paretosearch
Find points in Pareto set
Syntax
Description
defines a set of lower and upper bounds on the design variables in
x
= paretosearch(fun
,nvars
,A
,b
,Aeq
,beq
,lb
,ub
)x
, so that x
is always in the range
lb
≤ x
≤ ub
.
If no linear equalities exist, set Aeq = []
and beq =
[]
. If x(i)
has no lower bound, set lb(i)
= Inf
. If x(i)
has no upper bound, set
ub(i) = Inf
.
applies the nonlinear inequalities x
= paretosearch(fun
,nvars
,A
,b
,Aeq
,beq
,lb
,ub
,nonlcon
)c(x)
defined in
nonlcon
. The paretosearch
function finds
nondominated points such that c(x) ≤ 0
. If no bounds
exist, set lb = []
, ub = []
, or both.
Note
Currently, paretosearch
does not support nonlinear equality
constraints ceq(x) = 0
.
Examples
Find Pareto Front
Find points on the Pareto front of a twoobjective function of a twodimensional variable.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; rng default % For reproducibility x = paretosearch(fun,2);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
Plot the solution as a scatter plot.
plot(x(:,1),x(:,2),'m*') xlabel('x(1)') ylabel('x(2)')
Theoretically, the solution of this problem is a straight line from [2,1]
to [1,2]
. paretosearch
returns evenlyspaced points close to this line.
Create Pareto Front with Linear Constraints
Create a Pareto front for a twoobjective problem in two dimensions subject to the linear constraint x(1) + x(2) <= 1
.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; A = [1,1]; b = 1; rng default % For reproducibility x = paretosearch(fun,2,A,b);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
Plot the solution as a scatter plot.
plot(x(:,1),x(:,2),'m*') xlabel('x(1)') ylabel('x(2)')
Theoretically, the solution of this problem is a straight line from [2,1]
to [0,1]
. paretosearch
returns evenlyspaced points close to this line.
Create Pareto Front with Bounds
Create a Pareto front for a twoobjective problem in two dimensions subject to the bounds x(1) >= 0
and x(2) <= 1
.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; lb = [0,Inf]; % x(1) >= 0 ub = [Inf,1]; % x(2) <= 1 rng default % For reproducibility x = paretosearch(fun,2,[],[],[],[],lb,ub);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
Plot the solution as a scatter plot.
plot(x(:,1),x(:,2),'m*') xlabel('x(1)') ylabel('x(2)')
All of the solution points are on the constraint boundaries x(1) = 0
or x(2) = 1
.
Create Pareto Front with Nonlinear Constraints
Create a Pareto front for a twoobjective problem in two dimensions subject to bounds 1.1 <= x(i) <= 1.1
and the nonlinear constraint norm(x)^2 <= 1.2
. The nonlinear constraint function appears at the end of this example, and works if you run this example as a live script. To run this example otherwise, include the nonlinear constraint function as a file on your MATLAB® path.
To better see the effect of the nonlinear constraint, set options to use a large Pareto set size.
rng default % For reproducibility fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; lb = [1.1,1.1]; ub = [1.1,1.1]; options = optimoptions('paretosearch','ParetoSetSize',200); x = paretosearch(fun,2,[],[],[],[],lb,ub,@circlecons,options);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
Plot the solution as a scatter plot. Include a plot of the circular constraint boundary.
figure plot(x(:,1),x(:,2),'k*') xlabel('x(1)') ylabel('x(2)') hold on rectangle('Position',[1.2 1.2 2.4 2.4],'Curvature',1,'EdgeColor','r') xlim([1.2,0.5]) ylim([0.5,1.2]) axis square hold off
The solution points that have positive x(1)
values or negative x(2)
values are close to the nonlinear constraint boundary.
function [c,ceq] = circlecons(x) ceq = []; c = norm(x)^2  1.2; end
Find Pareto Front Using Options
To monitor the progress of paretosearch
, specify the 'psplotparetof'
plot function.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; options = optimoptions('paretosearch','PlotFcn','psplotparetof'); lb = [4,4]; ub = lb; x = paretosearch(fun,2,[],[],[],[],lb,ub,[],options);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
The solution looks like a quartercircular arc with radius 18, which can be shown to be the analytical solution.
Find Pareto Front in Function Space and Parameter Space
Obtain the Pareto front in both function space and parameter space by calling paretosearch
with both the x
and fval
outputs. Set options to plot the Pareto set in both function space and parameter space.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; lb = [4,4]; ub = lb; options = optimoptions('paretosearch','PlotFcn',{'psplotparetof' 'psplotparetox'}); rng default % For reproducibility [x,fval] = paretosearch(fun,2,[],[],[],[],lb,ub,[],options);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
The analytical solution in objective function space is a quartercircular arc of radius 18. In parameter space, the analytical solution is a straight line from [2,1]
to [1,2]
. The solution points are close to the analytical curves.
Monitor Pareto Set Solution
Set options to monitor the Pareto set solution process. Also, obtain more outputs from paretosearch
to enable you to understand the solution process.
options = optimoptions('paretosearch','Display','iter',... 'PlotFcn',{'psplotparetof' 'psplotparetox'}); fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; lb = [4,4]; ub = lb; rng default % For reproducibility [x,fval,exitflag,output] = paretosearch(fun,2,[],[],[],[],lb,ub,[],options);
Iter Fcount NumSolutions Spread Volume 0 60 11  3.7872e+02 1 386 12 7.6126e01 3.4654e+02 2 702 27 9.5232e01 2.9452e+02 3 1029 27 6.6332e02 2.9904e+02 4 1357 36 1.3874e01 3.0070e+02 5 1690 37 1.5379e01 3.0200e+02 6 2014 50 1.7828e01 3.0252e+02 7 2214 59 1.8536e01 3.0320e+02 8 2344 60 1.9435e01 3.0361e+02 9 2464 60 2.1055e01 3.0388e+02 Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
Examine the additional outputs.
fprintf('Exit flag %d.\n',exitflag)
Exit flag 1.
disp(output)
iterations: 10 funccount: 2464 volume: 303.6076 averagedistance: 0.0250 spread: 0.2105 maxconstraint: 0 message: 'Pareto set found that satisfies the constraints. ...' rngstate: [1x1 struct]
Obtain Pareto Front Residuals
Obtain and examine the Pareto front constraint residuals. Create a problem with the linear inequality constraint sum(x) <= 1/2
and the nonlinear inequality constraint norm(x)^2 <= 1.2
. For improved accuracy, use 200 points on the Pareto front, and a ParetoSetChangeTolerance
of 1e7
, and give the natural bounds 1.2 <= x(i) <= 1.2
.
The nonlinear constraint function appears at the end of this example, and works if you run this example as a live script. To run this example otherwise, include the nonlinear constraint function as a file on your MATLAB® path.
fun = @(x)[norm(x[1,2])^2;norm(x+[2,1])^2]; A = [1,1]; b = 1/2; lb = [1.2,1.2]; ub = lb; nonlcon = @circlecons; rng default % For reproducibility options = optimoptions('paretosearch','ParetoSetChangeTolerance',1e7,... 'PlotFcn',{'psplotparetof' 'psplotparetox'},'ParetoSetSize',200);
Call paretosearch
using all outputs.
[x,fval,exitflag,output,residuals] = paretosearch(fun,2,A,b,[],[],lb,ub,nonlcon,options);
Pareto set found that satisfies the constraints. Optimization completed because the relative change in the volume of the Pareto set is less than 'options.ParetoSetChangeTolerance' and constraints are satisfied to within 'options.ConstraintTolerance'.
The inequality constraints reduce the size of the Pareto set compared to an unconstrained set. Examine the returned residuals.
fprintf('The maximum linear inequality constraint residual is %f.\n',max(residuals.ineqlin))
The maximum linear inequality constraint residual is 0.000000.
fprintf('The maximum nonlinear inequality constraint residual is %f.\n',max(residuals.ineqnonlin))
The maximum nonlinear inequality constraint residual is 0.002142.
The maximum returned residuals are negative, meaning that all the returned points are feasible. The maximum returned residuals are close to zero, meaning that each constraint is active for some points.
function [c,ceq] = circlecons(x) ceq = []; c = norm(x)^2  1.2; end
Input Arguments
fun
— Objective functions to optimize
function handle  function name
Objective functions to optimize, specified as a function handle or function name.
fun
is a function that accepts a real row vector of
doubles x
of length nvars
and
returns a real vector F(x)
of objective function values.
For details on writing fun
, see Compute Objective Functions.
If you set the UseVectorized
option to
true
, then fun
accepts a matrix of
size n
bynvars
, where the matrix
represents n
individuals. fun
returns
a matrix of size n
bym
, where
m
is the number of objective functions. See Vectorize the Fitness Function.
Example: @(x)[sin(x),cos(x)]
Data Types: char
 function_handle
 string
nvars
— Number of variables
positive integer
Number of variables, specified as a positive integer. The solver passes row vectors of length
nvars
to fun
.
Example: 4
Data Types: double
A
— Linear inequality constraints
real matrix
Linear inequality constraints, specified as a real matrix. A
is an M
bynvars
matrix, where M
is the number of inequalities.
A
encodes the M
linear inequalities
A*x <= b
,
where x
is the column vector of nvars
variables x(:)
, and b
is a column vector with M
elements.
For example, to specify
x_{1} + 2x_{2} ≤ 10
3x_{1} + 4x_{2} ≤ 20
5x_{1} + 6x_{2} ≤ 30,
give these constraints:
A = [1,2;3,4;5,6]; b = [10;20;30];
Example: To specify that the control variables sum to 1 or less, give the constraints
A = ones(1,N)
and b = 1
.
Data Types: double
b
— Linear inequality constraints
real vector
Linear inequality constraints, specified as a real vector. b
is an M
element vector related to the A
matrix. If you pass b
as a row vector, solvers internally convert b
to the column vector b(:)
.
b
encodes the M
linear inequalities
A*x <= b
,
where x
is the column vector of N
variables x(:)
, and A
is a matrix of size M
byN
.
For example, to specify
x_{1} + 2x_{2} ≤ 10
3x_{1} + 4x_{2} ≤ 20
5x_{1} + 6x_{2} ≤ 30,
give these constraints:
A = [1,2;3,4;5,6]; b = [10;20;30];
Example: To specify that the control variables sum to 1 or less, give the constraints
A = ones(1,N)
and b = 1
.
Data Types: double
Aeq
— Linear equality constraints
real matrix
Linear equality constraints, specified as a real matrix. Aeq
is an Me
bynvars
matrix, where Me
is the number of equalities.
Aeq
encodes the Me
linear equalities
Aeq*x = beq
,
where x
is the column vector of N
variables x(:)
, and beq
is a column vector with Me
elements.
For example, to specify
x_{1}
+ 2x_{2} +
3x_{3} =
10
2x_{1}
+ 4x_{2} +
x_{3} =
20,
give these constraints:
Aeq = [1,2,3;2,4,1]; beq = [10;20];
Example: To specify that the control variables sum to 1, give the constraints Aeq =
ones(1,N)
and beq = 1
.
Data Types: double
beq
— Linear equality constraints
real vector
Linear equality constraints, specified as a real vector. beq
is an Me
element vector related to the Aeq
matrix. If you pass beq
as a row vector, solvers internally convert beq
to the column vector beq(:)
.
beq
encodes the Me
linear equalities
Aeq*x = beq
,
where x
is the column vector of N
variables x(:)
, and Aeq
is a matrix of size Meq
byN
.
For example, to specify
x_{1}
+ 2x_{2} +
3x_{3} =
10
2x_{1}
+ 4x_{2} +
x_{3} =
20,
give these constraints:
Aeq = [1,2,3;2,4,1]; beq = [10;20];
Example: To specify that the control variables sum to 1, give the constraints Aeq =
ones(1,N)
and beq = 1
.
Data Types: double
lb
— Lower bounds
[]
(default)  real vector or array
Lower bounds, specified as a real vector or array of doubles. lb
represents
the lower bounds elementwise in
lb
≤ x
≤ ub
.
Internally, paretosearch
converts an array lb
to the
vector lb(:)
.
Example: lb = [0;Inf;4]
means x(1) ≥ 0
, x(3) ≥ 4
.
Data Types: double
ub
— Upper bounds
[]
(default)  real vector or array
Upper bounds, specified as a real vector or array of doubles. ub
represents
the upper bounds elementwise in
lb
≤ x
≤ ub
.
Internally, paretosearch
converts an array ub
to the
vector ub(:)
.
Example: ub = [Inf;4;10]
means x(2) ≤ 4
, x(3) ≤ 10
.
Data Types: double
nonlcon
— Nonlinear constraints
function handle  function name
Nonlinear constraints, specified as a function handle or function name.
nonlcon
is a function that accepts a row vector
x
and returns two row vectors,
c(x)
and ceq(x)
.
c(x)
is the row vector of nonlinear inequality constraints atx
. Theparetosearch
function attempts to satisfyc(x) <= 0
for all entries ofc
.ceq(x)
must return[]
, because currentlyparetosearch
does not support nonlinear equality constraints.
If you set the UseVectorized
option to
true
, then nonlcon
accepts a
matrix of size n
bynvars
, where the
matrix represents n
individuals.
nonlcon
returns a matrix of size
n
bymc
in the first argument,
where mc
is the number of nonlinear inequality
constraints. See Vectorize the Fitness Function.
For example,
x = paretosearch(@myfun,nvars,A,b,Aeq,beq,lb,ub,@mycon)
,
where mycon
is a MATLAB^{®} function such as the following:
function [c,ceq] = mycon(x) c = ... % Compute nonlinear inequalities at x. ceq = [] % No nonlinear equalities at x.
For more information, see Nonlinear Constraints.
Data Types: char
 function_handle
 string
options
— Optimization options
output of optimoptions
 structure
Optimization options, specified as the output of optimoptions
or as a
structure.
{}
denotes the default value. See option details in
Pattern Search Options.
Options for paretosearch
Option  Description  Values 

 Tolerance on constraints. For an options structure, use  Positive scalar  
 Level of display.  'off'  'iter'  'diagnose'  {'final'} 
 Initial points for
 Matrix with 
 Maximum number of objective function evaluations. For an options structure, use  Positive integer  
 Maximum number of iterations. For an options structure, use  Positive integer  
 Total time (in seconds) allowed for optimization. For an options structure, use  Positive scalar  
 Tolerance on the mesh size. For an options structure, use
 Positive scalar  
 Minimum fraction of the pattern to poll.  Scalar from 0 through 1  
 Function that an optimization function calls at each iteration. Specify as a function handle or a cell array of function handles. For an options structure, use  
 The solver stops when the relative change in a stopping measure
over a window of iterations is less than or equal to
See Definitions for paretosearch Algorithm. The solver stops when the relative change in any
applicable measure is less than
Note Setting  Positive scalar  
 Number of points in the Pareto set.  Positive integer  
 Plots of output from the pattern search. Specify as the name of a builtin plot function, a function handle, or a cell array of names of builtin plot functions or function handles. For an options structure, use

With multiple objectives:
With a single
objective: 
 Polling strategy used in the pattern search. Note You cannot use MADS polling when the problem has linear equality constraints. 

 Compute objective and nonlinear constraint functions in parallel. See Vectorized and Parallel Options and How to Use Parallel Processing in Global Optimization Toolbox. Note You must set Beginning in R2019a, when you set the


 Specifies whether functions are vectorized. See Vectorized and Parallel Options and Vectorize the Objective and Constraint Functions. Note You must set For an options structure, use 

Example: options =
optimoptions('paretosearch','Display','none','UseParallel',true)
problem
— Problem structure
structure
Problem structure, specified as a structure with the following fields:
objective
— Objective functionnvars
— Number of variablesAineq
— Matrix for linear inequality constraintsbineq
— Vector for linear inequality constraintsAeq
— Matrix for linear equality constraintsbeq
— Vector for linear equality constraintslb
— Lower bound forx
ub
— Upper bound forx
nonlcon
— Nonlinear constraint functionsolver
—'paretosearch'
options
— Options created withoptimoptions
rngstate
— Optional field to reset the state of the random number generator
Note
All fields in problem
are required, except for
rngstate
, which is optional.
Data Types: struct
Output Arguments
x
— Pareto points
m
bynvars
array
Pareto points, returned as an m
bynvars
array, where
m
is the number of points on the Pareto front. Each row of
x
represents one point on the Pareto front.
fval
— Function values on Pareto front
m
bynf
array
Function values on the Pareto front, returned as an
m
bynf
array.
m
is the number of points on the Pareto front, and
nf
is the number of objective functions. Each row of
fval
represents the function values at one Pareto
point in x
.
exitflag
— Reason paretosearch
stopped
integer
Reason paretosearch
stopped, returned as one of the
integer values in this table.
Exit Flag  Stopping Condition 

1  One of the following conditions is met.

0  Number of iterations exceeds
options.MaxIterations , or the number
of function evaluations exceeds
options.MaxFunctionEvaluations . 
1  Optimization is stopped by an output function or plot function. 
2  Solver cannot find a point satisfying all the constraints. 
5  Optimization time exceeds
options.MaxTime . 
output
— Information about the optimization process
structure
Information about the optimization process, returned as a structure with these fields:
iterations
— Total number of iterations.funccount
— Total number of function evaluations.volume
— Hypervolume of the set formed from the Pareto points in function space. See Definitions for paretosearch Algorithm.averagedistance
— Average distance measure of the Pareto points in function space. See Definitions for paretosearch Algorithm.spread
— Average spread measure of the Pareto points. See Definitions for paretosearch Algorithm.maxconstraint
— Maximum constraint violation, if any.message
— Reason why the algorithm terminated.rngstate
— State of the MATLAB random number generator just before the algorithm starts. You can use the values inrngstate
to reproduce the output when you use a random poll method such as'MADSPositiveBasis2N'
or when you use the default quasirandom method of creating the initial population. See Reproduce Results, which discusses the identical technique forga
.
residuals
— Constraint residuals at x
structure
Constraint residuals at x
, returned as a structure
with these fields (a glossary of the field size terms and entries follows
the table).
Field Name  Field Size  Entries 

lower  m bynvars  lb – x 
upper  m bynvars  x – ub 
ineqlin  m byncon  A*x  b 
eqlin  m byncon  Aeq*x  b 
ineqnonlin  m byncon  c(x) 
m
— Number of returned pointsx
on the Pareto frontnvars
— Number of control variablesncon
— Number of constraints of the relevant type (such as number of rows ofA
or number of returned nonlinear equalities)c(x)
— Numeric values of the nonlinear constraint functions
More About
Nondominated
Nondominated points, also called noninferior points, are points for which no other point has lower values of all objective functions. In other words, for nondominated points, none of the objective function values can be improved (lowered) without raising other objective function values. See What Is Multiobjective Optimization?.
Algorithms
paretosearch
uses a pattern search to search for points on the
Pareto front. For details, see paretosearch Algorithm.
Alternative Functionality
App
The Optimize Live Editor task provides a visual interface for paretosearch
.
Extended Capabilities
Automatic Parallel Support
Accelerate code by automatically running computation in parallel using Parallel Computing Toolbox™.
To run in parallel, set the 'UseParallel'
option to true
.
options = optimoptions('
solvername
','UseParallel',true)
For more information, see How to Use Parallel Processing in Global Optimization Toolbox.
Version History
Introduced in R2018b
Open Example
You have a modified version of this example. Do you want to open this example with your edits?
MATLAB Command
You clicked a link that corresponds to this MATLAB command:
Run the command by entering it in the MATLAB Command Window. Web browsers do not support MATLAB commands.
Select a Web Site
Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .
You can also select a web site from the following list:
How to Get Best Site Performance
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.
Americas
 América Latina (Español)
 Canada (English)
 United States (English)
Europe
 Belgium (English)
 Denmark (English)
 Deutschland (Deutsch)
 España (Español)
 Finland (English)
 France (Français)
 Ireland (English)
 Italia (Italiano)
 Luxembourg (English)
 Netherlands (English)
 Norway (English)
 Österreich (Deutsch)
 Portugal (English)
 Sweden (English)
 Switzerland
 United Kingdom (English)