Minimize

<< Click to Display Table of Contents >>

Navigation:  Problem Descriptor Reference > The Sections of a Descriptor >

Minimize

Previous pageReturn to chapter overviewNext page

The MINIMIZE section is used to optimize parameters.  FlexPDE implements the Nelder-Mead "amoeba" algorithm to minimize an objective function. This is not the method of greatest speed, but it is very flexible, and allows FlexPDE to perform optimization searches in a wide range of problem environments.

 

MINIMIZE has the forms :

 

MINIMIZE  objective VS parameter ( start_value, value_range )    [ parameter2 ...]

MINIMIZE  objective VS parameter ( start_value, value_range, lower_limit, upper_limit )    [ parameter2 ... ]

 

The method creates a simplex of parameter values, each parameter assigned a random sampling of N+1 values centered about the start value with the stated range, where N is the number of parameters. The parameter values are modified until the value of the objective is minimized within the OPTERRLIM or the range of values of each parameter is bracketed to OPTERRLIM*parameter.

In the first form, the values of the parameter are allow to shift without limits on values. The second form is used to enforce an upper and lower limit to the acceptable values for the parameter. So as the optimization is taking place, the parameter will never go beyond these limits.  When multiple parameters are declared, each parameter may be either form.

Note that this feature cannot be used with the STAGED facility, since internally FlexPDE will be implementing stages in order to perform the optimization.

See also : MAXIMIZE

Examples:

Samples | Usage | Optimization | Criticality.pde

Samples | Usage | Optimization | Criticality_Size.pde

Samples | Usage | Optimization | Power_Control.pde

Samples | Usage | Optimization | Size_Control.pde