Introduction to Optimization in MATLAB
Introduction to MATLAB for Economics
Introduction to Optimization in MATLAB
Javier Barbero
Universidad Autónoma de Madrid
http://www.javierbarbero.net
October 31, 2014
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 1 / 17
Introduction to Optimization in MATLAB Introduction
Optimization in MATLAB
MATLAB can solve two types of optimization problems:
Zero finding: Find x such f (x) = 0.
Minimization: min f (x)
A maximization problem can be solved by minimizing the negative of
the function: max f (x) = min −f (x).
Zero finding of functions of one variable and minimization can be
solved with basic MATLAB.
For finding the zeros of functions of several variables and
minimization with constraints a toolbox is needed. The official
Optimization Toolbox provides this functionality.
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 2 / 17
Introduction to Optimization in MATLAB Zero Finding
Zero finding
The function fzero is used to find the zero of a 1-D function.
The objective function can be a function in a separate file or an
anonymous function.
For the solver to work, you must supply an initial guess of where the
zero is or an interval in which the zero is located.
If a function has several zeros, the result is conditioned by the initial
guess.
The function returns the value of x where the zero is located. The
value y = f (x) is returned as the second argument.
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 3 / 17
Introduction to Optimization in MATLAB Zero Finding
Zero Finding. Example
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 4 / 17
Introduction to Optimization in MATLAB Zero Finding
Zero Finding. Example
% Define the objective function as an anonymous
function
objFun = @( x ) sin ( x ) + cos ( x ) ;
% Plot the objective function
ezplot ( objFun , [ -3 ,3]) ;
% Optimize it with zero as initial guess
[x , val ] = fzero ( objFun , 0)
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 5 / 17
Introduction to Optimization in MATLAB Zero Finding
Zero Finding. Example
% Find the zero at the left
>> [x , val ] = fzero ( objFun , 0)
x =
-0.7854
val =
-1.1102 e -16
% Find the zero at the right
>> [x , val ] = fzero ( objFun , [0 3])
x =
2.3562
val =
-5.5511 e -16
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 6 / 17
Introduction to Optimization in MATLAB Minimization
Minimization
Minimization in one dimension is performed with the fminbnd
functions.
The function searches a minimum between a given interval (bound).
Minimization of functions of several variables are performed with the
fminsearch function.
For the function, each variable is an element of a vector x.
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 7 / 17
Introduction to Optimization in MATLAB Minimization
Minimization. Example 1-D
% Define the objective function as an anonymous
function
objFun = @( x ) sin ( x ) + cos ( x ) ;
% Plot the objective function
ezplot ( objFun , [ -3 ,3]) ;
% Search for a minimum between -3 and 0
[x , val ] = fminbnd ( objFun , -3 , 0)
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 8 / 17
Introduction to Optimization in MATLAB Minimization
Minimization. Example 1-D
% Search for a minimum between -3 and 0
>> [x , val ] = fminbnd ( objFun , -3 , 0)
x =
-2.3562
val =
-1.4142
% Find for a maximun between -1 and 2
>> [x , val ] = fminbnd (@( x ) - objFun ( x ) , -1 , 2)
x =
0.7854
val =
-1.4142
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 9 / 17
Introduction to Optimization in MATLAB Minimization
Zero Finding. Example 2-D
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 10 / 17
Introduction to Optimization in MATLAB Minimization
Minimization. Example 2-D
% Define the objective function as an anonymous
function
objFun = @( x ) sin ( x (1) ) * cos ( x (2) ) ;
objFunPlot = @( x , y ) sin ( x ) .* cos ( y ) ;
% Plot the objective function
ezsurf ( objFunPlot , [ -3 ,3]) ;
% Search for a minimum between -3 and 0
[x , val ] = fminsearch ( objFun , [0 ,0])
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 11 / 17
Introduction to Optimization in MATLAB Minimization
Minimization. Example 2-D
% Search for a minimum with guess (0 ,0)
>> [x , val ] = fminsearch ( objFun , [0 ,0])
x =
-1.5708 0.0000
val =
-1.0000
% Search for a maximum with guess (0 ,0)
>> [x , val ] = fminsearch (@( x ) - objFun ( x ) , [0 ,0])
x =
1.5708 -0.0000
val =
-1.0000
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 12 / 17
Introduction to Optimization in MATLAB Configuring the Optimizer
Configuring the Optimizer
The optimizer functions can be configured by passing them an
options structure.
The most important ones are:
Display: to display output at each iteration, only the final output, or
only if there is a problem.
Tolerance both of the X and the Function Value.
Maximum number of iterations.
etc.
Options structures are created with the optimset command.
The full list of options is available at:
http://www.mathworks.com/help/matlab/ref/optimset.html
or typing doc optimset
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 13 / 17
Introduction to Optimization in MATLAB Configuring the Optimizer
Configuring the Optimizer. Example
% Define the objective function as an anonymous
function
objFun = @( x ) sin ( x (1) ) * cos ( x (2) ) ;
% Configrue the optimizer
% Display results at each iteration
options = optimset ( ' Display ' , ' iter ' ) ;
% Search for a minimum between -3 and 0
[x , val ] = fminsearch ( objFun , [0 ,0] , options )
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 14 / 17
Introduction to Optimization in MATLAB Configuring the Optimizer
Configuring the Optimizer. Display Options
options = optimset ( ' Display ' , value )
Value Displays
’off’ Nothing
’notify’ A message only if the function fails
’final’ Only the final message
’iter’ Output at each iteration
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 15 / 17
Introduction to Optimization in MATLAB Steps in optimization
Steps in optimization
1 Look at your problem: It is a zero finding problem or a minimization
one? Does the problem have constraints?
2 Choose the adequate optimizer for the problem you have.
3 Configure the optimizer.
4 Define an initial guess for the solution.
5 Optimize it!.
6 Check if results are correct or if they look strange.
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 16 / 17
Introduction to Optimization in MATLAB References
References
Hanselman, D. and Littlefield, B. (2012).
Mastering MATLAB.
Prentice Hall.
The MathWorks, Inc. (2012).
MATLAB — The Language of Technical Computing , Version R2012b
(8.0).
Natick, Massachusetts.
Javier Barbero (UAM) Introduction to MATLAB for Economics October 31, 2014 17 / 17
