = Matlab is a pocket calculator
 Files in the directory

A. Fournier-Gagnoud Numerical Methods 2016-2017 15

Beginning with Matlab

Command window

Variable Window

Click on help/MATLAB help

To reach on line help

Commands history

A. Fournier-Gagnoud Numerical Methods 2016-2017 16

MATLAB Help

Getting started
Introduction
Matrices and Array
Graphics
Programming
…….

A. Fournier-Gagnoud Numerical Methods 2016-2017 17

Matlab : variables

 Variables are not declared prior

Easy – convenient
Careful users
Good use of Malab is a difficult task.

 Basic data type : n-dimensional array of

double precision numbers
A. Fournier-Gagnoud Numerical Methods 2016-2017 18
Interactive system : vectors

>>a=[1 4 8 ]
a=
1 4 8
>>b=[3 ; 2 ]
b=
3
2
>>c=pi

A. Fournier-Gagnoud Numerical Methods 2016-2017 19

 Matrix definition

>>A=[1 1 1; 2 4 8; 3 9 27 ]
A=
1. 1. 1. Matrix definition
2. 4. 8.
Each componant is separated by space
3. 9. 27.
Each line by ;

>>A' '
1. 2. 3.
Transposed Matrix
1. 4. 9. ' is the tranposition operator
1. 8. 27.

A. Fournier-Gagnoud Numerical Methods 2016-2017 20

Calculation: vectors-matrices

>>a=[1 4 8 ]
>>b=[3 ; 2 ;5]
>>a*c
??? Undefined function or variable ‘c’
>>a*b
ans=54
>>A=b*a
A=
3 12 24  3  3
2 8 16 a  1 4 8 b  2 A  b  a  2  1 4 8
5 20 40
5 5

A. Fournier-Gagnoud Numerical Methods 2016-2017 21

Calculation:
Matrix – Vector Product
>>b=[1 0 0 ]; No echo if ;
>>A*b
??? Error using ==> * |1 0 0|
Inner matrix dimensions must agree. | 1. 1. 1. |
>>A*b' | 2. 4. 8. |
1.
2.
| 3. 9. 27. | = ?
3.

Arithmetic operators + - * /

A. Fournier-Gagnoud Numerical Methods 2016-2017 22

Linear system: solver

--> b= [1 0 0]’;
transposed

--> x = A \ b
Resolution of A*x = b
x =

3.0 Verification that it runs !

- 2.5
0.5

A. Fournier-Gagnoud Numerical Methods 2016-2017 23

Matrix and Vectors building

[] [ ]
1 1 2 3
Defining a vector V: V = −2 a Matrix A:A= 2 −3 5
3 4 −3 1
• Display the 2nd component of V
• Calculate norm of V (3 methods)
• Build the vector W such as W=A . V and G=V ^ W

•Display the 2nd line, 3rd column element of A matrix

•Display first column of A
•Display 2nd line of A

•Calculate determinant of A
•Solve A X = V (2 methods)
A. Fournier-Gagnoud Numerical Methods 2016-2017 24
Matrix and Vectors building
>> A=[1 2 3;2 -3 5;4 -3 1]
>> V=[1 -2 3]’ or V=[1;-2;3]
>> A(2,3)
>> V(2) ans =
ans = 5
-2
>> A(:,1)
>> norm(V) ans =
>> sqrt(V’*V) 1
>> sqrt(dot(V,V)) 2
ans = 4
3.7417
>> A(2,:)
ans =
2 -3 5

A. Fournier-Gagnoud Numerical Methods 2016-2017 25

Matrix and Vectors building

>> W=A*V

W=
6
23
13

>> G=cross(V, W)
-30
-6
6

A. Fournier-Gagnoud Numerical Methods 2016-2017 26

Matrix and Vectors building
Build matrix T
Calculate with Matlab the dimension of T matrix
List of Matlab functions : zeros, ones, diag, eyes, size

2 -1 0 0 .... 0
-1 2 -1 0 0
0 -1 2 -1 0
T= 0 0
0 -1
0 ... .. -1 2

A. Fournier-Gagnoud Numerical Methods 2016-2017 27

First program « Hello World ! »

Program edition : File  New  M-file hello.m

Saving : File  Save , write hello.m

Matlab code interpretation with editor

debug menu

Or by typing the name of the program file,

hello.m in the Matlab command window

A. Fournier-Gagnoud Numerical Methods 2016-2017 28

 Matlab variables

All variables in Matlab are tables (matrices),

Matlab is case sensitive
variable names : I non equal to i
Auto-declaration of variables when initialized
Example : Matlab comments

t=0; % numeric variable

s=‘Hello World’; % table of characters variable
Type Overloading of a variable is possible

A. Fournier-Gagnoud Numerical Methods 2016-2017 29

 Matlab variables

A. Fournier-Gagnoud Numerical Methods 2016-2017 30

 Error ! Ok !

Look help informations for

zeros, ones et diag

A. Fournier-Gagnoud Numerical Methods 2016-2017 31

Variable: influence zone in program

Initialization of the variable N

Influence : from the
Initialization until end of file

 No local declaration inside

! a bloc !!!

Danger : ‘:’ and , in a ‘for Loop’

have not the same effects !
for i=1, 10
i
end;

A. Fournier-Gagnoud Numerical Methods 2016-2017 32

 Binary operators

binary operators

A<B
A>B
Operators : A && B
A <= B
A || B
A >= B
Arithmetic operators A == B
Logical operators A ~= B
Relational operators

A. Fournier-Gagnoud Numerical Methods 2016-2017  33

 Control structures
compare.m

Condition :
if
Enter
true i & j

i<j ?

false Block of instructions

A. Fournier-Gagnoud Numerical Methods 2016-2017 34

 When different instructions
Conditional connection have to be executed according
tochoice2.m
a variable value
choice1.m In replacement of
‘if, elseif, else, end’
control structure

no break instruction in a
block case as in C++!
A. Fournier-Gagnoud Numerical Methods 2016-2017 35
Loops

loop1.m
loop2.m

A. Fournier-Gagnoud Numerical Methods 2016-2017 36

Loops

loop3.m sum1.m

What is calculated ?

Break : loop breaking

A. Fournier-Gagnoud Numerical Methods 2016-2017 37

Mathematical functions

|x| abs(x)
arctg x atan(x)
cos x cos(x)
exp x exp(x)
ln x log(x)
log x log10(x)

x sqrt(x)
y
C&
x x^y C++

A. Fournier-Gagnoud Numerical Methods 2016-2017 38

Mathematical functions

 : continuation following line

num2str : numeric conversion  string

i ≡ sqrt(-1.)

A. Fournier-Gagnoud Numerical Methods 2016-2017 39

Writing a function

% Solving a linear system

% comments

function [x]=resolution(A, b) Entries are separated by a comma

x = A\b;

Outputs separated by a comma

• Save in a file with “resolution.m” name for example

• In matlab, select “resolution.m” with the menu : file  open

• Call the resolution function from the Matlab command window :

x = resolution(A, b)
A. Fournier-Gagnoud Numerical Methods 2016-2017 40
 Writing a function
function linsolver linsolver is the principal function
A=[1 2 3;2 -3 5;4 -3 1]
calling the resolution function
b=[1 -2 3]'
[x, ier]=resolution(A, b)
if (ier)
error('Pb resolution');
end;
% solving a linear system Conditional instruction
function [x, res]=resolution(A, b) if (condition)
if (det(A)==0) bloc;
res=1; determinant else
x=0; bloc;
else end
res=0;
x=A\b; What happens if det(A)=0 ?
end
A. Fournier-Gagnoud Numerical Methods 2016-2017 41
Graphical functions

For x[-50,50], calculate f=x*sin(x) and draw f(x)

List of Matlab functions : plot, xlabel, ylabel, title, grid on

x = linspace(-50,50,100);
y = x*sin(x);
plot(x,y)
hold on
title('Graph of f=x*Sine')
xlabel(' x ') % x-axis label
ylabel('f(x)=x*sin(x)') % y-axis label

A. Fournier-Gagnoud Numerical Methods 2016-2017 42

 Graphical functions

For [0,2] et [0, ] calculate e=sin()^4*cos(2*)

Present e(, ) with a surface and isovalues of e(, ) as contour lines

List of Matlab functions : meshgrid, surf, contour

% Isovalues
clc
close all figure
colormap(jet)
[phi,tet]=meshgrid(0:0.02:2*pi,0:0.01:pi); [c,h]=contour(phi,tet,e,v);
e=sin(tet).^4.*cos(2*phi); set(h,'ShowText','on','TextStep',...
get(h,'LevelStep')*2);
% Fixed Isovalues
v=-1.:0.2:1; figure
colormap(jet)
colormap([0 0 0]) h=surf(phi,tet,e);
[c,h]=contour(phi,tet,e,v); shading interp
set(h,'ShowText','on','TextStep',...
get(h,'LevelStep')*2);
A. Fournier-Gagnoud Numerical Methods 2016-2017 43
Simplified Man-Machine-Interface

Example n°1

choice_list ={'banana', 'pear', 'apple', 'cherry'}

[rg, ok] = listdlg('PromptString','Select a fruit ?',...

'SelectionMode','single',...
'ListString', choice_list,...
'InitialValue',2,...
'listSize',[400 80],...
'Name','dialogue');
choice_list{rg}
ok % =1 if OK, =0 if CANCEL

A. Fournier-Gagnoud Numerical Methods 2016-2017 44

Simplified Man-Machine-interface

Example n°2

nb_lignes=1;
val_defaut={'293'};
txt = inputdlg('temperature' ,‘window_title, ...
nb_lignes, val_defaut);
T=str2double(txt) % convert a string into a real

A. Fournier-Gagnoud Numerical Methods 2016-2017 45

 Structures

Point in 2D coordinates

M.x=input('abscisse du point : ');

x M.y=input('ordonnee du point : ');

y disp(M);

A. Fournier-Gagnoud Numerical Methods 2016-2017 46

 M.x=input('abscisse du point : ');
M.y=input('ordonnee du point : ');
Structures
disp(M);

Point in 2D coordinates

Automatic declaration and

Dynamic memory allocation
of a table composed of point
structures

A. Fournier-Gagnoud Numerical Methods 2016-2017 47

 Reading a Point table in a file
Objective: filling a data structure ‘tab’ from the data file ‘tab.dat’

Number of points ‘ tab.dat’ t=

63
Coordinates_of_points NP: 63
1 -3.1415927e+000 -1.2246468e-016 x: [1x63 double]
2 -3.0415927e+000 -9.9833417e-002
y: [1x63 double]
3 -2.9415927e+000 -1.9866933e-001
4 -2.8415927e+000 -2.9552021e-001
5 -2.7415927e+000 -3.8941834e-001
6 -2.6415927e+000 -4.7942554e-001
7 -2.5415927e+000 -5.6464247e-001
8 -2.4415927e+000 -6.4421769e-001
9 -2.3415927e+000 -7.1735609e-001
…
63 3.0584073e+000 8.3089403e-002

A. Fournier-Gagnoud Numerical Methods 2016-2017 48

function [tab, err]=reading(nomfich)
clear tab;

fich = fopen(nomfich);
if ( fich == -1 )
errordlg([‘Invalid file name : ' nomfich]); % !!does not stop
err = 1;
else
% reading the comment
comment = fscanf(fich, '%s', [1])

% reading the number of points

tab.NP = fscanf(fich,'%d', [1]) % [1] only one scalar value

% reading the comment function tab_point1

comment = fscanf(fich, '%s', [1]) clc
nom='tab.dat'
% reading point coordinates [t, ier]=reading(nom);
for np=1:tab.NP if (ier==1)
n = fscanf(fich,'%d', [1]) error(‘reading error);
end;
x = fscanf(fich,'%g', [1])
y = fscanf(fich,'%g', [1])
plot(t.x, t.y)
tab.x(n)=x;
tab.y(n)=y; t=
end; NP: 63
err = 0;
end; x: [1x63 double]
A. Fournier-Gagnoud Numerical Methods 2016-2017 49
fclose(fich); y: [1x63 double]
 Reading a Point table in a file
Objective :
Filling a data structure ‘curve’ reading the data file ‘tab.dat’

curve= curve.NP = 63

NP: 63
M: [1x63 struct] curve.M =

1x63 struct array with fields:

x
y

curve.M(i) donne accès au ième point M(i) :

for ex. curve.M(1) = curve.M(1).x =
x: -3.1416
y: -1.2246e-016 -3.1416
A. Fournier-Gagnoud Numerical Methods 2016-2017 50
function [curve, err]=reading(nomfich)
clear curve;

fich = fopen(nomfich);
if ( fich == -1 )
errordlg([‘incorrect file name : ' nomfich]); % !! Does not stop
err = 1;
else
% reading a comment
comment = fscanf(fich, '%s', [1])

% reading the number of points =

curve.NP = fscanf(fich,'%d', [1]) % [1] only one scalar value

% reading a comment
comment = fscanf(fich, '%s', [1])
function tab_point2
% reading points coordinates clc
for np=1:curve.NP nom='tab.dat'
n = fscanf(fich,'%d', [1]) [curve, ier]=reading(nom);
x = fscanf(fich,'%g', [1]) if (ier==1)
y = fscanf(fich,'%g', [1]) error(‘error reading’);
curve.M(n).x=x; end;
curve.M(n).y=y;
end; for i=1:curve.NP
err = 0; x(i)=curve.M(i).x;
end; y(i)=curve.M(i).y;
fclose(fich); end;
A. Fournier-Gagnoud plot(x, y)
Numerical Methods 2016-2017 51
MATLAB
Website:
http://www.mathworks.com/academia/student_center/t
utorials/starting_matlab.html
 On this web site read the different tutorials
and stop after: Creating scripts with
MATLAB Editor/Debugger
 Two important points:
 Visualizingdata
 Creating scripts with MATLAB Editor/Debugger

A. Fournier-Gagnoud Numerical Methods 2016-2017 52

MATLAB
 Indocumentation MATLAB
 Getting started
IntroductionDocumentationProgramming
fundamental Functions and scripts
In this part the 5 sequences are very
important:
 Program Development
 Working with M-files
 M-files Scripts and Functions
 Calling Functions
 Function Arguments
A. Fournier-Gagnoud Numerical Methods 2016-2017 53