Set options for patternsearch by using
optimoptions.
options = optimoptions('patternsearch','Option1','value1','Option2','value2');
Some options are listed in
italics. These options do not appear
in the listing that optimoptions returns. To see why
'optimoptions hides these option values, see Options that optimoptions Hides.
Ensure that you pass options to the solver. Otherwise,
patternsearch uses the default option
values.
[x,fval] = patternsearch(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon,options)
PlotFcn specifies the plot function or functions called at each
iteration by patternsearch or
paretosearch. Set the PlotFcn option to be a
built-in plot function name or a handle to the plot function. You can stop the
algorithm at any time by clicking the Stop button on the plot
window. For example, to display the best function value, set
options as follows:
options = optimoptions('patternsearch','PlotFcn','psplotbestf');To display multiple plots, use a cell array of built-in plot function names or a cell array of function handles:
options = optimoptions('patternsearch','PlotFcn', {@plotfun1, @plotfun2, ...});where @plotfun1, @plotfun2, and so on are
function handles to the plot functions. If you specify more than one plot function,
all plots appear as subplots in the same window. Right-click any subplot to obtain a
larger version in a separate figure window.
Available plot functions for patternsearch or for
paretosearch with a single objective function:
'psplotbestf' — Plot the best objective function
value.
'psplotfuncount' — Plot the number of function
evaluations.
'psplotmeshsize' — Plot the mesh size.
'psplotbestx' — Plot the current best point.
'psplotmaxconstr' — Plot the maximum nonlinear
constraint violation.
You can also create and use your own plot function. Structure of the Plot Functions describes the structure of a custom plot function. Pass any custom function as a function handle.
For paretosearch with multiple objective functions, you can
select a custom function that you pass as a function handle, or any of the following
functions.
'psplotfuncount' — Plot the number of function
evaluations.
'psplotmaxconstr' — Plot the maximum nonlinear
constraint violation.
'psplotdistance' — Plot the distance metric. See paretosearch Algorithm.
'psplotparetof' — Plot the objective function values.
Applies to three or fewer objectives.
'psplotparetox' — Plot the current points in parameter
space. Applies to three or fewer dimensions.
'psplotspread' — Plot the spread metric. See paretosearch Algorithm.
'psplotvolume' — Plot the volume metric. See paretosearch Algorithm.
For patternsearch, the
PlotInterval option specifies the
number of iterations between consecutive calls to the plot function.
The first line of a plot function has the form
function stop = plotfun(optimvalues, flag)The input arguments to the function are
optimvalues — Structure containing
information about the current state of the solver. The structure
contains the following fields for
patternsearch:
x — Current point
iteration — Iteration number
fval — Objective function
value
meshsize — Current mesh size
funccount — Number of function
evaluations
method — Method used in last
iteration
TolFun — Tolerance on function value
in last iteration
TolX — Tolerance on
x value in last iteration
nonlinineq — Nonlinear inequality
constraints, displayed only when a nonlinear constraint function
is specified
nonlineq — Nonlinear equality
constraints, displayed only when a nonlinear constraint function
is specified
The structure contains the following fields for
paretosearch:
x — Current point
fval — Objective function
value
iteration — Iteration number
funccount — Number of function
evaluations
nonlinineq — Nonlinear inequality
constraints, displayed only when a nonlinear constraint function
is specified
nonlineq — Nonlinear equality
constraints, always empty []
volume — Volume measure (see Definitions for paretosearch Algorithm)
averagedistance — Distance measure
(see Definitions for paretosearch Algorithm)
spread — Spread measure (see Definitions for paretosearch Algorithm)
flag — Current state in
which the plot function is called. The possible values for flag are
'init' — Initialization state
'iter' — Iteration state
'interrupt' — Intermediate
stage
'done' — Final state
For details of flag, see Structure of the Output Function.
Passing Extra Parameters explains how to provide additional parameters to the function.
The output argument stop provides a way to
stop the algorithm at the current iteration. stop can
have the following values:
false — The algorithm continues
to the next iteration.
true — The algorithm terminates
at the current iteration.
Poll options control how the pattern search polls the mesh points at each iteration.
PollMethod specifies the pattern the algorithm uses to
create the mesh. There are two patterns for each of the classes of direct search
algorithms: the generalized pattern search (GPS) algorithm, the generating set
search (GSS) algorithm, and the mesh adaptive direct search (MADS) algorithm. These
patterns are the Positive basis 2N and the Positive basis
N+1:
The default pattern for patternsearch,
'GPSPositiveBasis2N', consists of the following
2N vectors, where N is the
number of independent variables for the objective function.
[1 0 0...0][0 1 0...0] ...[0 0 0...1][–1 0 0...0][0 –1 0...0][0 0 0...–1].
For example, if the optimization problem has three independent variables, the pattern consists of the following six vectors.
[1 0 0][0 1 0][0 0 1][–1 0 0][0 –1 0][0 0 –1].
The 'GSSPositiveBasis2N' pattern is similar to
'GPSPositiveBasis2N', but adjusts the basis vectors
to account for linear constraints. 'GSSPositiveBasis2N'
is more efficient than 'GPSPositiveBasis2N' when the
current point is near a linear constraint boundary.
The 'MADSPositiveBasis2N' pattern consists of
2N randomly generated vectors, where
N is the number of independent variables for the
objective function. This is done by randomly generating
N vectors which form a linearly independent set,
then using this first set and the negative of this set gives
2N vectors. As shown above, the
'GPSPositiveBasis2N' pattern is formed using the
positive and negative of the linearly independent identity, however, with
the 'MADSPositiveBasis2N', the pattern is generated using
a random permutation of an N-by-N
linearly independent lower triangular matrix that is regenerated at each
iteration.
The 'GPSPositiveBasisNp1' pattern consists of the
following N + 1 vectors.
[1 0 0...0][0 1 0...0] ...[0 0 0...1][–1 –1 –1...–1].
For example, if the objective function has three independent variables, the pattern consists of the following four vectors.
[1 0 0][0 1 0][0 0 1][–1 –1 –1].
The 'GSSPositiveBasisNp1' pattern is similar to
'GPSPositiveBasisNp1', but adjusts the basis vectors
to account for linear constraints. 'GSSPositiveBasisNp1'
is more efficient than 'GPSPositiveBasisNp1' when the
current point is near a linear constraint boundary.
The 'MADSPositiveBasisNp1' pattern consists of N
randomly generated vectors to form the positive basis, where
N is the number of independent variables for the
objective function. Then, one more random vector is generated, giving
N+1 randomly generated vectors. Each iteration
generates a new pattern when the 'MADSPositiveBasisNp1'
is selected.
For paretosearch only, the default
'GSSPositiveBasis2Np2' pattern uses the GSS
2N patterns, and also uses the [1 1 ...
1] and [-1 -1 ... -1] patterns.
UseCompletePoll specifies whether all the points in the current mesh must
be polled at each iteration. UseCompletePoll can have the values
true or false.
If you set UseCompletePoll to true,
the algorithm polls all the points in the mesh at each iteration and chooses
the point with the smallest objective function value as the current point at
the next iteration.
If you set UseCompletePoll to false,
the default value, the algorithm stops the poll as soon as it finds a point
whose objective function value is less than that of the current point. The
algorithm then sets that point as the current point at the next
iteration.
For paretosearch only, the
MinPollFraction option specifies the fraction of poll
directions that are investigated during a poll, instead of the binary value
of UseCompletePoll. To specify a complete poll, set
MinPollFraction to 1. To specify
that the poll stops as soon as it finds a point that improves all objective
functions, set MinPollFraction to
0.
PollOrderAlgorithm specifies the order in which the algorithm searches the
points in the current mesh. The options are
'Consecutive' (default) — The algorithm polls
the mesh points in consecutive order, that is, the
order of the pattern vectors as described in Poll Method.
'Random' — The polling order is random.
'Success' — The first search direction at each
iteration is the direction in which the algorithm found the best point at
the previous iteration. After the first point, the algorithm polls the mesh
points in the same order as 'Consecutive'.
The paretosearch solver mainly uses patternsearch
options. Several of the available built-in plot functions differ; see Plot Options. The
following options apply only to paretosearch.
In the table, N represents the number of decision
variables.
Multiobjective Pattern Search Options
| Option | Definition | Allowed and {Default}
Values |
|---|---|---|
ParetoSetSize | Number of points in the Pareto set. | Positive integer | {max(60,number of objectives)
} |
ParetoSetChangeTolerance | Tolerance on the change in volume or spread of solutions. When
either of these measures relatively changes by less than
ParetoSetChangeTolerance, the iterations end.
For details, see Stopping Conditions. | Positive scalar | {1e-4} |
MinPollFraction | Minimum fraction of the pattern to poll. | Scalar from 0 through 1 | {0} |
InitialPoints | Initial points for
If there are missing entries in the
| Matrix with |
The SearchFcn option specifies an optional search that the algorithm can
perform at each iteration prior to the polling. If the search returns a point that
improves the objective function, the algorithm uses that point at the next iteration
and omits the polling. If you select the same SearchFcn and
PollMethod, only the Poll method is used, although both are
used when the selected options differ.
The values for SearchFcn are listed below.
[], the default, specifies no search step.
Any built-in poll algorithm: 'GPSPositiveBasis2N',
'GPSPositiveBasisNp1',
'GSSPositiveBasis2N',
'GSSPositiveBasisNp1',
'MADSPositiveBasis2N', or
'MADSPositiveBasisNp1'.
'searchga' specifies a search using the genetic
algorithm. You can modify the genetic algorithm search using two additional
parameters:
options = optimoptions('patternsearch','SearchFcn',... {@searchga,iterlim,optionsGA})
iterlim — Positive integer specifying
the number of iterations of the pattern search for which the genetic
algorithm search is performed. The default for
iterlim is 1. The recommendation is not to
change this value, because performing these time-consuming searches
more than once is not likely to improve results.
optionsGA — Options for the genetic
algorithm, which you can set using
optimoptions. If you do not specify any
searchga options, then
searchga uses the same
UseParallel and
UseVectorized option settings as
patternsearch.
'searchlhs' specifies a Latin hypercube search.
patternsearch generates each point for the search
as follows. For each component, take a random permutation of the vector
[1,2,...,k] minus rand(1,k),
divided by k. (k is the number of
points.) This yields k points, with each component close
to evenly spaced. The resulting points are then scaled to fit any bounds.
Latin hypercube uses default bounds of
-1 and 1.
The way the search is performed depends on the setting for the
UseCompleteSearch option.
If you set UseCompleteSearch to
true, the algorithm polls all the points that
are randomly generated at each iteration by the Latin hypercube
search and chooses the one with the smallest objective function
value.
If you set UseCompleteSearch to
false (the default), the algorithm stops the
poll as soon as it finds one of the randomly generated points whose
objective function value is less than that of the current point, and
chooses that point for the next iteration.
You can modify the Latin hypercube search using two additional parameters:
options = optimoptions('patternsearch','SearchFcn',... {@searchlhs,iterlim,level})
iterlim — Positive integer specifying
the number of iterations of the pattern search for which the Latin
hypercube search is performed. The default for
iterlim is 1.
level — The level is
the number of points patternsearch searches, a
positive integer. The default for level is 15
times the number of dimensions.
'searchneldermead' specifies a search using
fminsearch, which uses the Nelder-Mead algorithm.
You can modify the Nelder-Mead search using two additional
parameters:
options = optimoptions('patternsearch','SearchFcn',... {@searchneldermead,iterlim,optionsNM})
iterlim — Positive integer specifying
the number of iterations of the pattern search for which the
Nelder-Mead search is performed. The default for
iterlim is 1.
optionsNM — Options for fminsearch, which
you can create using the optimset
function.
Custom enables you to write your own search
function.
options = optimoptions('patternsearch','SearchFcn',@myfun);To see a template that you can use to write your own search function, enter
edit searchfcntemplate
The following section describes the structure of the search function.
Your search function must have the following calling syntax.
function [successSearch,xBest,fBest,funccount] = ...
searchfcntemplate(fun,x,A,b,Aeq,beq,lb,ub, ...
optimValues,options)The search function has the following input arguments:
fun — Objective function
x — Current point
A,b — Linear inequality
constraints
Aeq,beq — Linear equality
constraints
lb,ub — Lower and upper
bound constraints
optimValues — Structure
that enables you to set search options. The structure contains the
following fields:
x — Current point
fval — Objective function
value at x
iteration — Current iteration
number
funccount — Counter for
user function evaluation
scale — Scale factor used
to scale the design points
problemtype — Flag passed
to the search routines, indicating whether the problem is 'unconstrained', 'boundconstraints',
or 'linearconstraints'. This field is a subproblem
type for nonlinear constrained problems.
meshsize — Current mesh
size used in search step
method — Method used in
last iteration
options — Pattern search options
The function has the following output arguments:
successSearch — A Boolean
identifier indicating whether the search is successful or not
xBest,fBest — Best point
and best function value found by search method
funccount — Number of user
function evaluation in search method
See Search and Poll for an example.
The UseCompleteSearch option applies when you set
SearchFcn to 'GPSPositiveBasis2N',
'GPSPositiveBasisNp1',
'GSSPositiveBasis2N',
'GSSPositiveBasisNp1',
'MADSPositiveBasis2N',
'MADSPositiveBasisNp1', or
'searchlhs'. UseCompleteSearch can have
the values true or false.
For search functions that are poll algorithms,
UseCompleteSearch has the same meaning as the poll option
UseCompletePoll. For the meaning of
UseCompleteSearch for Latin hypercube search, see the
'searchlhs' entry in Search Options.
Mesh options control the mesh that the pattern search uses. The following options are available.
InitialMeshSize specifies the size of the initial mesh,
which is the length of the shortest vector from the initial point to a mesh
point. InitialMeshSize must be a positive scalar. The
default is 1.0.
MaxMeshSize specifies a maximum
size for the mesh. When the maximum size is reached, the mesh size does not
increase after a successful iteration. MaxMeshSize must
be a positive scalar, and is only used when a GPS or GSS algorithm is
selected as the Poll or Search method. The default value is
Inf. MADS uses a maximum size of
1.
AccelerateMesh specifies whether,
when the mesh size is small, the MeshContractionFactor is
multiplied by 0.5 after each unsuccessful iteration.
AccelerateMesh can have the
values true (use accelerator) or false
(do not use accelerator), the default.
AccelerateMesh applies to
the GPS and GSS algorithms.
MeshRotate applies only when the
PollMethod is
'GPSPositiveBasisNp1' or
'GSSPositiveBasisNp1'. MeshRotate
= 'On' specifies that the mesh vectors are multiplied by
–1 when the mesh size is less than 1/100 of the
MeshTolerance option after an unsuccessful poll. In
other words, after the first unsuccessful poll with small mesh size, instead
of polling in directions ei (unit
positive directions) and –Σei, the
algorithm polls in directions –ei
and Σei.
MeshRotate can have the
values 'Off' or 'On' (the
default).
MeshRotate is especially
useful for discontinuous functions.
When the problem has equality constraints,
MeshRotate is
disabled.
ScaleMesh specifies whether the algorithm scales the
mesh points by carefully multiplying the pattern vectors by constants
proportional to the logarithms of the absolute values of components of the
current point (or, for unconstrained problems, of the initial point).
ScaleMesh can have the values
false or true (the default). When
the problem has equality constraints, ScaleMesh is
disabled.
MeshExpansionFactor specifies the factor by which the
mesh size is increased after a successful poll. The default value is
2.0, which means that the size of the mesh is
multiplied by 2.0 after a successful poll.
MeshExpansionFactor must be a positive scalar and is
only used when a GPS or GSS method is selected as the Poll or Search method.
MADS uses a MeshExpansionFactor of
4.0. See Mesh Expansion and Contraction for more information.
MeshContractionFactor specifies the factor by which the
mesh size is decreased after an unsuccessful poll. The default value is
0.5, which means that the size of the mesh is
multiplied by 0.5 after an unsuccessful poll.
MeshContractionFactor must be a positive scalar and
is only used when a GPS or GSS method is selected as the Poll or Search
method. MADS uses a MeshContractionFactor of
0.25. See Mesh Expansion and Contraction for more information.
For information on the meaning of penalty parameters, see Nonlinear Constraint Solver Algorithm.
InitialPenalty — Specifies
an initial value of the penalty parameter that is used by the nonlinear
constraint algorithm. InitialPenalty
must be greater than or equal to 1, and has a default of
10.
PenaltyFactor — Increases
the penalty parameter when the problem is not solved to required accuracy
and constraints are not satisfied.
PenaltyFactor must be
greater than 1, and has a default of
100.
TolBind specifies the tolerance for the distance from
the current point to the boundary of the feasible region with respect to linear
constraints. This means TolBind specifies
when a linear constraint is active. TolBind
is not a stopping criterion. Active linear constraints change the pattern of points
patternsearch uses for polling or searching. The default
value of TolBind is
1e-3.
The pattern search algorithm can keep a record of the points it has already polled, so that it does not have to poll the same point more than once. If the objective function requires a relatively long time to compute, the cache option can speed up the algorithm. The memory allocated for recording the points is called the cache. This option should only be used for deterministic objective functions, but not for stochastic ones.
Cache specifies whether a cache is used. The options
are 'On' and 'Off', the default. When you set
Cache to 'On',
the algorithm does not evaluate the objective function at any mesh points that are
within CacheTol of a point in the
cache.
CacheTol specifies how close a mesh point must be to a
point in the cache for the algorithm to omit polling it.
CacheTol must be a positive scalar.
The default value is eps.
CacheSize specifies the size of the cache.
CacheSize must be a positive scalar.
The default value is 1e4.
See Use Cache for more information.
Stopping criteria determine what causes the pattern search algorithm to stop. Pattern search uses the following criteria:
MeshTolerance specifies the minimum tolerance for mesh size. The GPS and
GSS algorithms stop if the mesh size becomes smaller than
MeshTolerance. MADS 2N stops when the mesh size becomes
smaller than MeshTolerance^2. MADS Np1 stops when the mesh size
becomes smaller than (MeshTolerance/nVar)^2, where
nVar is the number of elements of x0. The
default value of MeshTolerance is 1e-6.
MaxIterations specifies the maximum number of iterations the algorithm
performs. The algorithm stops if the number of iterations reaches
MaxIterations. The default value is 100 times the number of
independent variables.
MaxFunctionEvaluations specifies the maximum number of evaluations of the
objective function. The algorithm stops if the number of function evaluations
reaches MaxFunctionEvaluations. The default value is 2000 times
the number of independent variables.
MaxTime specifies the maximum time in seconds the pattern search algorithm
runs before stopping. This also includes any specified pause time for pattern search
algorithms.
StepTolerance specifies the minimum distance between the current points at
two consecutive iterations. Does not apply to MADS polling. After an unsuccessful
poll, the algorithm stops if the distance between two consecutive points is less
than StepTolerance and the mesh size is smaller than
StepTolerance. The default value is
1e-6.
FunctionTolerance specifies the minimum tolerance for the objective
function. Does not apply to MADS polling. After an unsuccessful poll, the algorithm
stops if the difference between the function value at the previous best point and
function value at the current best point is less than
FunctionTolerance, and if the mesh size is also smaller than
StepTolerance. The default value is
1e-6.
See Setting Solver Tolerances for an example.
ConstraintTolerance is not used as stopping criterion. It is used to
determine the feasibility with respect to nonlinear constraints. The default value
is 1e-6.
OutputFcn specifies functions that the pattern search algorithm calls at
each iteration. For an output function file myfun.m, set
options = optimoptions('patternsearch','OutputFcn',@myfun);For multiple output functions, enter a cell array of function handles:
options = optimoptions('patternsearch','OutputFcn',{@myfun1,@myfun2,...});To see a template that you can use to write your own output function, enter
edit psoutputfcntemplate
at the MATLAB® command prompt.
To pass extra parameters in the output function, use Anonymous Functions.
Your output function must have the following calling syntax:
[stop,options,optchanged] = myfun(optimvalues,options,flag)
MATLAB passes the optimvalues, state,
and flag data to your output function, and the
output function returns stop, options,
and optchanged data.
The output function has the following input arguments.
optimvalues — Structure
containing information about the current state of the solver. The
structure contains the following fields:
x — Current point
iteration — Iteration number
fval — Objective function
value at x
meshsize — Current mesh
size
funccount — Number of function
evaluations
method — Method used in
last iteration, such as 'Update multipliers' or 'Increase
penalty' for a nonlinearly constrained problem, or 'Successful
Poll', 'Refine Mesh', or 'Successful
Search', possibly with a '\Rotate' suffix,
for a problem without nonlinear constraints
TolFun — Absolute value
of change in function value in last iteration
TolX — Norm of change in x in
last iteration
nonlinineq — Nonlinear inequality
constraint function values at x, displayed only
when a nonlinear constraint function is specified
nonlineq — Nonlinear equality
constraint function values at x, displayed only
when a nonlinear constraint function is specified
options — Options
flag — Current state in
which the output function is called. The possible values for flag are
'init' — Initialization
state
'iter' — Iteration state
'interrupt' — Iteration
of a subproblem of a nonlinearly constrained problem
When flag is 'interrupt',
the values of optimvalues fields apply to the subproblem
iterations.
When flag is 'interrupt', patternsearch does
not accept changes in options, and ignores optchanged.
'done' — Final state
Passing Extra Parameters explains how to provide additional parameters to the output function.
The output function returns the following arguments to patternsearch:
stop — Provides a way to
stop the algorithm at the current iteration. stop can
have the following values.
false — The algorithm continues
to the next iteration.
true — The algorithm terminates
at the current iteration.
options — patternsearch options.
optchanged — Boolean flag
indicating changes to options. To change options for
subsequent iterations, set optchanged to true.
Display specifies how much information is displayed at the command line
while the pattern search is running. The available options are
'final' (default) — The reason for stopping is
displayed.
'off' or the equivalent 'none'
— No output is displayed.
'iter' — Information is displayed for each
iteration.
'diagnose' — Information is displayed for each
iteration. In addition, the diagnostic lists some problem information and
the options that are changed from the defaults.
Both 'iter' and 'diagnose' display the following
information:
Iter — Iteration number
FunEval — Cumulative number
of function evaluations
MeshSize — Current mesh
size
FunVal — Objective function
value of the current point
Method — Outcome of the
current poll (with no nonlinear constraint function specified). With
a nonlinear constraint function, Method displays
the update method used after a subproblem is solved.
Max Constraint — Maximum
nonlinear constraint violation (displayed only when a nonlinear constraint
function has been specified)
You can choose to have your objective and constraint functions evaluated in serial, parallel,
or in a vectorized fashion. Set the UseVectorized or
UseParallel options to true to use
vectorized or parallel computation.
Note
You must set UseCompletePoll to true for
patternsearch to use vectorized or parallel polling.
Similarly, set UseCompleteSearch to true
for vectorized or parallel searching.
Beginning in R2019a, when you set the
UseParallel option to true,
patternsearch internally overrides the
UseCompletePoll setting to true so it polls in
parallel.
When UseVectorized is false,
patternsearch calls the objective function on one
point at a time as it loops through the mesh points. (This assumes
UseParallel is at its default value of
false.)
UseVectorized is true,
patternsearch calls the objective function on all
the points in the mesh at once, i.e., in a single call to the objective
function.
If there are nonlinear constraints, the objective function and the nonlinear constraints all need to be vectorized in order for the algorithm to compute in a vectorized manner.
For details and an example, see Vectorize the Objective and Constraint Functions.
When UseParallel is true,
patternsearch calls the objective function in
parallel, using the parallel environment you established (see How to Use Parallel Processing in Global Optimization Toolbox). At the command line, set
'UseParallel' to false to compute
serially.
Note
You cannot simultaneously use vectorized and parallel computations. If you set
UseParallel to true and
UseVectorized to true,
patternsearch evaluates your objective and constraint
functions in a vectorized manner, not in parallel.
How Objective and Constraint Functions Are Evaluated
Assume UseCompletePoll = true | UseVectorized = false | UseVectorized = true |
|---|---|---|
UseParallel = false | Serial | Vectorized |
UseParallel = true | Parallel | Vectorized |
Option Availability Table for All Algorithms
| Option | Description | Algorithm Availability |
|---|---|---|
AccelerateMesh | Accelerate mesh size contraction. | GPS and GSS |
Cache | With | All |
CacheSize | Size of the cache, in number of points. | All |
CacheTol | Positive scalar specifying how close the current mesh
point must be to a point in the cache in order for | All |
ConstraintTolerance | Tolerance on nonlinear constraints. | All |
Display | Level of display to Command Window. | All |
FunctionTolerance | Tolerance on function value. | All |
InitialMeshSize | Initial mesh size used in pattern search algorithms. | All |
InitialPenalty | Initial value of the penalty parameter. | All |
MaxFunctionEvaluations | Maximum number of objective function evaluations. | All |
MaxIterations | Maximum number of iterations. | All |
MaxMeshSize | Maximum mesh size used in a poll/search step. | GPS and GSS |
MaxTime | Total time (in seconds) allowed for optimization. Also includes any specified pause time for pattern search algorithms. | All |
MeshContractionFactor | Mesh contraction factor, used when iteration is unsuccessful. | GPS and GSS |
MeshExpansionFactor | Mesh expansion factor, expands mesh when iteration is successful. | GPS and GSS |
MeshRotate | Rotate the pattern before declaring a point to be optimum. | GPS Np1 and GSS Np1 |
MeshTolerance | Tolerance on mesh size. | All |
OutputFcn | User-specified function that a pattern search calls at each iteration. | All |
PenaltyFactor | Penalty update parameter. | All |
PlotFcn | Specifies function to plot at run time. | All |
PlotInterval | Specifies that plot functions will be called at every interval. | All |
PollOrderAlgorithm | Order in which search directions are polled. | GPS and GSS |
PollMethod | Polling strategy used in pattern search. | All |
ScaleMesh | Automatic scaling of variables. | All |
SearchFcn | Specifies search method used in pattern search. | All |
StepTolerance | Tolerance on independent variable. | All |
TolBind | Binding tolerance used to determine if linear constraint is active. | All |
UseCompletePoll | Complete poll around current iterate. Evaluate all the points in a poll step. | All |
UseCompleteSearch | Complete search around current iterate when the search method is a poll method. Evaluate all the points in a search step. | All |
UseParallel | When | All |
UseVectorized | Specifies whether objective and constraint functions are vectorized. | All |